OReilly Media, Inc.

2/26/2015

Preface

Design & Data: A Perfect Synergy

All design is informed by data. Sometimes the data is based in the

experiences and observations of the designer, its craft knowledge.
When design innovations yield great benefits to users and to companies,
the practice of applying such craft knowledge derived from years of
informal observation has fueled a mystique of the genius designer.
However, we live in data rich times. With the advent of large scale data
collection and with the increasing fascination with measures and metrics,
such craft knowledge can appear ungrounded. An unintended side effect
for some has been the assumption that design as a practice is never an
empirical discipline, that design is never grounded in data, and that design
is practice is opposed to data science. Some believe that large scale
experiments can be run to gather data from millions of users to answer all
design questions that such analytics can replace design. Some believe that
any and all questions regarding, for example, font types, colours and sizes,
questions such as Should we have a blue or a red dialog box?, Do
people engage more with a list or a carousel? or Does a wizard help with
new user on-boarding flow? are not the purview of designers but can be
more effectively answered by engineering and data science. Let the crowd
speak with their clicks, and what emerges will necessarily be the best
design.

1
OReilly Media, Inc. 2/26/2015

[Insert diagram illustrating the chicken and egg problem, eternally shifting
venn diagrams, laddering the order of data and design]

We have a different view. Rather than an either/or, a divide between

analytics and design insight, we believe designers have a significant role
to play in the design of data gathering activities (what should be captured
and why), in the design of data itself (what format should data take, what
data types need to be triangulated) and in the presentation and
communication of insights and results that derive from the data capture
and analysis. We believe that data capture, data validation, data curation
and management, and data analysis for insights are central to design
practice. We believe that design practice is central to effective data design.
Therefore, in this book we ask: What is the relationship between data,
design and ultimately the user?
We concede that, on the surface, data practice and design practice might
seem incompatible. Much design practice centralises the creation of
empathy with users focus on creating artful solutions for them, addressing
their needs by creating and crafting effective products. For many
designers, its emotive, its a non-linear exploratory journey, it cannot be
articulated as a set of procedures or steps to be followed, it cannot be
rationalized and constrained, its human, its a fluid practice. Design
practice results in products that are experienced in ways that cannot be

2
OReilly Media, Inc. 2/26/2015

reduced to numbers, cannot be captured with one size fits all measures and
metrics. By contrast, the common belief is that data is rational, is about
measurements and numbers that are irrefutable. Data is black and white.
Data reveals the truth, it is not a fluid craft but a proceduralized scientific
endeavor where rigor leads to irrefutable results, to certainty. Data science
leads to truth, and is trustworthy. For some designers most measures do
not capture the essence of the users experience and for some data scientists
the only way to understand experience is to measure activity.
Both of these stories are, of course, myths. In reality, data sciences and
design practices are working toward the same goal: understanding users
and creating positive engagements with users. They are better together, in
dialogue. Designers know that design practice has always been about
different forms of data. Designers also know that we are in ever-changing
marketplace and industry. This means that new technologies, new
applications and therefore new behaviours are constantly emerging. We
are designing in a dynamic environment where we need to quickly adapt
and learn what works well or doesnt work for our users. Data can play a
big role in helping us learn and respond as quickly as possible. By
harnessing and leveraging the power of data at scale, there are new ways
to understand people, to understand users.
Designers also know that too often data gathered at scale does not
accurately represent the human experience. One response to this is to
reject and/or shy away from engaging with data at scale. But this is a two-
way conversation: design can bring deeper meaning to data. Data at all
scales, systematically collected, analyzed, communicated and leveraged
can empower design. Designers have to engage in the practice and
business of designing data, that is being part of the conversation about
what data should be collected, when and most importantly why.

Who this is for

As we wrote this book for you, we imagined a few things. That you felt a
strong sense of responsibility towards representing user needs. Because of
this, you see yourself as an advocate for users and want to ensure that
youre representing them accurately and appropriately. By being
accountable to your users, youre probably also eager to find ways to
reflect them in business goals, measures and metrics. Ideally, you want
data to be disciplined and in turn, you want to be disciplined about how
you use data. We hope that the audience for this book is familiar with the
concepts of iteration and continuous learning and that you want to bring
that perspective to the design of data gathering and analysis and the
practice of design itself.

3
OReilly Media, Inc. 2/26/2015

For Designers
We are writing for designers who are largely unfamiliar with leveraging
data or have very limited exposure to it in their day-to-day work. Perhaps
you started your career with a strong art background (as opposed to
engineering) or while you have done some qualitative user research, you
have only worked at companies that havent had access to rich data.
The target audience for this book is not designers who have already gone
through a corporate boot camp on data or regularly program APIs and
consider Python a good friend of yours. While we have many friends who
are like you, we feel like you are probably already familiar with much of
what we will cover, so this book will be of limited use to you.
For User Researchers
We are also writing this book for user researchers who are focused on
peoples everyday experience. Youre already empathetic and a humanist,
with a desire to truly understand how people feel. Youre probably
looking to get a stronger grasp of data because you want to learn more
about how rich data at scale can augment your job.
Were not writing this book for user researchers who are very familiar
with statistical analysis, lab studies and rich survey analysis. Again, like
so many of our designer friends, this book is will cover some of the basics
of what you already know.
For Product Managers and Developers
If you are interested in the blending of design and data, then this book
could help you gain insight into how designers are starting to approach
their job by incorporating data. We know that great product is only built
when you have product, technology and design working together hand in
hand, so were of course excited to have as many people from other
disciplines engage with this book as well.

All this being said, even if you are in our not column, you might be
interested because you get the other side of the coin on how people view
you.

Ultimately, this book is written for people involved in launching digital

products - designers and product managers - who have little to no
experience with taking a data aware approach to product development.
You might be part of a small start up with just a few folks working on
building your product or you might be part of a team in a larger company

4
OReilly Media, Inc. 2/26/2015

and looking to bring some of the data methodology to your group. You
might work as an in-house designer or you may be working in an agency
with clients. So long as you have an interest in understanding how
blending data and design can help you solve problems for your product,
then we believe that you will find this book useful.

Scope
Although this book is titled Designing with Data, we are not going to
talk very much about design principles or patterns and were not going to
teach you how to design. We are assuming that you have a basic
understanding of design thinking and best practices, and working with
partners in product management and technology to build your products.
Based on our experience, the biggest adjustments for a designer or
organization looking to leverage data is that they first need to get a solid
understanding of data types and how they work together. Very rarely does
it require a significant change to the fundamentals of how they design.
Our aim in this book is to help you understand the basics of designing with
data, recognize the value of incorporating data into your workflow and
avoid some of the pitfalls as you do. Although there are many types of
data that you might use as a designer, we are focusing this book on large
scale experiments and AB testing. Were doing this specifically because
we have found this to be where analytics and design have converged the
least, but where we think that some of the greatest gains can be made. We
will, however, provide a broad overview of other kinds of data in our
appendix as a list of resources and recommendations that you can look
into further on your own if you like.
At the end of this book, you should understand at a basic level the kinds of
large scale data you can gather and when youll want to use various
methods. You should get a sense of the relationship between different
data gathering methods, and have a deeper sense about methods like AB
testing which will include how to structure tests and what variables to
consider as well as how to interpret your results. You probably already
have a high-level an understanding of quantitative and qualitative data
engagements, but you want to know more about the former. We aim to
give some perspectives on how to incorporate a data aware approach into
your organization, the long term benefits of it and how to consider or
avoid some of the pitfalls as well. Throughout the book well share our
experiences and the experiences of others, suggest ways that your team
can change the way you work, and get the most benefit from your data.

5
OReilly Media, Inc. 2/26/2015

We want you to become more effective at what and how you design, and a
better advocate for your work through data. We want you to get the best
feedback from data to help you evaluate whether or not you are
succeeding at what you set out to achieve. We want you to be able to
argue your case with confidence about what should be measured, when
and how to best illustrate, test and further your design intention.
Ultimately, we would love to make you feel excited and eager to embark
on a data aware approach within your company or organization.
Remember, this is a nascent field and you can be a central part of how data
and design come together in the future.

What we believe
We started to write this book because we wanted to encourage designers to
influence and change the conversation around measures and metrics that
purport to reflect the value of a product or service.
Between the two of us, we have managed and driven user-centered design
and evaluation processes in the Internet industry, we both care deeply
about understanding the ways in which people interface with, interact with
and derive value from technology. We also feel strongly that carefully
gathered and analysed data can and does help develop that understanding
but that we need to broaden the conversation. Ultimately though, we hope
that by sharing our enthusiasm and passion for deeply integrating different
forms of systematically and programmatically gathered data into the
design process we can persuade others who have a human-centered
approach to design to be more centrally part of the data design process.
We believe this will lead to better products and better business. We
believe data and design need to be seen as two sides of the same coin. We
believe that designers, data scientists, developers and business leaders
need to work together to determine what data to collect, when and why
and how to manage, curate, summarize and communicate with and
through data.

We also believe that data is by definition broad. We take into account

everything from qualitative to quantitative data, bearing in mind that all
qualitative data can be distilled into quantitative measures and all
quantitative data are the result of qualitative judgments about what is
measureable and what should or should not be measured. You may be
gathering qualitative feedback by talking with your customers and users in
a usability session about how effective or simple your product seems to
them. Or you might be capturing quantitative data through tracking that
youve instrumented in your product to measure exactly what actions your
customers are doing in your product. If you only consider one form of

6
OReilly Media, Inc. 2/26/2015

data, for example click data, you are not getting a full picture of your
customers experience. Therefore we believe that the definition of data
needs to be broad and constantly reviewed for completeness and
addressing the questions posed.
In our daily work, we have heard concerns about incorporating data into
the design process. Some of the most often discussed concerns about
having a data orientation to design are:
It enslaves you believing in data as way to evaluate designs takes
the power away from design as a practice
Having a data orientation diminishes design as a practice, ie
experience and instinct arent valued
Having a data orientation stifles creativity and removes the art
from the design process
Focusing on data as a measure of user experience de-humanizes
the design process, reducing experience and design assessment to
just numbers
Having a data orientation drives incrementalism, leads us to always
get stuck in the minutia and focus on tweaking small variations in
the design
While each of these concerns has some grain of truthwe have
experienced some circumstances and work situations where these
criticisms heldwe believe that this was the case because designers were
not part of the data collection and analysis conversation. That design intent
and design evaluation was poorly matched to the data capture and analysis
because designers with a desire to understand user experience were not in
effective dialogue with data scientists and machine learning experts. By
developing an awareness of and an affinity for data, such conversations
will be possible and will be beneficial for both disciplines. Design practice
can be enhanced, managed well, data science can demonstrate the worth
of creativity in design rather than stifle it.
For us, the strongest benefits of taking a data aware approach are that it
can help you:
create a direct feedback loop with your users elevating the way
you understand and think about customer behavior, ultimately
helping you to hone your instinct about users over time
create a stronger bond between the user needs and how your
business measures success, helping to align a cross-functional team

7
OReilly Media, Inc. 2/26/2015

create a rigorous approach which can help to eliminate hierarchy

and rank from the process of decision making, allowing you to
focus on your users needs
We believe that a data aware approach gives you a framework to have
discussions around what your product needs to be or could be for a better
business and user value overall result. Carefully captured, curated and
analysed data can help you with future product innovations. We hope that
by the end of this book you will feel like you are not just leveraging data,
but also designing a new approach to crafting products.

Why us/About us
We are excited to write this book because we are passionate about
empowering designers to be able to formulate, inform and evaluate their
approach to product design with data. Both of us have had a lot of
experience working with data in both qualitative and quantitative
capacities. We believe that our backgrounds were natural compliments to
each other and as representatives of both design and user research in the
companies weve worked in weve had a first row seat at witnessing the
evolution of how designers have changed their relationship with data over
the past 15 years.

A word from Elizabeth

My original training as an undergraduate was in experimental
psychology. Therefore, carefully constructed questions that are
programmatically addressed using the scientific method of data collection
and analysis is baked into my way of thinking. I believe systematic data
collection can help us understand people, their traits and motivations, their
experiences and behaviors.
In this book we focus on user-testing at scale using logged behavioral data
from online activity. My desire to write this book stems from a frustration
I have with the separation of design practice and data science. As someone
concerned with users and their experiences, it is incomprehensible to me
that the two disciplines that are most responsible for users experiences are
not in a more productive dialogue.
More specifically, while I am a strong advocate of data collection and data
analysis to address questions regarding human behavior and to determine
the effectiveness of design options, I have too often experienced the
powerful tools of data science being used to address inappropriate

8
OReilly Media, Inc. 2/26/2015

questions and poorly applied. While data science has become a well-
regarded tool for companies, the art and science of asking the right
questions (posing hypotheses), designing an effective study to address
those questions, including designing the right measures to determine
answers to the questions has not been equally emphasized. I have seen
designers shy away from the evaluation process which evaluates their
carefully crafted designs. I have read too many reports, sat in too many
meetings and reviewed too many slide decks that contain summaries of
user-behavior that at best make no sense and at worst reveal nothing about
human behavior but everything about the biases and lack of training of the
data science.
On the other side, I have seen many designers whose intuitions are
excellent, whose skills are evident in the products they create, but who shy
away from or actively resist answering the questions What can we expect
to see in terms of changes in user behavior from this roll-out? What should
we look for as indicating people are learning the product? What kinds of
error do you think people will make and what behavior should we
instrument for, should we measure, to see if our hunch and our hypotheses
are correct?. Some designers dont believe that is part of their job. They
are happy to hand off to analysts and take on their next assignment. This
book is not for them. Some designers do want to be part of that
conversation but dont feel empowered. I am writing this book to appeal to
them. I am part of this project because I believe they are the best user
advocates and because I believe that all questions about all products
should start with the value to the user first. Given most businesses that
offer discretionary products live and die by their user base, in these
competitive markets, it only makes sense to develop a deeper connection
between design, user experience and analytics. These together will provide
the foundations for a responsive, customer centric business.
A word from Rochelle
I remember the first time I started to take a more data driven approach to
design. It was in 2001. I was at a small startup and someone had
mentioned how Amazon was applying AB testing as a technique to make
decisions about their user experience. In AB testing you were exposing
your users to several different kinds of experiences and then measuring
which one performed the best. Our start-up was always great about
reviewing the business metrics on a weekly basis and we were keen to get
our hands on as much data as possible. However with the introduction of
AB testing, we really got our heads around using data in an even more
effective and sophisticated way than we previously had. During the next
few years, our start-up took a very DIY and self-taught approach to

9
OReilly Media, Inc. 2/26/2015

figuring the ins and outs of being more data driven. We studied and
learned as much as we could about what other data driven companies like
Netflix and Amazon were doing and tried to apply those learnings in
practice.
Over the years, Ive gotten more exposure to companies that really
excelled in gathering data and information from their users - my company
was acquired by Intuit which had a well established and widely respected
approach towards user research and then later I joined Netflix which is one
of the most disciplined technology companies at using data to make
decisions. Having witnessed world class data driven environments has
made me appreciate how incredible a tool it can be in helping to transform
the way that an entire product organization works. Ive learned that there
are many nuances to combining data and design, and that while there are
many benefits, there are also many pitfalls as well. My goal is to help
more designers appreciate and take advantage of the benefits while
avoiding some of the pitfalls that I, and others, have made in the past.
Im writing this book, because I hope that I can share my
enthusiasm and passion for working in a data driven environment with
other designers and product managers and that I can help you elevate the
way in which you work with data.

