Implementing
Search for Your
Knowledge Base
Implementing powerful search for your knowledge base is easy to understand
in theory, but can be intimidating in practice. This ebook walks through the
steps required, outlines available options, and provides guidelines to ensure
smooth implementation.
READING TIME: 20 MINUTES
Table of contents
Introduction ................................................................................................................3
Build vs. buy: Which implementation makes sense for you?.................................4
Out-of-the box search ......................................................................................5
Cloud-based search..........................................................................................6
Next steps: The four phases of implementing search ...........................................7
Planning: Designing search and establishing timelines .........................................9
Implementation: Realizing your vision ....................................................................11
Testing: Qualifying and quantifying search quality ...............................................13
Maintenance: Long-term optimization and operations .......................................15
2
Introduction
Congratulations on making the decision to implement search for your knowledge base! You’re
likely already aware of the benefits of having effective search, such as fewer inbound support
tickets, streamlined customer support, and a better overall customer experience.
You may even have seen stats like this: 45 percent of companies offering web and mobile
self-service reported an increase in site traffic and reduced phone inquiries.
However, when it comes time to implement your search, you may encounter a whole set of
issues you’ve never had to deal with before. Fortunately, you’ve come to the right place for help.
In this ebook we’ll talk about the different search implementation options available for your
knowledge base and take you through the four phases of search implementation:
1. Planning
2. Implementation
3. Testing
4. Maintenance
Where do I start?
3
Build vs. buy
Which implementation is right for you?
The type of search implementation you choose will depend on the goals you’ve defined for
your search, the resources you have available, and your unique working environment.
Let’s take a quick look at two different types of search implementation: out-of-the-box content
management system (CMS search and cloud-based search. The basic differences are outlined
below, with more in-depth explanation on the pages that follow.
Out of Cloud
the box search
No work required Faster implementation
One size fits all No ongoing maintenance costs
Cannot be customized Built-in relevance and language
Weak text analysis and relevance recognition technology
Unable to customize search results Customizable via dashboard
Weak analytics Analytics dashboard built into product
4
Build vs. buy
Out-of-the-box CMS
During the planning process, some companies may feel that the out-of-the-box search built
into their CMS is adequate for their customers and support agents. While in some cases this
may be true, many companies soon discover that out-of-the-box search is inherently inflexible
and cannot be customized for the specific needs or desires of your support team.
As a one-size-fits-all solution, CMS search relevance models cannot be changed, leaving
support teams guessing about how they might influence which search results appear at the top
of the list for a given query. Perhaps more importantly, out-of-the-box search is often built with
little to no support for common language problems (such as spelling correction, synonym
recognition, or basic language parsing technology). Companies that want a customized search
solution with more advanced relevance and language modeling technology will need to invest
in building new search on their own or look to a third party to build and host search.
5
Build vs. buy
Cloud search
If you want the benefits of an advanced search engine and language modeling technology but
don't have the engineering time to build and subsequently maintain search, then cloud-based
search is probably the best option for you.
Cloud search solutions allow complete control over your search experience, but free your
engineers up from building advanced relevance models or developing natural language
processing for your search engine. Hosted search also saves your engineers from dealing with
recurring software updates and ongoing server maintenance, and often comes with advanced
analytics and optimization dashboards.
password rest
FAQ
Documents
Videos Documentation: Steps to rest
Forum
FAQ: Can I reset?
6
Next steps
The four phases of implementing search
Now that you understand which implementation options are available, it’s time to learn more
about the four phases of search implementation: planning, implementation, testing, and main-
tenance. Below are the basic elements of each stage, with more in-depth explanation of each
on the pages that follow.
Planning
Gather stakeholders, design search, set goals, propose timeline
Implementation
Indexing, frontend coding, backend server configuration
Testing
Check search result quality, test common language problems
Maintenance
Optimize search results, maintain servers and software
7
Planning
What is the goal?
The first step in the planning phase is to look at your own people, processes, and systems and
develop a set of best practices for your own knowledge base to ensure a smooth
implementation. At the end of planning, you should know how your new search will work, who
will be responsible for executing what, and how long the project will take.
Who is involved?
Once these are in place, you can move on to gathering your project team members. Resources
you’ll need to involve include:
VP of support to champion and guide
Software engineer for development work
UX designer for frontend work
Customer service representatives and support content creation team to provide input
on requirements
8
Planning
How will search change?
With your team in place, you can move on to one of the most important steps in the planning
process — deciding how you want your search to work. Why is this step so critical? Because what
you define in this step determines how visitors will experience your search when they visit your
knowledge base.
For example, you might want customers to be able to search by topic, content type, or date/time
frame, as well as have the ability to filter results by topic, tags (such as “getting started” or
“troubleshooting”), content type, or reviews.
To help customers find your content faster, you might also want to implement an autocomplete
experience that features results rather than suggested queries. This removes a step in the search
process and immediately gets visitors to their destination.
Once you’ve determined how you want your search to work, it’s time to think about how you’ll index
your knowledge base content — a topic discussed in depth in the implementation section.
9
Planning
Who will do what, and when?
In the last step of the planning process, you’ll estimate project timelines, as well as develop
training and communication schedules. For the overall project timeline, expect to allocate
between one week for a very simple paid solution to several months for an open source search
solution.
You’ll also want to allocate time for training your customer service representatives and your
support content creation team so they can understand how customers will be searching and
the process they’ll need to follow when adding new content to your knowledge base.
Finally, you’ll need to develop a communications strategy in order to let people in your company
know that the new search is coming, and how much easier it will be for them and their custom-
ers to quickly find the information they need.
What’s
DESIGN INDEXING FRONT END TRAINING
happening?
Who’s
Support Engineering Engineering Support
involved?
10
Implementation
How can I index my website?
Data feed: Establish an XML, CSV, or RSS file transfer protocol
If you’re working with a third-party provider, they may