How to Use this Book

Chapters 1-3 will give you a foundation for working with data. This not
only includes some basic instruction on the different kinds of data and
terminology, but will also cover how data relates to your business. We
will discuss the importance of grounded question asking and creating
hypotheses, of determining what critical mechanisms or levers for human
behavior, about designing and developing for meaningful measures and
metrics, and in general applying design thinking to the design of data
capture, management and interpretation. In Chapters 3, 4 and 5 we look at
more in-depth case studies to illustrate how to think about data capture,
analysis and communication strategies during early, mid and late design
evaluation. The last part of the book looks at some pitfalls and highlights
some cautionary tales around data at scale. Here, we want you to be able
to think critically and sharpen your data design lens. We highlight areas of
concern both internally (how you work with your data) as well as
externally (what does it mean to collect large amounts of data on your
customers). We will conclude with some summary overview.

10
OReilly Media, Inc. 2/26/2015

Chapter by Chapter Overview

Weve divided the book into eight chapters and give a brief summary of
each one below.

Chapter 1 Why Does Data Matter?

This chapter is an introduction to data. We talk about the kind of data that
you have access to as a designer, how different roles in a company interact
with data and cover some of our basic terminology. Well also share two
personal case studies with you to illustrate the importance of data, design
and business.

Chapter 2 Business Matters

Here we cover how businesses can be transformed as a result of data. We

talk about the key concepts of the customer funnel and what kind of data
you might measure at each stage. We also try to help you identify what
kind of business you are and therefore how to determine which kind of
data might be most important for you at this stage.

Chapter 3 Foundations and Basics

In this chapter well give you the necessary foundation around data. We
talk in more depth about data types and how you collect them. We
balance the pros and cons of qualitative and quantitative data and
introduce the experimental methodology which is necessary to make sure
you are using your data to its fullest.

Chapter 4 Putting it to Use

Here we take the ideas from the previous chapters and put it into practice.
Well use an in-depth case study or two to show how to walk through a
design process from creating a hypothesis to analysis of results and roll
out.

Chapter 5 Culture and Communication

In this chapter we focus on communicating what youve learned to a

broader audience. We also write about how to drive a culture focused on
learning and making data a core part of that learning. We also talk about
some of the softer side of culture including the kinds of people that work
well in that environment and the longer-term benefits of using data in the
design process.

11
OReilly Media, Inc. 2/26/2015

Chapter 6 Pitfalls and Cautions

In this chapter, we cover the most common pitfalls that people can fall into
when incorporating data into their design practice. We talk about both
easy mistakes that people or teams can make and some of the common
misconceptions about the power of data.

Chapter 7 The Ethics of Data

Here we acknowledge and examine some of the ethics around using data.
As designers we want to make sure that we are always advocating for the
user, and that also means being considerate about how we use their data
and build trust with them.

Chapter 8 Conclusions

We will summarize and consolidate the thoughts and focus of the previous
chapters. Well leave you with some of the key highlights and some open
areas for you to explore further.

How to read this book

If you dont know much about designing with data, wed suggest that you
start by reading the first 5 chapters in order. After that, you can jump to
the relevant parts or things youre interested in in the second part of the
book, referring back to the basics and how to sections as needed.
No matter how experienced you are with incorporating qualitative and
quantitative data into your design process, we always find actual examples
and interviews to be a great way to learn more about how the theories are
being applied in reality. Vignettes are scattered throughout the book and
aimed at showing you working examples of how data helped to facilitate
design decisions. You can see the hypothesis that they were looking to
prove and then see how they were articulated in the product or in the
design and then see the results and what decisions were made. Remember
that the results of each case study may not be applicable to you - you may
have a different audience, product or constraints. The takeaway should be
what the framework is and how you can apply it to your own situation or
product. Weve tried to pick case studies that illustrate key points, so
youre also welcome to just scan the case studies as a way to get an
overview of the themes covered in this book as well.
Its always fun for us to hear other people talk about how they think about
the balance between data and design. We love to hear perspectives on how

12
OReilly Media, Inc. 2/26/2015

data and analytics are changing our field. Weve also included a few
sidebars where weve asked people who are involved in the industry and
familiar with the topic of designing with data to share their thoughts on
how they leverage this data in their day-to-day work. These sidebars are
also distributed throughout the book and can be read ad hoc on their own.
We hope this book will give you what you need to start out on this journey
and to build a shared understanding with your peers of your product, based
on objective feedback direct from your customers.

13
OReilly Media, Inc. 2/26/2015

Why Does Data Matter?

Introduction
In this chapter we want to give you a high level introduction into how bringing data into
the design process can transform your business. To do this, well give some historical
examples of the critical role that data played in illuminating core problems which will
hopefully show you how powerful data can be. Well also cover how digital interfaces
have fundamentally changed our ability to use data and some of the emerging tools and
services which allows even the smallest companies to gather their own data at scale.
Because we are using this chapter to set the foundation for the rest of the book, we will
also use it to provide some basic terminology and touch on the different philosophical
approaches to data and design. Ideally wed like to take these data terms and transform
them into design terms which you will feel comfortable with.

This chapter is much more about giving you a basic understanding of the relationship
between data, business and design, rather than teaching you how to design or making you
an expert statistician or data scientist. Wed like to show you how data and design are just
tools that you use to build really great experiences for your users. If you are building
great experiences for your users, then you have a great foundation for your business. We
will feel successful if we can get you to leave this chapter convinced that understanding
the data that you can gather about your users will make your design and therefore your
business better.

The Power of Digital Interfaces

It might feel like using data is big news now, but the truth is that weve been using data
for a long time already. For the past 20 years, weve been moving and replicating more
and more experiences that we used to have in the physical world into the digital world.
Sharing photos, having conversations, duties that we used to perform in our daily work
have all become digital. We could probably have a separate discussion as to how much
the migration from the physical real world to the digital world has benefitted or been

1
OReilly Media, Inc. 2/26/2015

detrimental to our society, but you cant deny that its happening and only continues to
accelerate at a breakneck pace.

Lets take a look at what it means for these experiences to be moving from the physical to
the digital. Not too long ago, the primary way that you shared photos with someone was
that you would have to have used your camera to take a photo at an event. When your
roll was done, youd take that film to the local store where you would drop it off for
processing. A few days or a week later you would need to pick up your developed photos
and that would be the first time youd be able to evaluate how well the photos that you
took many days prior actually turned out. Then, maybe when someone was at your
house, youd pull out those photos and narrate what each photo was about. If you were
going to really share those photos with someone else, youd maybe order duplicates and
then put them in an envelope to mail to them and a few days later, your friend would
get your photos as well. If you were working at a company like Kodak that had a vested
interest in getting people to use your film, processing paper or cameras more, then there
are so many steps and parts of the experience that I just described which are completely
out of your control. You also have almost no way to collect insight into your customers
behaviors and actions along the process.

Now lets take the same example of sharing a photo in the digital world. Your user will
take out their phone and take a photo. They may open up your Instagram, apply some
filters to the photo and edit it on the spot before adding a caption and then sharing it.
They might also choose to share it on different channels, like Twitter or via email. The
entire experience of sharing a photo has been collapsed and condensed into one
uninterrupted flow and a single screen, one that you can hold in the palm of your hand.
And because all of this is digital, data is continuously being collected along the way.
You have access to all kinds of information that you wouldnt have had before. Location,
time spent in each step, which filters were tried but not used, what was written about the
photo and to whom the photo was sent. You can also gather consumption data on the
photo, how many people viewed it or liked it? Not only are you able to gather that
information on just one user, but you can gather it for each and every single user. And
that data is both precise as well as dynamic so you could get an instant understanding of
how your customers behaviours and interactions might be changing and evolving with
respect to your product and in reaction to changes you make to your product.

All this data can be really powerful, and BECAUSE digital interfaces have made data
collection so easy we have to make sure that we dont fool ourselves into thinking that
data interpretation is easier than it actually is. There is a danger that the ease in gathering
data also makes it easer to make bigger mistakes with that data. It becomes our
responsibility to make sure we use that data responsibly. That we are clear and careful
about how we use data. We are just seeing the beginning of this time where data at scale
is just as accessible to small startups as it is to well established large companies and the
future holds so much promise for what designers will be able to do with access to all this
information.

2
OReilly Media, Inc. 2/26/2015

Commoditization of data
We are seeing that data is quickly becoming commoditized with companies and services
springing up to help you with your data. There are so many tools and services available
now that can really help you gather data about your customers that there is almost no
excuse to not be leveraging data more in your design and product development cycle.
Were seeing this commoditization of data happening in all aspects,

Companies like usertesting.com and others are making it easier to get qualitative data
even if you dont have dedicated user research facilities, in fact you could argue that
services like this can be stronger than traditional labs because the virtual nature of the
service allows you to gather feedback from customers around the world. Optimizely, and
other companies are making quantitative data collection easier by allowing companies to
run quick and easy AB tests on their websites as well. We see data boot camps springing
up and there are also companies where you can outsource your data analysis to as well
so that you dont need to bear the expense of hiring and keeping data analysts on staff.

Finally we see more and more places where there are new data degrees emerging.

Quantitative data at scale

We thought it would be fun to include a little bit of history in this chapter, that by taking
a step back, way back, it might help to give our readers some perspective into how the
smart use of data has been something we have been doing for a very long time. As
designers, we pride ourselves on being excellent at solving problems, seeing how others
have used data to illuminate and solve the problems they face in other industries can be
quite enlightening.

The 15th century marked the beginning of the Age of Discovery, when Europeans
embarked on expeditions to explore the world. However, on especially long trips where
they could not store fruits and vegetables, scurvy a serious disease, was a significant
problem. In May 1747, aboard the British Navy ship HMS Salisbury, naval surgeon
James Lind conducted an experiment to identify a cure for scurvy. He chose six pairs of
seamen suffering from the disease and gave a different remedy to each pair, in addition
to their normal rations. Five of these pairs of sailors showed no significant improvement,
but one pair, who had been prescribed oranges and lemons, quickly showed signs of
recovering from the disease.

On the 20th of May, 1747, I took twelve patients in the scurvy, on

board the Salisbury at sea. Their cases were as similar as I could have them.
They all in general had putrid gums, the spots and lassitude, with weakness of
the knees. They lay together in one place, being a proper apartment for the sick
in the fore-hold and had one diet common to all two of these were ordered
each a quarter of cyder a day. Two others took twenty-five gutts of elixir vitriol
three times a day Two others took two spoonfuls of vinegar three times a day
Two of the worst patients, were put under a course of sea-water. Two others

3
OReilly Media, Inc. 2/26/2015

each had two oranges and one lemon given them every day. The consequence
was, that the most sudden and visible good effects were perceived from the use
of the oranges and lemons. 1
By this simple experiment, Lind managed to demonstrate that oranges and lemons were a
more effective cure for scurvy than any of the other known remedies. Eventually, the
Navy started giving citrus fruit to all sailors on long voyages, to protect against the
disease.
In another historical example, Dr. John Snow was working with the Reverend Henry
Whitehead in the midst of the 19th century cholera outbreak in London. As described by
Steven Johnson in his book The Ghost Map, the two men worked together to take a
truly multidisciplinary approach. John Snows scientific understanding behind the
transmission of the disease was powerfully coupled with Rev. Henry Whiteheads human
understanding of the local community and their behaviors to help uncover the source and
ultimately stop the spread of the disease.
The link between British scientists and how we use data in design today might not seem
immediately obvious, but Lind is credited with not only conducting the first ever clinical
trial, but the first controlled experiment on multiple groups, where all factors remained
the same aside from a single variable. As for Dr. Snow and Rev. Whitehead, you can see
here how qualitative data was gathered and used to both provide more insight into
uncovering what was going on in the communities and how that qualitative data collected
at scale was able to bring insight and clarity to a phenomenon that was initially
confounding.
The history of the modern data scientist is much more recent. 2 It was in the 1960s that
statisticians like John Tukey started to think more about what it was to bring a scientific
approach to data analysis. In the 1970s we see these analysts start to recognize that the
power they can bring to the data is to fill it with insight and more information. As we
jump ahead to today, there is no question that data science is a term which now captures
all the work that is done to both capture, measure and to interpret the vast amount of data
that represents our users on a daily basis.
We now take it for granted that in medicine, new treatments will be fully researched and
rigorously evaluated against other options, before being adopted. We expect the same
level of rigour in the design and engineering of safety-critical systems like aircraft,
automobiles or nuclear power stations. But in the design of consumer-facing and
recreational software and web sites, where human lives are rarely at stake, the pursuit of
the best possible designs through a similar approach of testing various options using a
scientifically structured methodology is a relatively new phenomenon. Based on the sheer
volume of articles, publications, talks that we see generated about data and business, data
and technology, data and marketing its clear that data is a very hot topic and the
currency of the day. What we want to do with this book is to take the term data science
and expand it even further to move beyond the realm of people who consider themselves
statisticians and to something that designers will start to embrace as part of their skillset
as well.

1
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1081662/pdf/medhist001130029.pdf
2
http://www.forbes.com/sites/gilpress/2013/05/28/a-very-short-history-of-data-science/

4
OReilly Media, Inc. 2/26/2015

Data roles
As you start to incorporate the usage of data in your day to day, youll find that in
addition to navigating the data itself, youll also need to navigate the various people who
work with data in different ways. In some cases there might be just one person who
represents each of the roles we describe, in other cases there might be an entire team
based around a particular role and in some cases you will find one person might play
multiple roles. We break these roles up into two key groups; those that help to create and
capture the data (producers) and those who use the data and information generated from it
in their work (consumers).

Data can be generated in so many ways through tracking of user behaviors in your
product, or through interviews with customers or by soliciting user feedback through
surveys. The people we often find associated with the generative side of data are; data
analysts, data scientists, user researchers, designers and marketers.

Data analysts and scientists 3 should be involved throughout the lifecycle of a product.
They help to take the mass of data that is collected from a product and then help to clean,
interpret, transform, model and validate that data. All of this work is done with the intent
of helping the business to make better decisions. Data analysts and scientists often bring
insight which can help the business predict where it needs to go, but they can also help to
analyze the resulting data from a business decision to understand if the business
accomplished what it set out to do. Common background for data analysts and scientists
will include statistics, information management , technology and business intelligence.

User researchers 4 are complimentary to data analysts and scientists and in some cases
will overlap with them in terms of skills and interests especially as user researchers
start to do more of their work on a bigger and bigger scale. Typically user researchers are
championing the user by seeking to understand who your users are and what they want.
They are interested in both attitudinal and behavioral information about your users. They
will tend to do this by focusing on qualitative information gathering via interviews,
surveys, diary studies and other forms of ethnographic research. Common backgrounds
for people involved in user research might include psychology, cognitive science, as well
as the standard backgrounds that you often find for designers.

Designers, as weve been insisting throughout this book need to also be concerned with
the generation of data. Its not new for designers to seek information about our users or
to gather information about our designs as we evaluate them. In some companies that
dont have the room for dedicated user researchers or analysts the designer will have to
step into those roles occasionally to get the information that they need to generate their

3
http://shop.oreilly.com/product/0636920028529.do
4
http://www.fastcodesign.com/3032719/ui-ux-who-does-what-a-designers-guide-to-the-tech-
industry

5
OReilly Media, Inc. 2/26/2015

designs. Designers can also play a key role in the generation of data based on what they
design and choose to prototype. Designers can just as easily come from a tech or art
background. Human factors or HCI are also fairly common as well.

Marketers can also be great collaborators in terms of data generation and too often we
find that a weak tie between product and marketing means that a lot of valuable
information which could be shared by both teams can get lost. Marketers are often
seeking to understand the target audience, the market size of that audience and as a result
often generate a lot of data around customers. Many people in marketing will have a
strong business background.

On the consumption side of data roles, we find people who are actively taking the
insights generated from the data to help them make decisions about how to push the
business forward. In this bucket, we typically find; business managers, product managers
and, of course, designers.

Business managers and product managers look to data to get stronger insights into
how the business is performing. Business metrics (which we cover in more depth in
Chapter 2) are monitored and used to provide a health check on how the business is
performing. They also look for impact of business decisions and to see if the changes that
are being done in the product are performing as expected.

Designers should not only be caring about the business metrics, but also thinking about
any additional metrics with respect to usability or the user interface. Of course designers
should also be actively involved in understanding the data that comes back from user
researchers or analysts on their own design throughout the product development cycle.
The key for designers is to interpret these results and understanding them within the
context of the larger business.

User researchers and analysts should also be consumers of each others data. Looking to
see how a broader understanding of the data that is being collected on their product can
provide clarity into why they might be seeing locally in their own data.

These roles do not have strict boundaries or definitions. However, as you might find
yourself playing into either a consumer or producer of data it can be helpful to understand
which side of this divide it is that you are fitting into at that specific point in time.

Our use of terminology in this book

Terms like big data and design thinking are in fairly common use now as is the
distinction between data driven and data informed design. However, wed still like to
take the opportunity to define these terms and a handful of others that well be relying on
pretty heavily throughout the book. The terms we define in this chapter are especially
important as they represent some of the key philosophical foundation of what you will
find in the book.

6
OReilly Media, Inc. 2/26/2015

Big data 5 - it seems that finding a formal definition of big data can be quite difficult.
The term first emerged in 2007 and has generally come be used to describe very large
data sets (structured and unstructured) which can has the potential to be analyzed and
mined for information.

Design thinking also a very common term which is used to describe a process for
creating new ideas and solving problems. The term was popularized by Tim Brown 6 and
the folks at IDEO to describe a process where innovation comes from direct observation
of users and relies on creating strong empathy for them. What we hope to do in this book
is to not only show how all kinds of data fits nicely into the construct of design thinking,
but how design thinking as a practice can also be applied to the data itself.

The design council in the UK introduced the notion of the double diamond 7 design
process model in 2005. The phases of this model are as follows:

Phase 1: Discovery where activities like market research, user research help to identify
the user needs. Much of what we discuss in this book is leveraged at this stage of the
design process.

Phase 2: Definition in this stage you are aligning the user needs to the business. Its for
this stage that its especially important to understand what your business objectives are
and which metrics will help you and your team align to them best. We cover this in more
detail in Chapter 2.

Phase 3: Development the design solutions are developed and iterated on at this stage
of the process and its in this phase that the designer should be thinking proactively about
the kind of data that they will need to capture to best understand the effectiveness of their
work.

Phase 4: Delivery this represents the time where the product is finalized and launched.
At this stage we think its most important for the designer to be able to close the loop and
determine the effectiveness of what they have done. Understanding the resulting data
from their design and how it provides insight into which next steps they should or
shouldnt take are the key things at this stage.

Quantitative and qualitative data at scale for the purposes of this book we really
focus on gathering either qualitative or quantitative data at scale. For us what that means
is looking to getting enough information that it can be interpreted at a level that gives you
the confidence that whatever you interpret as a result from that data will scale. That the

5
http://www.forbes.com/sites/gilpress/2014/09/03/12-big-data-definitions-whats-yours/
6
https://hbr.org/2008/06/design-thinking
7

http://www.designcouncil.org.uk/sites/default/files/asset/document/ElevenLessons_Design_Counci
l%20%282%29.pdf

7
OReilly Media, Inc. 2/26/2015

data you are gathering and interpreting is statistically significant. Well talk more about
the details around making sure that your data is sound in Chapter 3.

Data vs. Design vs. designer vs. designing it may seem obvious, but we do want to
call out that the design that is produced by a designer is a tool (in the same way that we
are advocating for the fact that data is also a tool) that is used to help the designer craft
a solution for their users. The act of designing is something that we want to redefine as a
process that not only speaks to the work of crafting a design, but also taking into account
and incorporating data.

Data driven vs. Data Informed vs. Data Aware

Before proceeding wed like to spell out some differences we have perceived in how data
and design have been positioned in the industry. The three terms are data driven, data
informed and one we have coined, data aware. You are probably already somewhat
familiar with the debate between data driven and data informed and you are also
probably aware of the different views that surround those terms as well as the potential
pitfalls of taking too binary approach to data.
One of the best descriptions that weve ever seen on the difference between data-driven
and data-informed comes by way of Andrew Chen. In a well referenced post entitled
Know the difference between data-informed and data-driven 8 , he explains that
[]the difference [] in my mind, is that you weigh the data as one piece of a messy
problem youre solving with thousands of constantly changing variables. While data is
concrete, it is often systematically biased. Its also not the right tool, because not
everything is an optimization problem. And delegating your decision-making to only
what you can measure right now often de-prioritizes more important macro aspects of the
problem.
The words not everything is an optimization problem sum up the philosophy behind
this book. Our intent is to broaden the definition of data from being an overly strong
association with AB test results to acknowledging that what companies need and should
gather are multiple forms of data gathered from many different sources. In reality, many
companies are already collecting all kinds of data on their customers behaviors:
business analytics, trace log data, user experience lab-based data, interview data,
customer feedback data, NPS and other industry standard data formats, call center data
around issues and complaints. However one of the things we are concerned with in this
book is that AB testing at scale has captured the imagination as THE tool to use even
when not systematically conducted because of the enticement of large numbers and the
obsession in or industry with scale as if large numbers cant be wrong. We want to
acknowledge that the intersection of data and design can be so much more than AB
testing and to recognize that it can be equal parts art and science.
Data driven design implies that the data that are collected are what drives
design decisions. In some instances, this is the right way forward. We will give
examples through the book where data are gathered and directly effect the

8
http://andrewchen.co/2012/05/29/know-the-difference-between-data-informed-and-versus-data-
driven/

8
OReilly Media, Inc. 2/26/2015

design decision that was made next. The data is used to answer a specific
question and the results are unequivocal.
In some instances however, things are more nuanced and the data might suggest
an answer but it is not cut and dried. This is what we call data-informed
design, where a team decides to take the data gathered with a pinch of salt, and
perhaps set up another iteration of investigation. This is when more research
may need to be done, different kinds of data gathered, and/or an informed
creative leap is taken. We will give examples of this kind of situation in the
book too.
Finally, we use the term data aware design to underscore the fact that the
design process is a creative one, where design decisions need to be taken not just
from data but back to data that how a system is instrumented, that what data
types are being captured and how they are combined is itself a design problem.
So, designers and data scientists need to work with developers and business
strategists to actively design systems so that the right data types forms are
collected to address the right questions. Designers always need to be designing
their hypotheses and the data they would like to see collected to test their
assumptions.
In this book, we want to encourage you to inform your design decisions by specific,
objective evidence; data. Data comes in many forms, and it is collected in a number of
different ways, and it should be used to give teams a better understanding of what
customers are doing with a product. We want to acknowledge that data is not
information until it is effectively managed and analyzed and it is not knowledge until it is
put into the business and decision making context. Using data in our decision making
entails that we reflect on the quality of the data, on what data is right for the decision
making setting that we critically engage with question relevance (as we asking the right
questions?), data appropriateness (does it answer our questions) and data quality (is the
data reliable? Did we lose something in data collection/curation?). It also requires we ask:
would different data and/or a different analysis be more appropriate? Are we doing what
is convenient rather than what is right?
Being smart about data in your decision making has considerable advantages. Having
common success metrics within your company can also help designers and the broader
product team to align around common goals and to understand what kind of data is the
most important to track and follow. So more specifically, for our purposes, we hope to
present a framework that you can use as a designer to help you hone your understanding
of customer behavior, align you and your team to the larger company objectives and
business goals.

Personal Stories
We thought it would be useful to share a personal story from each of us, to help share
with you why we are so passionate about this topic. Each of us has seen how leveraging
data smartly in the product development cycle has made an impact on our ability to react
to our customers and make smart decisions.

9
OReilly Media, Inc. 2/26/2015

A note on Why I love data from Rochelle

During the summer of 2011 a couple of the designers at Netflix were working on a major
redesign of the website. At that time, the fundamental bones of the website hadnt
been touched in over 3 years even though the business had changed quite a bit. Netflix
had started as a service that offered DVD-by-mail and was going through the transition of
becoming more known for its streaming service rather than its DVD business. When
Netflix started to offer a streaming service, the existing design for the DVD service was
simply replicated with the buttons that originally said Add to Queue being replaced by
buttons that said Play instead. There were a number of other artifacts from the DVD
design as well, some of these artifacts were based on the technical limitations or web
standards that existed at the time that the service was created and some were based on
usability considerations (e.g. more information provided up front so that a customer could
make an informed decision about whether or not it was worth adding a particular movie
to their queue and then waiting the 2 days it would take to arrive in their mailbox.)
The team was really excited to begin working on a new design and we started by clearly
defining a hypothesis for the redesign:
A cleaner UI which showcases the content will lead to more viewing.
There were a number of deeper implications with this overarching statement. For
example, in a world where you could just stream the movie or TV show instantly, you
didnt need to have a lot of additional information front and center - the act of playing
could be the primary means of determining whether or not a piece of content was worth
watching. By enlarging the box art used to represent the movies or TV shows, we could
showcase the content in a more visual way which would be a more compelling way of
showing off the content. Before launching the AB test, we were gathering feedback
through user research and sharing prototypes with customers to see what their reaction
was to the different changes we were making. Feedback generally came back positive.

Figure 1 Original Watch Instantly Web Page

10
OReilly Media, Inc. 2/26/2015

Figure 2 New Clean Design

Figure 3 Play button, stars, text title all appear on hove

The team rolled out several versions of the new design to a small group of users - running
an A/B test to understand which design performed best. The results of these new
designs were compared to the existing design to see if our hypothesis was correct and that
we indeed could move our core metrics. We knew that people who streamed more from
our service were more likely to retain as customers, so hours played was a great proxy

11
OReilly Media, Inc. 2/26/2015

for retention. As the results came in it was clear that the new design indeed worked
better than the existing design. The viewing hours were up and we saw a small lift in
retention. Given the positive results, the team was eager to roll out the winning design.

Announcement of New Netflix Design

However, the response on the Netflix blog and on twitter was strongly negative.

User feedback
In fact, this change to the website resulted in the fastest accumulation of negative
comments we had ever seen (more than 1000 posts within 24 hours). Imagine how the
team felt at that moment. The flood of criticism really shook our confidence and made us
question our design. You could imagine that one outcome would be to roll back all the
changes we had just pushed out. However, because of the prior A/B testing we had done,
AND because of how we had rolled out the new design, the data showed us that the new

12
OReilly Media, Inc. 2/26/2015

design was still performing better than the original design. Most designers know that
when you make a major change to a design that people are comfortable with, that there is
often a strong negative reaction because people generally dont like change. That being
said, it never feels good to be in the middle of all that criticism. The data helped to give
us confidence that the reaction we were getting must have been from a strong, vocal
minority.
But as we picked up the pieces of our broken souls off the ground and we got the strength
to look at the feedback that we had gathered from customers through a number of sources
- we did user testing after the launch to better understand the reactions and to observe
how people were using the product to see if we could identify other issues that didnt
emerge from the AB testing. We learned that the sortable list and scrolling were they key
issues with the design. Luckily, when we had done the testing, these were variables that
we had controlled for. We knew that we could isolate these items and work to improve
on them. We were able to respond to the criticism, and do more isolated testing and
iteration on the parts that users told us were problematic.
If you look at the design of the website today, youll notice that its largely the same as
what we originally launched.
This story illustrates a number of the reasons that I really love and believe in data-aware
design. A/B testing and data from user research and surveys of a broader population
gave us the confidence that our decision was correct and allowed us to keep pushing
forward despite the vocal, passionate reaction of some of our customers. Looking at the
data helped to give us clarity around what the key issues really were and it helped us to
refine our understanding of what the customer found valuable and what didnt make an
impact on them.
A note on data aware design from Elizabeth
In 2006, a prototype product was launched by a research scientist at Yahoo! Lab, then
located at Berkeley. This prototype was a plug-in to Yahoo!s then popular Messager
service that allowed users to watch video in real time together in a chat space. Here is the
scenario (see Figure X):
I am chatting with you in Yahoo! Messenger, and share a link to a video with you. Rather
than opening another browser, pasting the URL in, waiting for the page to download,
clicking on the play button and then watching the video and reporting back to you, I
could simply click on a button that said Watch with Me and the video would load in the
chat, and wed be able to watch it together, You hit pause, it pauses for me. I hit play it
plays. And we can text chat alongside the video as it plays.

13
OReilly Media, Inc. 2/26/2015

Figure X Zync synchronous media chat space

This kind of synchronized media consumption is commonplace now but in 2006 it
wasnt. After a soft launch later, of course we wanted to evaluate users reactions to it.
Numerous measures were taken, from links shared to time spent per session, to number of
links shared, to amount of chat.
After 2 months, we had around .5 million active users a month spending on average 11
minutes a session and sharing a median of 3 videos. 43.6% of the sessions the invitee
played at least one video back to the sessions initiator., we showed a 77.7% sharing
reciprocation and that pairs of people often exchanged more than one set of videos in a
session. We looked at content categories; in the categories of Nonprofit, Technology and
Shows, the invitees shared more videos back to the initiator (5:4, 9:7, and 5:2
respectably). We later developed social network analyses of the diffusion of media
content through different subgroups online.
Back to 2007, the lead scientist, David Ayman Shamma, created various vizualizations of
the data to try and tease out the shapes of user engagement with the product we note
that data visualization is another area where design skills are central to the advancement
and improvement of the data sciences. One of the visualizations is shown in Figure Y.

14
OReilly Media, Inc. 2/26/2015

Figure Y: Percentage of actions over time

This visualization shows the relative amounts of time that people spent rewinding,
playing, pausing and chatting per session. Clearly this shows that people main action was
chatting. However, the more interesting story arose when we looked at the data over time.
This is shown in Figure Z. When we first looked at this plot, we looked at actions as
pulled from the data server according to what was then a standard session. Here the
time of a session for the data pull was set to be a certain number of minutes (describe a
session). We were curious to see the sharp drop of activity, a sign of low engagement for
most end-user applications where action equates to engagement. However, in tandem
with activity data and summary action data analysis we were conducting observational
studies of users trying Zync. And we made an observation that seems obvious in
hindsight, but was not obvious until we did the observational studies: for actions like
watching a video, inaction or doing nothing, is a sign of engagement. To put it simply, if
you were sitting on a couch watching a video and the person next to you keep chatting,
nnin that content and doing nothing. With this insight in hand, we returned to our data
and widened the time window for the log analysis, we created a human-activity driven
session window rather than a system session window, and we saw something very
interesting: the tall spike in activity that you see at the far right of Figure Z. When the
video stopped playing, people started chatting, they resumed their engagement with each
other. We thus defined a whole new set of methods for determining user engagement
with the product, and created a set of design guidelines for data analysis for engagement
as inaction. This example illustrates several things: First, we designed several content
recommendation features for Zync based on what we learned, (2) we derived a set of
instrumentation for experience mining recommendations, including consideration of what
constitutes a session (3) data awareness here was the understanding of how to triangulate
the different data types and reconsider the first analysis as potentially misleading based
on an insights from another data set.

15
OReilly Media, Inc. 2/26/2015

Figure Z: Trace log of activity in a Zync session

Upcoming chapters
In Chapter 2, well shift to focusing on the business side of things. Well cover the
different kinds of businesses you might be in and cover how that might affect the data
that you gather. Well also discuss how the maturity of your business might affect the
metrics that you gather and how to solve for different aspects of your customer funnel.
Were also going to take a look at a few really interesting case studies from different
kinds of companys a t different stages in their lifecycle including starting with one of the
most (in)famous stories of data and design.

Summary/Key Takeaways
We hope this chapter has given you a good foundation in not only some of the
terminology that well be using throughout the book, but also in the philosophical
approach that we would like to encourage as you start to work with data in your design
decisions. We want you to feel empowered by data and to recognize that by adapting this
framework, youre actually engaging in a two way conversation with your users where
both data and design can be a useful tool in that communication.

16
OReilly Media, Inc. 2/26/2015

The history of using data to bring insight and information which facilitates
problem solving is long, we can learn a lot by looking at how other industries
have used data.

Though incorporation of data into design is relatively recent, we believe this is

the beginning of an exciting and long era to come.

As a designer, you will play both the role of a producer and a consumer of data.

There isnt a one size fits all approach to data and design, and understanding
the nuances of data driven vs. data aware vs data informed can be a powerful
tool in design.

Fundamentally, the difference between data-aware and instinct driven design comes
down to what you rely on to inform your design decisions. With data-aware design, data
is a creative production and is the primary decision making tool when and only when the
data have been themselves well designed and have been proven to be what we call fit for
purpose; with instinct- or experience-driven design, decision-making is more
experimental. Both paths can lead to great design. Is one way right? Absolutely not.
Are these methods mutually-exclusive? No. For us, the right approach to design will
vary depending on the nature of the problem you are trying to solve, how you operate
best and will almost always requires a balance between leveraging experience, instinct
and data. You need to have great instinct and experience to be a great designer
regardless. Data can be one more (great) tool that you can add to that toolkit.

Questions to ask yourself

We thought it would be helpful to include a section on Questions to ask yourself at the
end of each chapter. These questions are included as a way to spark some conversation
either with yourself or within your company.

How is your company currently using data?

What kinds of data does your company use?

Who is currently producing and consuming data in your company?

Are there data gaps?

How could you close those gaps?

What kinds of questions would you like to answer about your users?

Additional resources
Content

17
OReilly Media, Inc. 6/15/2015

Chapter 2
Business Matters

Introduction
Smart data practices are key when it comes to having a strategic impact in any business.
If you have a great user experience, then you have the foundation of a great business.
User centered design and activity data analytics are both focused on working out whether
your idea of a successful, rewarding and repeat-worthy experience aligns with the views
of your intended or current customer base. In a monopoly environment, a business can
offer unsatisfying service and still seem apparently successful. Thankfully, todays
competitive landscape means the company that provides a better service will win.
Data capture, management and analysis is the best way to bridge between design, user
experience and business relevance. A data aware approach to design is a good foundation
for cross-functional collaboration within your business, whether large, medium or small.
Its also an excellent way to have impact upon and create alignment between design and
business goals, focusing on the critical part of any business: providing the best possible
service to your customers and clients, understanding their goals and concerns, and
addressing their frustrations. A user centered design practice centralizes empathy and
customer understanding, and as such, is best positioned to also work to define and design
the right measures that accurately reflect the customer experience.
Being user focused and data aware means you and the people you work with should also
be actively contributing to the creation of meaningful business goals that are focused on
the greatest asset of any business: users, clients and customers.
Our bottom line: dont just design the user experience, sketch out what data you will need
that will help you test your design. Design the data capture, analysis and questions as part
of your design process. Be clear about the data that will best help you measure and
articulate the effect of your design on your customers and then through that to the
business.
In this chapter we will give you one take, based on our experiences of how data in
combination with design can impact business.

1
OReilly Media, Inc. 6/15/2015

Thinking about your business

As you can imagine, the kind of business youre in will influence how you measure the
health of your business. Business health is an extremely complex concept, that includes
myriad different measures, that roll up into a bottom line that defines whether the
business is viable or not. Such measures include engineering analytics (service delivery
and robustness metrics), business analytics (which in the end is about balancing profit
and loss), markets and competitiveness, and so on. Ultimately though, whether you are
focused on one audience or many, the dynamic quality of the market today means our
focus, what customers users do, is key.
We kick off this chapter by first asking you to take a step back and to reflect on the
fundamentals of your business. We ask three questions in this section:
1) What kind of business are you?
2) What is your revenue model?
3) How mature is your business?
Your answers to these questions will determine the kind of data that you will want to
collect.
What kind of business are you in?
The data you collect, the measures you define and the metrics you track will, to some
extent, depend on the industry or space that you operate within. Though not a
comprehensive list of business types, the following are some of the most common.
Transactional Companies are on selling goods to customers or creating a
marketplace. Well-known examples are Amazon, eBay and Zappos.com.
User Generated Content (UGC) User generated content sites often, but not
always, use social networks for content distribution. UGC sites rely on their
users to create content that is stored, disseminated, annotated and mined by the
service. That content might take many different forms whether its video, blogs,
audio files, images, commentary, and so on. The best known examples here are
Twitter, Facebook, YouTube, Vimeo, Instagram and Flickr.
Media Sites Companies that provide access to curated media, such as news,
music and video. Examples here are Hulu, Spotify and the NY Times.com.
Software as a Service (SAAS) Companies that provide a service or software
for their customers, hosted by the company. Examples here include
Salesforce.com, Intuit, Squarespace and Zendesk.
Once you have decided what kind of business you are in, looking at what competitors and
others in your industry value will also provide you with important clues as to what might
be of greatest value and thus what to focus on.
What is your revenue model?
The next question to ask yourself is about your revenue model. What is the primary
source of revenue for your business? Some of the most common models are:
Production - the primary source of revenue comes from selling goods.

2
OReilly Media, Inc. 6/15/2015

Subscription your company collects a recurring fee for services or access to

content and tools. Factors to consider here are how often does the subscription
renew? Within the subscription model, some companies will use a Freemium
pricing model where the service is free, but customers might pay for premium or
proprietary features or functionality.
Advertising the primary source of income will be from advertising revenue.
Advertisers might be charged for the number of times that their ads are seen or
based on the number of times that their ad is clicked. You might also get
revenue from advertisers that are looking to get exposure through sponsorship.
Pay as you go customers will pay for access to a service or content on a one
off basis, as needed.
How mature is your business?
In addition to being thoughtful about the kind of business you are in, youll also want to
take into consideration the stage of growth your business is in. The maturity (or
immaturity) of your business and your customer base will have a critical role in the way
you apply a data aware framework. Here are some of the key considerations at various
phases 1 :
Young businesses in the earliest stages of a business, youll be most
concerned with establishing a presence and acquiring customers. Youll be
focused on ensuring that you have a viable business so market fit might also be a
consideration at this stage. A solid foundation between your product and the
market is essential before scaling a business, and the best way to test this fit is to
do carefully designed user testing of your product to go alongside more
traditional marketing analytics.
Mid stage businesses after establishing that you have enough demand from
customers for your product or service and that you are somewhat stable enough
to keep these customers, your focus in this stage may switch from growth and
acquiring customers to optimizing the relationship between your revenue and
expenses. In order to do this, you will need to find out what aspects of your
product or service offering is returning robust and reliable results, and if there
are opportunities for growth.
Mature businesses - if your business is at a fairly mature state then its likely
that you are more concerned with retention of your customers, protecting them
and not losing ground to new competitors. In order to do this, you will need to
find out which aspects of your product are not working as well as which are and
ask yourself, where am I losing potential or actual customers.
With all three of the areas we outline above, you will find that placing yourself into any
of those groups is not clear cut. Some UGC businesses might also have a transactional
component to them. You might be a new business, but that doesnt mean that you
shouldnt care about retaining the customers you have.

1
https://hbr.org/1983/05/the-five-stages-of-small-business-growth

3
OReilly Media, Inc. 6/15/2015

Business influences on data, measures and metrics

Naturally, the kind of business youre in has a huge impact on what metrics and data it
makes sense for you to collect, manage and report. The considerations of the businesses
are very different across different industries, but there may be overlap. Customer/user
engagement, for example, is something that all businesses care about.
A basic engagement measure is the Active User (AU). The idea is to capture how many
people use your product or service on a daily or monthly basis. Business reports often
include summaries of Daily Active Users (DAU) and Monthly Active Users (MAU),
potentially across many categories if the nature of the business is complex. Of course,
defining active can get complicated. It can vary dramatically from company to
company, and can change within a company over time. For example, to Wikipedia, an
active user may be someone who contributed to more than one article. According to the
Wall Street Journal, Twitter considers a user active if they log in once a month. For a
social platform, an active visitor is someone who has come back to the platform at least
once within 30 days. For an ecommerce platform, a metric like active browsing 2 days
out of 7 may be the metric the business considers successful. For a news media outlet,
active engagement with stories once a day may be sufficient.
Drilling deeper we can define engagement metrics that address and measure a broader
spectrum of user activities. The amount of a user profile that is filled out could be one
such measure. Engagement for a social network may mean number of contacts, numbers
of connections or communications made with those contacts, time spent on site, volume
of content posted, produced and/or consumed and so on. For an ecommerce site,
engagement means transactions (purchases, sales, returns, refunds and so on). For a
business with an online presence where revenue is drawn from advertising, clicks (which
are a proxy for interest piqued) and conversions (transactions that include purchase for
example) may be the best metric of success.
The definition and characterization of active for business health and longevity is a
prime example of a metric that needs user centered design thinking, and design expertise
in the creation of measures that will deliver meaningful results and therefore a
meaningful and business relevant set of metrics.
Business types: positive and negative measurements
Lets take a more structured look at the kinds of things that different businesses will
measure. This is not meant to be a comprehensive list, but rather a structure that you
might use to consider what measurements are key for your business. In this structure,
were looking at positive and negative measurements. Positive measurements are the
factors that the businesses are looking to see trend upward and have positive impact on
the business. Negative measurements are correlated to negative impact on the business
and are things that businesses would look to reduce.
Table 2-1
Example Positive Example Negative
Type of business
Measurements Measurements

Transactional Number of products viewed Bounce rate (number of

visitors who come to your
Conversions (emails given,

4
OReilly Media, Inc. 6/15/2015

shopping cart purchases, site but leave without doing

etc.) anything further)
Failed purchases (note that
failures might happen for a
number of reasons on the
customer side (like a failed
credit card) or on your side
(like poor usability or an
error))

User Generated Viral growth Attrition

Content
Content submissions Spam/Bad content
New visitors

Media Sites Time engaged with content Abandonment

Virality of content Disengagement
Number of ads seen (in an
ad supported model)

Software as a Retention Churn

Service (SAAS)
Productivity Support calls
Performance Refunds

Businesses overlap in various respects. To more closely illustrate how one might use the
framework above, lets focus on businesses that centers on user generated content. UGC
product metrics are primarily driven by audience and behavior. Audience numbers are
relatively pervasive across different UGC sites, whereas behavioral metrics tend to be
unique to the business being built. For example, many UGC sites have a SaaS-like
componentthat is, an area where users complete tasks such as writing and publishing a
story, uploading an image, posting a video, and so on. However, the SaaS component of
these sites merely supports the discussion and spread of content. Users creating content
for other users is what sets these sites apart from a SaaS only business.
A business that is focused on user generated content aims to increase the number of
content submissions from their users. The primary question is How can we get our users
to submit more content? As a designer, youll be trying to solve this business
imperative through design, including understanding the trade-offs such as how increasing
the number of content submissions may result in a decrease in quality and a poorer user
experience. As a designer, youll therefore need to craft a solution that encourages people
to submit more content while keeping the content at an acceptable level of quality and
ensuring content relevance. Being able to articulate and ultimately measure how your
design work directly impacts the unique goals of your business is one of the key benefits
of taking a data aware approach to design.

5
OReilly Media, Inc. 6/15/2015

Business types: critical moments

Depending on your business, there are also differences in what is and what is not critical
to measure and monitor. It is useful to think about whether your product or service is
critical for your primary target user groups activities. An emergency helpline for
example need to be always available; it is critical that the helpline does not have down
time. A discretionary use, social sharing platform may be able to afford a small amount
of down-time, without the results being critically negative for the users. However you're
your business is a transaction-based marketplace, where personal allegiance is exceeded
by customer satisfaction with an efficient transaction, the longer you have downtime, the
more likely your audience will go elsewhere. The point here is that, what is critical to
your business is where your resources will be allocated, and where your most fine
grained and nuanced data gathering may need to be.
Take for example, the differences between what mattered to businesses and services like
Twitter and Gmail in their early days. As we all know, critical for most businesses is up
time, that is making sure your services are available a key part of the engineering
analytics we mentioned above. The worst thing that can happen is when the services go
down, that is are not available. We have all clicked through to a URL to see an error, and
it is jarring.
When Flickr, the photo sharing site goes down, a playful image is shown with the
message Flickr is having a massage., intended to let users know someone is working on
the problem and the site will be available again soon. The message is personal and
personable. In the early days of Twitter, the platform went down very, very often. The
root cause came down to scale too many people trying to use the product at once. The
underlying technology couldnt support it, which lead to many users getting error
messages. The error message was a picture created by Yiying Lu, later dubbed the Fail
Whale (see figure 2-1) by twitter user @qrush (Nick Quaranto) 2 . Biz Stone, the co-
founder of Twitter told an NPR interviewer that their idea was to convey that its a big
job but were all working together to do it." It was intentionally fun, but not so jokey.

2
https://twitter.com/qrush/status/822613478

6
OReilly Media, Inc. 6/15/2015

Figure 2-1 The whale

As a result, when Twitter used to go down, many people seemed to be delighted. Shouts
of glee would hit walls on Facebook when Twitter went down. Business Insider, The
Next Web, BBC News, Times Magazine and Read Write Web are all examples of
publications that have called the Fail Whale Iconic at least once. 3 In short, people love
the fail whale, so much so that it has been the inspiration for jewelry4 .,

Figure 2-2 Tweets from when Twitter was down.

Figure 2-3 Tweets from when Gmail is down.

By contrast, when a service like Gmail goes down, even in the early days, it caused
anxiety. This is because theyre two fundamentally different kinds of business. Although
it is now a central source of news and therefore more critical, Twitter started as a
recreational, user generated content business. Peoples livelihood generally does not
depend on a UGC or social networking service especially for those in early stages, so
downtime is more easily tolerated.

3
https://www.google.ca/search?q=fail+whale+iconic&oq=fail+whale+iconic
4
https://www.etsy.com/ca/listing/60239227/hoist-fail-whale-necklace

7
OReilly Media, Inc. 6/15/2015

SaaS products like gmail are products people rely on. When it goes down, it means
someone cant do something that theyve generally paid (or are getting paid) to do.
The moral here isnt that its more fun to work at Twitter than at gmail. The moral is that
SaaS companies care about things like performance and user experience. Social
networks care about completely different things, like sharing, return visits, or reducing
spam content.

Metrics for different business stages

Drilling a little deeper into the stage where you are in your business, one of the most
common frameworks for growth is looking at the customer funnel through what have
been called AARRRC metrics. These letters stand for Acquisition, Activation,
Retention, Referral, Revenue and Conversion. In detail:.
Acquisition How well is your business doing at acquiring new users? If your
business is relatively new, then you might be most focused on building up this
part of your customer funnel acquiring new users to your service or product.
o Examples of typical acquisition metrics include number of new
users/day or week, number of sign-ups, etc.
Activation - How do you know if the user you just acquired is really going to
stick around? What do you measure to gain confidence that they are going to
actively use your product or service on an ongoing basis?
o When measuring activation, a good way to approach it is to think of the
key value that you want to make sure your customers get exposure to in
their first usage of your product or service. Usually this gives you a
great way to measure activation: Is it sharing something to another
person? Is it registering more of their friends? Is it playing a song or
watching a video? Is it uploading something to your service?
Retention - How do you understand if your users will be coming back to you on
a regular basis and that you are creating a sustainable business? Retention is
fundamentally about continuing to build your user base, retaining your
customers. Theres no sustainable way to grow your customer base if you have a
leaky bucket and therefore retention is very often a key metric for many
companies.
o Typical retention metrics might look at how many new users remain
active 2 weeks after first using your service/product or 1 month after
using your service/product. You may want to tie the timing of your
retention numbers directly to the length of your trial period.
Referral this looks at how many people are referring your product or service
to others. Its a good predictor of growth and typical metrics looked at here
might be shares, viral coefficient or NPS (net promoter score).
Revenue this is the fairly obvious measurement of how much revenue your
product or service is generating. In as much as most businesses aspire to
eventually be self-sustainable, revenue is a key metric to track. You might look
at revenue growth over different periods of time.
Conversion - This is most relevant for businesses that have a free trial or a
freemium model and it refers to converting your customers from a free customer
to a paying one. With conversion metrics, you may want to consider what time
period youd like your customers to convert in - do you want them to convert as

8
OReilly Media, Inc. 6/15/2015

fast as possible or is it ok if they are a free customer for a long time before
deciding to pay?

Depending on the kind of business youre in and the maturity of your company, some of
these stages in the customer funnel will have different levels of importance for you.
Good data-driven decision making isnt even possible at certain stages in the
development of a company. Early-stage companies have, in most cases, far more pressing
issues than split testing the color of a button. They may also not have access to good,
robust and reliable data at scale. They may also be designing their metrics as they go or
want to take leaps of creative inspiration that go counter to what the data say. In some
cases, the data may also be focused around a local minima because the company simply
hasnt been around long enough, or tried enough divergent variations to find their sweet
spot yet.
For example, if you are part of a start up that just launched you are most likely focused
on getting new users and establishing your customer base. In this situation you would
focus on acquisition and activation. The data that you collect and the kinds of
experiences that you create will most likely be around trying to find insight into how you
can improve this early part of the customer funnel.
If youre part of a company that has been around for a long time and has a substantial
user base youre probably thinking about how you can retain the customers that youve
worked so hard to get. Youll want to collect information about the retention and
conversion sections of the customer experience as well and youll want to understand
what is working or not working for your customers. By understanding what data youre
using to monitor the health of your business youll be able to align your design thinking
to the very same problem space and metrics that youre using to judge your business on.

Having a solid understanding of the different metrics that you can influence at the
different stages of the customer funnel can also give you a perspective on the longer term
consequences of your design. As an example - lets say that you are trying to optimize
your sign up flow. Right now, you ask customers to fill in their password twice to
confirm that theyve entered it correctly. However, you believe that every extra step (no
matter how seemingly small) causes friction for your user that you want to eliminate - so
you devise a design that doesnt have the field to repeat the password collection. You
might find that this actually works, and you are getting more users to complete the flow
so the metric of # of users who sign up increases. However, if you think about the full
customer funnel, you might start to find that the users who didn't have to enter their
password were more likely to call customer support to retrieve their password or that they
didnt come back because they couldnt remember it and didnt bother to return. In this
situation, you may have successfully increased one part of the funnel, but without
considering the full set of metrics that are relevant to your business you may choose
metrics that satisfy a local goal but harm the overall business goal. This is why its key
to plan ahead and make sure that you understand how the suite of metrics related to you
business relate to each other and potentially interact.

9
OReilly Media, Inc. 6/15/2015

Other relevant business metrics

There are other business metrics that are really important to track when driving a holistic,
user centered data aware design framework. Two worth highlighting are:
Net Promoter Score (NPS), which we mentioned in brief above
Customer Lifetime Value (CLV)

Net promoter score the net promoter score or NPS has become a fairly common way
for companies to measure satisfaction and the general sentiment that their customers
have. Its based on the question, How likely are you to recommend this service/product
to a friend or colleague. This is measured on a 10 point scale. The score is calculated by
taking the number of promoters of a brand or company (those who rate it 9 or 10) and
subtracting the percentage of people who are detractors (those who rate it 0-6).
Customer Lifetime Value the customer lifetime value helps you to measure the
amount of profit that your company will derive from a customer. CLV is a metric that
many companies will optimize for by ensuring that they are attracting and retaining the
customers that will provide the most value for them in the long term.
As businesses today develop ecosystems, its going to be increasingly important that you
think about all the inputs from various devices, locations and experiences that contribute
to the full picture of how your business is doing and what is important to that business.
Collecting data from a single source (whether it be a platform or a market) will not give
you an accurate impression of the health of your business and you might find yourself
optimizing for the wrong things. The emergence of the internet of things and new cross-
device and platform experiences means that we will be generating new sources and new
forms of measure and metrics that should all be taken into account when considering the
input to your designs.

Focused metrics
It is important to be clear about what are, for your business, key metrics that you and your
cross functional partners can focus on driving. As weve pointed out in this chapter, being
aware and recognizing all the variables around the type of metrics you may care about
from a business point of view can help. But its easy to get caught up in a long list of
everything that you could possibly measure and care about - so make sure that you have
carefully selected the metrics that you think are most representative of what you need to
focus on for your business to be successful and that you are not picking too many. Be
disciplined about identifying what you need to improve. Everyone has a limited amount
of time and resources, so you will want to keep the team focused on driving the metrics
that really matter and not getting caught up in moving things that arent key or wont
impact your business.
These metrics should be fairly stable; resist changing them month over month.
You should ALWAYS be monitoring and measuring your company key metrics.
Tracking these metrics on everything you do will help to keep you thinking
about the larger picture so that you dont fall into the trap of over optimizing for
one piece of the puzzle rather than the whole thing.

10
OReilly Media, Inc. 6/15/2015

Ask yourself: Are you selecting metrics that get you to focus on the right things?
The metrics that you pick to focus on will affect your behavior and the kind of
solutions you design. If you focus on click through of a button, you will make
designs that only focus on the button. If you focus on improving the retention
metric, you will think of designs that improve the overall user experience.
Some companies have tried the exercise of having one key metric the
bottom line metric that if that was the one thing to be measured it would reflect
the business health. Think about what would be your businesses one metric and
how you might design your experiences to impact that metric.
Why spend so much time on metrics?
There are three key reasons why we think metrics matter.
1. Alignment
2. Focus
3. Consistency (for future learning)
Ensuring that you and anyone else you work with are all in agreement about which
metrics matter most for your business will keep you aligned. There are many times (not
just in a data aware environment) where teams can get distracted or waste time because
they dont realize that they are solving different problems. By having a clearly defined
and objective measurement of success like a specific set of metrics - it can help to ensure
that you and your partners are aligned and have the same goal. You have agreement that
you are trying to impact the same thing in your business.
Having commonly defined metrics can also help to ensure that you stay focused. Its
very easy to think of additional things that you want to change about your customer
experience. You start to layer in feature after feature and improvement after
improvement, without being disciplined about making sure that those improvements tie
back clearly to your original goal.
Finally, by having clear metrics you also allow for consistency over time. Hopefully, you
arent changing the metrics for success all the time. Occasionally these metrics will
change, but ideally, thats the exception and not the rule. This way, if you are working
on a series of explorations and you have a consistent set of metrics that youre using to
measure them against, then you actually get to understand what kind of impact (delta)
you can expect to get from certain kinds of design changes. You can compare results
from one design to another. Metrics should provide you with a solid framework for
ideation. By knowing what metrics you want to affect, you should be focusing on the
kinds of ideas that will actually have impact on the things you measure and consider to be
important. By having all of your design explorations evaluated by the SAME metrics
you can compare their impact against each other and have a good baseline so that all of
your ideas are being judged on a level playing field.

Summary/Key Takeaways
Understanding your business and the maturity of your company (early stage, growing or
established) are critical factors in determining the measures and the metrics that will best
reflect how your designs are impacting your business goals.

11
OReilly Media, Inc. 6/15/2015

Common metrics and measures help to align and focus your work, and ultimately they
should tie directly back to the user experience. It is important to think carefully about the
customer journey with your product(s) for different user groups. What is the customer
experience that youre trying to drive and how does that relate to some of the business
metrics that your company cares about?
You want to make sure that everyone in your team agrees ahead of time on how you are
measuring the success or failure of your design. Its important to iron out any differences
at the beginning of your project and to continue to build alignment at every step of the
process. That means aligning on the key metrics at the company and business level.
To do this youll talk to key stakeholders, CEOs and other people who craft the
product vision. Its best if you actually do this together so that you have
complete buy in and understanding of what the key metrics are at all levels of
your organization. If you were to share the metrics you think are most important
back to the CEO will he or she agree that those metrics are an accurate way to
measure their vision?
Its also good to build alignment with not just the top people in your
organization, but also the people who are building the actual product. If they
buy into the way you are measuring success then they will make better tradeoffs
and push for the right decisions as the product gets built. Once youve built
alignment on the key metrics its not likely youll need to do this again unless
your business model shifts.
Finally and critically, you should always try to put yourself in a situation where a good
customer experience will be reflected in good metrics. When you start to find that you are
creating bad designs in order to move your metrics you should question the metrics you
need to stop and assess what it is that you are doing wrong. As a designer, you are
advocating for the customer experience first and foremost, a successful business should
be giving its customers a great experience.

Questions to ask yourself

What kind of business is your company in?
o What are the positive and negative metrics that matter most to your
business?
What stage of maturity is your business at?
o Are you starting up (i.e. acquiring new users?) or are you a more
established business and care about retaining existing users? Or is it
some combination of those two?
What does success look like and how are you measuring it?
Who is the audience you hope to affect?
o What do you know about them as a demographic? Their habits?
What are you trying to accomplish? What is your goal and what does success
look like?
o What is the customer behavior that you want to encourage?
o Are you trying to directly increase a metric that is important to your
business (e.g. getting more people to sign up, getting your customers to
return more often to your product?)
o What will you do to encourage the user behavior that you are seeking?

12
OReilly Media, Inc. 6/15/2015

How are you measuring success?

o Which metrics are you going to look at to understand if the impact you
hope to make through design is the right one or will have a big enough
impact?
o How or will you measure or factor in user sentiment? Will you use
surveys, focus groups, and interviews?
How do the various measures and metrics relate to each other?
o How are you ensuring that a positive result in a specific arena (e.g.,
password setting) does not have down-stream negative consequences
on the business?
o How do your behavioral and usage measures relate to the attitudinal
measures reflected in your customer satisfaction surveys, NPS surveys
and user verbatims?
If you had just one metric to measure the health of your business, what would it
be?
All of the above questions are implicated by the business youre in and the way you
approach gathering data on your design and your experience will have clear impact on
your business. In Chapter 4, well revisit a number of these questions as we begin to offer
some more specific and practical advice about how you apply a data aware framework to
design.

13
OReilly Media, Inc. 4/8/2015

Chapter 5
Culture and Communication

Introduction
As we mentioned at the beginning of the book, the strategic and programmatic capture,
management and analysis of user behavior data are typically not considered to be a part
of good design practice. Indeed, quantitative data collection and analysis to drive design
decisions in particular has been contentious and has sometimes even had a polarizing
effect on designers and product managers.
However, we believe that designers need to play a part in the design of data capture and
analysis, as we have argued in earlier chapters. We also believe that designers can play a
key strategic role in companies by engaging in the design of data communication
strategies and practices. However, to communicate most effectively you need to be
willing to interrogate and get your hands into the data yourself, to be part of designing the
tests so you know where the results came from.
This chapter follows on from our description of business cultures in Chapter 2 to address
company culture and specifically group and company communication cultures. We focus
on communicating data driven and data informed ideas, showcasing where product
excellence is derived from systematic and programmatic engagement with data and how
to put the foundations in place for a data aware group and company culture. Creating a
culture where all of this comes together well really depends on three key things: the
material to be presented, the presentation format and process, and the people.
When you have a culture that supports data communication and prioritizes collaborative
learning, it is easier to create strategic, tactical and operational business alignment. More
than that, as a design practitioner, data is one of the best ways to persuade people of your
case, to support your arguments and this to more effectively advocate for your users and
create the most effective and delightful experiences. Data allows you to bring your users
voice into the conversation, helps you know which interactions are the most important
and when. In some senses the usage data you capture is a representation of your user, of
what they care about and when, of what they need and dont need. So you need to design
your data to help you know your user better.
Although we focus on communication here, we dont intend this chapter to offer you the
foundations for becoming a marketer/PR person, nor a data visualization expert. These

1
OReilly Media, Inc. 4/8/2015

are both of skillsets and competencies you may want to hire for and/or leverage as you
design your companys internal- and external-facing communication strategies. Rather,
we are advocating that you work by enacting 6 principle activities:
(1) aim for shared understanding of the value of user centered as a company universal,
by
(2) laying the foundations for engaged and transparent conversations about systematic
data collection, in order to
(3) create alignment, and locate and establish allies and co-partners across company
functions.
Also aim to
(4) hire and internally recruit the right people, and to
(5) articulate the long term benefits and the ability to do effective pattern finding across
different kinds of data set and to
(6) drive a culture of learning, and communicate what youve learned with clarity and
integrity.
On integrity, in our next chapter we will investigate more on the issues to do with ethical
data collection.
In this chapter, well cover two things:
1) What a data informed culture looks like
2) Some tactics that you can use to get there

Benefits of designing your culture of communication

When it comes to user data, summarization and communication are key to having an
impact on business.
In a recent Harvard Business Review article (April 2014) the benefits of being a design
led organization were outlined. It is clear that those design organizations that are most
successful are ones where communication is carefully crafted.
What are the benefits of carefully crafting and designing communication around data? A
culture of communication allows deeper business understanding, including understanding
of actual and hoped-for users, and for better product experience over a customer life-
time. Creating a culture of questioning, learning and data innovation, and a culture that
develops a common set of goals, and agreed on set of user-focused metrics and a
common vocabulary will lead to a more successful business. There are a number of ways
of achieving this that we will discuss in this chapter:
Aim for shared understanding of the value of user centered as a company
universal which requires you to
Lay the foundations for engaged and transparent conversations about systematic
data collection. You need to create a dialogue, set expectations and develop a
common vocabulary.
Create alignment, and locate and establish allies and co-partners across
company functions.and.

2
OReilly Media, Inc. 4/8/2015

Build a team through strategic hiring to reflect, amplify and grow the
above..and
Articulate the long term benefits and the ability to do effective pattern finding
across different kinds of data set by establishing a disciplined and consistent
approach to leveraging data that is programmatic, flexible and not rigid ..by
Creating a learning organization to ensure you perennially grow your
knowledge base.
Well walk through each of these points to help you think about which things you want to
adopt for your organization and which may not make as much sense for you or your
team. At the end of the chapter, well suggest some tactics that you might use as you
build up your own culture of data informed design.
Principle #1
Aim for shared understanding of the value of user centered as a company
universal
For a data informed approach to be successful in any organization, it must be both
universally embraced and understood. By universal we mean both in terms of depth
(that is accepted at all levels, from senior executives to individual contributors) and
breadth (that it is accepted throughout the company, beyond the product development
organization as well). This means that people at all levels in the company and across a
broad set of functions within the company recognize that all kinds of data are triangulated
to inform product decisions, and that they have a part to play in designing, understanding
and interrogating data. It also means that they have at least a basic understanding of what
this process looks like and what the pros and cons are of using data are. This means that
there is a good support net for data informed design to happen having this level of
transparency in the process and common understanding can help to keep the design and
product teams accountable to good practices.

Depth: Top to bottom

Because a data and design framework is conceptualized as being in opposition to gut
decision making, it is often also about breaking the hierarchical decision making
structure. Traditional management-led, top-down decision-making based on hunches goes
away, and is replaced by company-wide, all-level data literacy which includes addressing
well-formulated, carefully investigated questions. While this is the norm for business
analytics, this needs to be developed for user behavior analysis for more effective
customer-centric learning to drive product decisions.
This data may be collected using any number of techniques (such as ethnographic field
research, field deployments of prototypes, surveys, lab studies and usability research,
testing at scale using AB testing, etc.) Such grounded and carefully gathered user data
can drive your decisions, and is a less risky prospect than relying on hunches from
individual executives working from gut feeling. An early paper that focused on AB
testing published in 2007 by Ron Kohavi, Randal Henne and Dan Sommerfield entitled
Practical Guide to Controlled Expriements on the Web: Listen to Your Customers not
to the HIPPO 1 , addressed this issue:

3
OReilly Media, Inc. 4/8/2015

Many organizations have strong managers who have strong opinions, but lack data, so
we started to use the term HiPPO, which stands for Highest Paid Persons Opinion, as a
way to remind everyone that success really depends on the users perceptions.
If you are trying to shift into a data informed framework, it will be critical to look at the
way product and design decisions were made previously made. If these decisions were
largely in the hands of one or two individuals then its worth taking the time to make sure
that they can understand and recognize the value of bringing in user/customer data into
the decision making process because in the end, they may find that their weight is
diminished (as is the weight of any individual opinion). This is a good point to check
with yourself as well are you comfortable and willing to see how depending more on
data will change the weight of your own opinion within the design and product
development process?

Breadth: beyond design and product

In addition to getting all levels of an organization on board with a data informed
approach, it can be just as important for teams outside of product and design to have a
basic understanding of a data informed framework and how its being employed.
Company cultures that embrace transparency are company cultures that encourage
participation and engagement from their staff.
Sharing well-summarised data will engage parts of the organization that are not normally
part of the design & product development process and make them feel like they have a
closer affinity with how the product is shaped. If people in organizations beyond product
and design feel like they can empathize with your users, then they can also feel
empowered as part of this process. A carefully articulated data and design framework is
something that many people beyond product and design can understand. By being more
transparent with the rest of the company about how product and design decisions are
being made (and especially in product and design driven companies), you can build
broader support within the company for the decisions that are being made.
There is another positive side effect of having your data informed approach being broadly
understood. It may initially seem like its superfluous to have teams like recruiting, HR,
sales, or finance fundamentally understand a data driven approach. However, it is always
helpful to have any team that you work with in any capacity understand how you do your
work. Having recruiting and HR understand the companys engagement with data helps
with attracting, attaining and retaining top talent, something we will turn to later in this
chapter. When product development teams understand what questions are being posed by
different parts of the company and how those questions are prioritized and then answered
through data helps staff understand the various aspects of the business. This, in turn
empowers them further in the idea generation stage and better aligned around product and
business success. Imagine the sales executive who feels that they know exactly what you
need to build into your product but may be overly swayed by a key vocal client. If that
sales executive understands that you build product by looking at the data, then they
realize that they need to start to articulate their arguments into the same framework - and
it helps to get you and them aligned with what you are building, why you are building it
and what impact you think it will have on your overall user base.
There is a difference between using data to make and inform decisions in one part of an
organization (e.g. business strategy) and having it be universally embraced by the entire
organization. To get this level of understanding throughout the company, using data has
to be something that you actively share within all parts of the company and has to be

4
OReilly Media, Inc. 4/8/2015

something that is talked about and participated in broadly (e.g. in company all hands
meetings and other public forums).

Principle #2
Create a dialog, set expectations, establish a shared vocabulary and lay the
foundations for engaged and transparent conversations about systematic data
collection.

It is important to make sure that the people in your organization understand at least
superficially whats involved and that they are essential to the process when it comes to
leveraging user data effectively. If a process isnt already established in a company there
will always be some level of skepticism around something new. You might find that there
are people who will say Yes! We should definitely start to take a data informed
approach to design, but we shouldnt do X, Y and Z or that what they think it means to
be data informed, might be something very specific (like only AB testing) and that they
arent considering the bigger picture where different kinds of data collection and analysis
fit together to answer strategically important questions.

Establishing a knowledge baseline

Therefore, having decided you want to enroll the company around data capture, analysis
and consumption, you will need to create a dialog. Find out what people really do know
and dont know about taking a data informed approach to design. Some people may
understand user research, but have much less experience with AB testing (or vice versa).
Find out what the biases and assumptions are about data informed design and do this at
all levels of the organization (from CEO to individual designer or product manager).
Having all of this understanding up front about what people think and know can really
save a lot of time in the long run: Youll be able to assess where you might want to focus
initially in terms of establishing best practices and youll also become aware of any
challenges you might face as you try to get a data informed design process established.
Once youve established what the knowledge level is, where there is deep knowledge and
where there are opportunities for teaching, learning and sharing, then explicitly talk about
what practices you want to establish and why. Getting your team and others that you
work with to agree to try out some of these things that might be new to them will be
important if theyre not familiar with some of these tactics. Its always good to set
expectations for what you will do and what you expect to see as a result, so saying Id
like to take an extra few days on this to make sure we have time for user feedback on
how usable this flow is. Makes it clear that the cost will be a little more time to
introduce something new and what level of results you hope to get from it. Be clear about
what method you will be using to answer which questions and why; too often A/B tests
are used to answer questions that would have been better suited small-N, lab based
usability research - or vice versa. This leads to inappropriate negative evaluations of
good methods. As we have shown in previous chapters, literacy in data collection and
analysis methods of all kinds means you can create a clear rationale, set appropriate
expectations and offset negative reactions and loss of trust. Data ignorance fuels data
reticence.

5
OReilly Media, Inc. 4/8/2015

Establishing a common vocabulary

A key part of establishing a shared set of values is the language that you use. Think
about the words that you most often hear at your company. What are the words and
phrases that are unique to your company? Which acronyms have you created and
embraced? When you first joined what were the things that stood out to you as unique?
Now think about how those words and phrases help to set a tone for how business is
conducted within the company and how usage of certain phrases make you feel like
youre part of an in-group or in the know. To make a data informed decision-making
framework is universally embraced, its then important to actually have a common
vocabulary and way of speaking about it.
Youll find that a common vocabulary will help in a number of ways, from making
brainstorms more effective to making debate over product and design decisions even
more rich because youll have a shorthand for the things that are most important to the
discussion. Seemingly simple questions that can get asked over and over again, How do
we know X?, How are we going to measure Y?, How do you expect the majority of
users will react?, What is the learning curve for this?, What user errors or stumbling
blocks do we think are reasonable or not?, What do you expect to learn about user
behavior from this design?, What does design success look like?, What does design
failure look like? can quickly help to train people and set expectations about the role that
data can play in decision making and in strategic, grounded design iteration.
Remember though, this should be as much about having common definitions as it is
about having a common vocabulary. Since a data informed environment requires a
certain amount of discipline and rigor to the practice, you cant ensure consistent
application of this framework if you dont also have alignment on your definitions as well
as your vocabulary. Dont get too caught up in the words and acronyms (especially
around metrics) such that you forget that the ultimate goal is to provide the best, the most
compelling experience for your users. Find ways to actively represent the user in your
definitions so that everyone recognizes that data itself does not humanize but its how we
interpret it and how it is applied and used to rationalize decisions about product design
and product quality that can be dehumanizing.
Key phrases and vocabulary for a data-aware approach have been defined throughout the
book, and these include terms like:
Design Space
Hypothesis - by stating my hypothesis is... it helps to place your idea
within the context of everyone elses ideas. Calling it a hypothesis
reminds you that it hasnt been proven yet and that it needs to be tested.
Hypothesis also has the added benefit of sounding more scientific, making
people less attached to their ideas and therefore more open to testing them
and being proven wrong.
Statistically significant - a great shorthand for knowing whether or not you
should act on the results from your test. You can always refer back to
whether or not the results were statistically significant.
Variables
Measures

6
OReilly Media, Inc. 4/8/2015

Metrics - Weve spent some time in this book pointing you to the fact that
the right user (rather than business bottom-line) metrics for your company
need to be designed. We have also spent some time identifying the
difference between core, proxy and secondary metrics. For the purpose of
this section, metrics is a key part of the data informed vocabulary because
the metrics that you use and measure represents how you measure success.
Clear understanding and agreement on how you are defining metrics (at the
company level and at the project level) will ultimately be an indicator of
how aligned you are on any given project or discussion and can therefore
play a significant role in how effective your discussions are. You might
find that you start to use questions like is this change really going to move
the metrics? to ask if youre really investing your time into a project that
will have a measurable impact on your users. Questions like which
metrics do you think this will affect? can help you to make sure youre
looking to impact the right level of user behavior.
Validity
Reliability
Soundness
Replication/reproduceable
Confound
Certainty
Ecological validity

In reality, the vocabulary itself may not be as important, its really more about consistent
usage of this vocabulary that helps to reinforce a specific mindset and a way of thinking.
Like working in any company, the vocabulary can help to shape the culture. What are the
words that are already part of your company culture that you can adopt to support a data
informed approach?

<Sidebar: Vocabulary>

Principles #3 and #4
Create alignment, and locate and establish allies and co-partners across
company function and build a team through strategic hiring to reflect, amplify
and grow the above
People matter. Look inside your company, seek out sympathetic and curious others,
recruit allies through conversation, question asking and engaging in their processes.
Invite colleagues to be pilots for studies and to comment on designs and plans. Small
pilot groups are invaluable. If your company or organization has not been data aware
historically, and/or has engaged in a data informed design practice before, youre likely
to find that there are some folks who are eager to try something new and others who will
have more resistance to it or not want to be bothered with the hassle of adopting a new

7
OReilly Media, Inc. 4/8/2015

practice. Generally, finding a few allies who are willing to try something out with you
and who are also going to be more forgiving in the initial stages when you are likely to
make a lot of mistakes can be really helpful. This will give you a chance to work out any
kinks in the process with a small group of supportive people and then have more
confidence as you try to roll it out to other designers or teams.
To find a good pilot project, there are a couple of characteristics to consider. Find
something small so that you can ideally get a quick win and share the positive results of
using data informed design to others. If the first project you pick to use data informed
design takes you a few months just to establish what data youre looking for or to do the
analysis on the results, its very unlikely that youll get a lot of enthusiasm from others to
adopt something new. Find a project which has well defined success metrics. This will
make it easier for you to explain what youve learned or not by applying a data informed
process. A project that allows you to do a couple of quick iterations can also be good
because it can demonstrate a data informed process over a couple of cycles. We really
try to encourage using data to do more than just optimization but when youre
establishing the practice for the first time these kinds of projects are perfect for
warming up the larger organization to the concept of data informed design and to build
confidence in adopting a new framework for design.
Finally a note on allies. As mentioned above, some people will be naturally drawn to a
data informed approach and others will not. The best allies exist in other teams. Having
many different voices who can advocate for a data informed process is very powerful.
Some of the best advocates are in analytics teams and in marketing. Seeking support
from the broader organization helps to make this more than just a design thing and
helps to elevate it to the level of a company wide initiative. This also helps with the
universally accepted aspect of culture we discussed earlier.

<Sidebar: Establishing a new function>

In addition to gaining internal allies and collaborators, new people matter. The people
you hire play a huge role in the success of your team. Building a team to work
effectively with data informed design requires hiring a certain kind of profile. There are
several qualities that are central to creating a strong team to drive a data aware design
team and a data informed design practice in your organization:
1. fundamental design skills the best designers know how to interpret a difficult
problem and create a solution that addresses it. They can also articulate and
justify the reasoning behind their design decisions. Its no different working in a
data aware environment. The best designers in a data aware environment will
leverage these fundamental design skills to address and articulate user problems
and needs that will often be identified and measured using data.
2. a passion for driving a successful business that is also consumer/user-centered.
Ideally the people you hire will have a passion and curiosity about the business
youre in. To build a data aware design environment, it becomes even more
important for people to have a certain level of business passion and desire to
contribute to strategic thinking about the business. This is because it all ties
back to being able to create hypotheses that have the intent of moving the
measures metrics you use to measure your business by. If the designers and

8
OReilly Media, Inc. 4/8/2015

product managers who are tasked with building the business solutions dont
have a baseline level of passion for the business bottom-line measures and
metrics, if they are happier being tasked and throwing designs over the wall,
then they will be unhappy in the environment you are creating.
3. an affinity for understanding, generalizing and being able to replicate design
success and drive consistent user centered design and product excellence. Being
programmatic and systematic through data is about understanding and being
able to derive general principles and being able to replicate your results and to
derive general principles. Even if results cannot be replicated for good data
reasons, it is important that general principles are extracted that could be
applicable to adjacent or downstream projects.
4. a desire to engage with and learn about the scientific method and develop some
skill or affinity for mathematics. It is true that having some basic understanding
of statistics and analysis can make a big difference in whether or not an
individual can succeed in a data-aware design environment. You dont need to
be great at it, but its important to have some basic level of math is just so that
you can have conversations with the folks that are doing the analysis of your
work. You need to be able to question how the success of your work is being
measured and you need to have an appreciation for it. Without some
fundamental understanding of the theory behind many of the techniques in a
data informed environment, you wont be able to engage in the more strategic
conversations.
5. an ability to be open-minded, iterative and engaged but not enslaved by initial
design inspirations. There are many kinds of designers that exist in companies
and all types of designers can be successful in their careers. For some designers,
a healthy ego and a deep belief in the fundamental rightness of their designs is
a necessary and positive characteristic as it can sometimes give designers the
confidence and ability to convince others of their vision and to sell them on that.
Ego is a necessary component for creativity to survive. In a data aware,
experimental, iterative design environment, over attachment to designs that can
result from ego can get in the way of success. This is because as part of a data
informed organization you often have to be willing to let the user and usage data
make the decisions. Just as the common phrase applied to authors that they
have to be the murderers of their children may be true, the same is true for
designers. Letting go can be hard but may be necessary.
6. an ability to be very focused on results as well as production. Especially in the
AB testing part of data informed design, it can be easy to get distracted with
different options and test cells and variations. By having a certain amount of
focus and discipline, your staff will be better equipped to resist the urge to fall
into the various pitfalls of AB testing around making a decision too early or
building out too many test cells.
You dont always need to hire people who have actually had experience in doing data
informed design, but you at least need to find people that seem to have an affinity for
working within that framework. If the folks that you hire arent fundamentally open to
the concept of data driven design then it can make it very difficult for them to be
successful in the team. This of course doesnt mean that they are bad designers, it just
means that they would probably be more effective (and more appreciated by their peers)
in a different environment. As we mention throughout the book, there are many ways to

9
OReilly Media, Inc. 4/8/2015

do design and to build product. What fundamentally determines the success of design or
product is finding the right fit between the individuals, organization and the processes
that you use to get there.

<Sidebar: accepting data>

Principle #5
Articulate the long term benefits and the ability to do effective pattern finding
across different kinds of data set by establishing a disciplined and consistent
approach to leveraging data that is programmatic, flexible and not rigid ..by
One of the aspects of creating a shared culture around user centered data is to be
programmatic about your data collection and sharing. What does this mean? We have
already talked about crafting a careful and consistent. Beyond this though you need to
create a framework for sharing results. The more you can create a set of data sharing
formats and practices the more people will sign up. For example, create a consistent look
and feel to your reports. Create a carefully managed repository of reports where people
can seek information as they need it. Create a rhythm for data sharing. Use physical
space to showcase results as they come in. Create what social scientist Nina Wakeford
calls atmospheres around the presentation of your data so that you can help others
understand and interpret their significance, and so that they can work to make the
connection to what the data suggest are actionable next steps. By careful management of
results over time you can avoid ephemerality and you can avoid the constant redoing of
work and the constant reinvention of the wheel. Systematic record keeping and sharing
in common formats makes your message potential stronger, and makes the business
stronger. We advocate putting considerable effort into your data hygeine.

Revisiting Data
One of the main reasons people consider adopting a data informed design approach is that
it can save you a lot of time in learning something about your product before you release
it. You can learn a lot from past internal examples as well as by looking externally at
what other companies are sharing about their learnings.
As data informed design is becoming an established process, weve found that an
effective technique is to take past projects and then retrospectively look at the data using
a data informed framework. Of course this may not always be possible, because you may
not have had all the proper tracking in place but many times the basic metrics will be
tracked and available. You can look at old projects and see if there was any way of
looking at the data ahead of time that might have affected the decisions you made on the
project. Often the data is available to but the team just wasnt considering it.
This might be especially effective on projects where the outcome wasnt what you were
expecting. For example, was there a project where you were hoping to see a big lift in
acquisition and it didnt happen? Going back to the data, was there anything you could
have looked at prior to making those changes that might have given you a hint as to the
fact that it wasnt going to be successful? How can you build that knowledge into future
tests? Or perhaps your users reacted very badly to a new feature you launched but really

10
OReilly Media, Inc. 4/8/2015

thought was going to make them happy applying some survey techniques or user
research in these instances (even though the feature has already been launched) will give
you insight into what your users might have been able to tell you ahead of time if you had
done this work pre-launch. When you can share these kind of insights back to the team
(even if it does feel too late) it makes people that much more enthusiastic to
incorporate a data informed approach earlier in the process the next time.
Another best practice which is especially effective with respect to AB testing is to keep
track of which cells you (and your team) think will win and then comparing it to what
actually happens. To cast your vote ahead of launching a test, ask all the members of
the team to write down what they think the results will be and why they think those
results will happen. When you get the actual results and can compare it with the votes
that were cast prior to the test being launched then it gives you a chance to see how well
you and your team were actually able to predict the impact to users.
Over time, its always interesting to keep track of who in the team has the best product
instincts. Theres of course a danger in this tactic in that it can create a competitive
environment, but the trick is to make it more about learning together and keeping each
other honest.

Developing a rhythm around data sharing

Finally you might find that there are some very easy ways to set up a rhythm around data.
There may be other things that you can look to piggy back on to set up an expectation
within your team about good habits that you can make part of a recurring rhythm.
For example, rather than setting up user research as needed, you could set up a monthly
user research session which either focuses on your baseline experience or that gets filled
with whatever happens to be going on that week. You might also consider sending out a
monthly update with insights from the various projects that youve been doing. By
starting to set up expectations that data will be collected on a regular basis or looking to
create dashboards that show how metrics might be changing over time given changes that
you are making to the product, you can start to establish a working rhythm in your
organization around data.

<Sidebar: Developing rhythm>

As you begin applying a data informed framework, theres no way to just turn it on so
youll start with selectively. However, your goal should be to make data aware design a
consistent part of product development. To do that, it will require discipline and
vigilance. This is probably true of any decision-making platform, but if youre going to
take a data informed approach to product and design its important to not apply this
framework selectively.
Fundamentally, taking a data informed approach to design is about trying to make
decisions about your product and for your customers using as much information as
possible along the way. Its about trying to mitigate the subjective aspects of product
development by incorporating data from your customers into the process. Its therefore
almost inherent to the nature of a data informed design process that you need to apply it

11
OReilly Media, Inc. 4/8/2015

consistently across all projects for it to work. Weve seen that using data inconsistently
within an organization can often cause confusion or skepticism about the larger process.
One of the critiques that weve heard about using data inconsistently is that the data is
only used when it supports the opinion of the person whos making the decision. This
can definitely happen in organizations and as many of us might know from lab courses in
college or high school, there are often ways to fudge the data or read it in a way that
makes it support the outcome youre looking for. As youre trying to establish the value
of data informed design, its even more critical that they way you leverage and interpret
data is such that it reinforces the objective value that it brings. As a designer or product
manager working in this environment, you might find that there can be a dose of
skepticism around which cells are being tested and why. Therefore, it becomes important
to treat all projects with the same amount of rigor across the board. Crafting solutions
and making decisions in absence of data and user feedback should become the exception
and when you do make those decisions without testing or user feedback youll want to be
able to articulate and defend why that was the case.
Especially when you are trying to establish the methodology of data informed design, it
can be important to showing how you consistently hold yourself accountable to the
methodology of data informed design, but ask others to hold you accountable as well.
This goes back to the first two points about having data informed design be universally
embraced and having a common vocabulary. If you have those two things in place, then
your whole company will be helping to keep you and your team accountable as well.

Principle # 6
Create a learning organization to ensure you perennially grow your
knowledge base.
Data aware design environments work best when they are seen as part of an overall
company culture that encourages learning. Many companies are focusing on creating
learning environments where there is a focus on building individual skills as well as
encouraging development. Data aware design lends itself to this environment because one
of the core attitudes that you need to have about using a data informed framework is to
recognize that you are constantly trying to build up and add to your knowledge about
your customers behavior. Using a data informed approach is fundamentally about
honing your consumer instinct. This is an ongoing activity and every project is an
opportunity to get better at developing that instinct through learning, information and
data.
There are some very practical habits and best practices that can really help to support this
culture. Some things that weve seen be especially effective are:
Sharing results and information broadly
Keeping sharp on both theory and practice
Self awareness/evaluation

Share broadly
If you are actively using a data informed design framework, one of the best things that
you can do for your company is to broadly share the results and to give everyone access

12
OReilly Media, Inc. 4/8/2015

to the data. In a data informed environment it becomes even more important that
designers understand how their designs are being judged and performing. Designers are
often very engaged with user research (which we also classify or include as part of the
data informed framework because the design is still in progress and user research is
often more integrated into the design process). However, there are many places where
once the design is done, the designer disengages from the product development process
and doesnt stay as engaged with the results post launch.
Designers need to get the results from any AB tests that are conducted over their designs.
Getting access to this data is a key shift in their ability to become more strategic.
Understanding (even if only at a high level) how to analyze this data and how to then use
that data to make informed decisions about what to design (and test) next is a key skill.
Habitual sharing/context setting/broad communication, company-wide, of test results -
reinforces that most decisions are influenced by data. It also helps folks to build up
shared knowledge about what works and what doesnt work. All this testing doesnt
really get you much further if youre only using it to make decisions on a micro level.
What should be happening is really taking away learning from these smaller instances
and then using them to hone your instincts about your users over time. The best way to
do this is to vet it with your peers and to subject your testing to debate and discussion.

Theory and practice

It is useful not only to share the results of ongoing work, but to also recognize that
operating in a data informed environment can be new for a number of people. Sharing
results from ongoing work means that youre sharing knowledge and insights which helps
people develop a better understanding when it comes to your particular product or
project, but it can also be helpful to make sure that people who have less experience with
data informed design have an opportunity to build a good foundation about the practice
itself.
Discussions and debate around more general topics and theory of data informed design
can be very helpful to make sure that everyone has a common foundation. Topics might
include everything from how to pick the right technique for different kinds of projects,
the pros and cons of different methodologies, or even details on how to pick a statistically
significant cell size or what a p-value is. The number and level of discussions that you
have on these best practices will be determined by how mature the organization is with
respect to data informed design. If everyone can have a common baseline of knowledge
on the theory behind the practice of data informed design youll find that your
discussions on actual projects will be much richer and much more efficient. Youll also
hopefully avoid some of the pitfalls of not using the data properly (and therefore being
more likely to subjective or mis-interpretation). Youll find plenty of potential topics for
these theory discussions from some of the earlier chapters in this book.

Self awareness/evaluation
As weve mentioned before, one of the biggest benefits of a data informed approach to
design is that it gives you a great tool to use to hone your consumer instinct over time. If
part of the culture of data informed design is about transparency and a more objective
evaluation of your work via the data then as weve hinted before accountability to
holding everyone to a similar standard is important. Its really helpful to create an
environment where you can encourage self-evaluation and to not be afraid of making

13
OReilly Media, Inc. 4/8/2015

mistakes and building out ideas that dont actually move the needle, that is positively
impact the business bottom line by increasing one of the established success metrics (e.g.
user retention or engagement, revenue, etc) so long as you are keeping track of them
and then learning from those experiences.
Weve discussed keeping track of track records as a good way to see how youre actually
doing. Being self aware of when your instinct has differed from what has been reflected
in the data is one of the fastest ways of improving and if youre in a culture where sharing
data and results is very common, then you should have a lot of tools available for you to
evaluate yourself. In the next chapter we will address some of the pitfalls of being in a
data informed environment if you DONT hold yourself accountable to your decisions
and if you dont do this constant self evaluation. For example, over-testing can actually
be a side effect of not having a strong enough instinct for what are the right things to test.
Developing a good instinct for what will give you the best return on your resources
investment (your ROI) will actually come from constant viligence and self-evaluation
as to how well your team is operating and how well your instincts are validated by data.
Not only should this be applied to the individual, but to the organization overall and its
best if you can explicitly think about fine tuning and improving your collective
understanding about your customers.

Ongoing education and establishing a rhythm around data

As weve mentioned before, creating an environment where its clear that ongoing
learning is supported can be one of the most effective tools to getting traction for a
framework like data informed design. This is best done through a series of ongoing
meetings, forums and occasional talks.
Lectures this might be more applicable to larger teams - introducing a topic to a broad
audience (say a mix of tech, product and design) can be an effective way to plant some of
the seeds of data informed design with that group. Lectures are good for:
Introduction of company wide metrics
Overview of best practices and introduction to data informed design processes
Overview of different methods and techniques and when to use them
However, lectures cant be the only means of introducing a new framework to an
organization. Lectures are ofen most useful for offering an introduction or overview.
Follow-up discussions with active project teams at the right time in their process are
essential. So, while its good to use an all-hands or team meeting to talk about the ways
that you might leverage early stage user research, it wont really feel concrete or useful
until you can find a team that is actually at a stage in their project where they can
leverage early stage research and speak directly to what techniques make sense for them
at that point in time.
Project review meetings
One of the most effective meetings for sharing knowledge is a recurring project review
meeting. This might be an ongoing meeting between a broad set of product managers and
designers to review a number of topics around ongoing work. The key is when you invite
people that may not normally work day-to-day with each other. Such meetings provide an
opportunity to share what they are learning in each of their areas to the broader
organization and it helps generally with building up that shared knowledge base of what

14
OReilly Media, Inc. 4/8/2015

is working and what doesnt work when it comes to your users. You might break this
meeting into two sections:
1) Ongoing or upcoming work
2) Results from launched projects
The meeting should encourage a lot of discussion and debate about the projects that are
being shared and it should feel more like a working meeting where everyone is vetting
what is being presented. Presenters should leave the room feeling more confident about
the approach that they are taking and perhaps have some new insights as well. To get you
started here are some of the kinds of things you might present and discuss.
On ongoing or upcoming projects your discussion might include the following:
Hypothesis
o What is the hypothesis?
o Background for what generated the hypothesis and any prior
work/data/research that influenced it.
o Does it resonate and contribute to the overall product goals? More
specifically, is the hypothesis valid?
Success metrics
o What are the success metrics? How were they derived?
o Are they right ones? Will these really measure the validity of the
hypothesis?
o What other things might you measure or use to measure success?
Methodology
o What techniques are you using to collect data for this project? (e.g.
user research, AB testing, etc.)
o What do you hope to learn?
o Are there other methods that could reveal better data? Would some
form of data triangulation give you deeper insights?
o To what extent were methods chosen as a result of extraneous
constraints (e.g., time constraints)?
o Given an infinite amount of time, what would the ideal method(s) be?
Design
o How effective is the design at reflecting the hypothesis or variations of
that hypothesis?
o How does the design support what you want to learn from this project?
Note: This meeting will be different from a design review
where you might be looking at consistency within your design
language, etc. and giving more pure design feedback. Those
meetings are still useful, but probably a separate meeting from
this one.

15
OReilly Media, Inc. 4/8/2015

For projects that have launched and where you are reviewing results, discussion might
revolve around the following:
Summary of methodology
o What methods were used and why? What other methods were
considered? Why were they not selected?
o How effective were the methods used at getting the insights that were
sought?
Results and analysis
o Was the hypothesis proven? Why or why not?
o What did the team learn and what can be applied to other work that is
going on?
o Did the results support any other larger trends that you might have seen
before?
o How do these results compare to prior work?
o What are the next steps?

By reviewing the work that is done in this manner on a regular basis, people can actively
learn about what works and doesnt work. It also encourages people to ask questions and
discuss, thereby learning together and actively learning while doing. When the
organization is newer to data informed design, having a meeting like this is a good way
for the few who might be more comfortable with the concepts to teach those who are less
familiar. When the organization is more mature at using data informed methodologies,
then the discussion can stimulate further learning and perhaps even innovation in
methods and analyses.
We advocate subscribing to user testing and design related blogs and services, and doing
regular searched for slide-decks and white papers online. Although results fro other
companies may not always be transferrable to your company, garnering an understanding
of what other companies have learned can be a good way to stimulate creative ways to
look at your own product(s) and your own practices. Looking to what may already be
written or shared can be a great way to pre-inform or to accelerate acceptance for
adopting a data informed design process in your company. For example, you might be
considering creating a side navigation system, but an article from Next Web suggests side
drawer navigation reduces user engagement. 2 Sharing this data back to your team should
could motivate running a test on your own navigational system.
Through these techniques you can establish your own internal process of data informed
design and ultimately build up your own knowledge base of things that work best for
your company and your product.

2
http://thenextweb.com/dd/2014/04/08/ux-designers-side-drawer-navigation-costing-half-user-
engagement/

16
OReilly Media, Inc. 4/8/2015

Conclusions/Questions to ask yourself

As you set this up, we also advocate you think through:
What are the things that will be challenges for you as an individual to change?
What are some of the things that will be challenging for your organization or
team in a move to a more data aware approach to design?
How does your organization currently communicate?
o Who initiates user testing?
o Who owns the results?
Who do you already have in your organization that might prove helpful for you
to pair with as you try to establish these practices?
What resources including skills do you and your team currently have in place
and what do you need to acquire, create and/or develop?
What existing habits and cultural aspects are already in place that might make a
good foundation upon which to build a culture of data informed design and
decision making?
As we mentioned when we started this chapter, culture is really something that is unique
to every organization and we would never dictate what the right solution might be for
you. You may want to take some of the practical things weve suggested and put them in
place right away. You will find that some of them will work well and endure, whereas
others will not get traction or will require considerable perseverance to establish.
In keeping with our theme of data aware design, weve always taken an experimental
approach to organizational processes too. We advise you to do the same. Try things out
get feedback from users (which may mean teams) as to which things are effective,
valuable and valued and for what reasons, and which are not. See if you can create ways
to measure your success, to measure what works and why. In the case of a cultural
transformation, these measures will not be as clear cut as a simple data pull of the number
of users on your site to reflect retention or acquisition, but you can find other ways to
measure your success. Ask yourself: How well has your common vocabulary taken
hold? Are people using the terms you have introduced in meetings and in company
correspondence in email, messaging and announcements? Are people actively asking to
bring data into the design and product development cycle? Has your time spent selling
the idea reduced? Are your outreach activities meeting with more positive responses?
How much does data informed design get attributed for the success of your product and
how much is it referred to by other organizations in your company?
Throughout this book we emphasise patience, persistence, resilience and tenacity. It will
take a while to see the fruits of your labor paying off (or not). It can take anywhere
upwards of 6 months to really see a change in the mindset of an organization take effect.
The length of time depends on the organization and the business you are in. It will
depend on the size of the organization. Aim for some small successes to start and build
from those. Establish, share and monitor your success metrics. With time, iteration
specification and allies, youll find that data informed design really does have a positive
effect not just on how you build your products, but on your culture as well.

17
OReilly Media, Inc. 4/8/2015

Summary
If you really want to be effective at integrating any kind of framework into your
company, youll want it in your DNA/Culture. Its a hundred times easier to adopt
something when the culture supports it and thats why the success of a data informed
framework is really going to depend on the culture at your company.
If youre interested enough in creating a lasting and deep culture around using a data
aware framework for design in your organization so you can make grounded, user-
centered product decisions, youll also invest some time in building up a culture which
can support it with the right processes as well as the right people. Even if your company
has the mechanics in place to embrace data as part of the decision making process, its
equally important that you have a culture that supports it as well. With respect to data,
there are a handful of tactics that weve seen work really well at different companies and
that are really important have as part of the company DNA. Some of these tactics apply
to carving out or transforming individual roles and responsibilities, some are more
directed at establishing collaborations and support from recruiting groups and teams.
We advocate the following to make this possible:
Aim for shared understanding of the value of user centered as a company
universal which requires you to
Lay the foundations for engaged and transparent conversations about systematic
data collection. You need to create a dialogue, set expectations and develop a
common vocabulary .in order to.
Create alignment, and locate and establish allies and co-partners across
company functions.and.
Build a team through strategic hiring to reflect, amplify and grow the
above..and
Articulate the long term benefits and the ability to do effective pattern finding
across different kinds of data set by establishing a disciplined and consistent
approach to leveraging data that is programmatic, flexible and not rigid ..by
Creating a learning organization to ensure you perennially grow your
knowledge base.

In the next chapter, we will turn to some pitfalls of uncritically overinvesting your faith
in a data driven approach. Part of being data informed and data aware is knowing what
can go horribly wrong.

Additional resources
Content

18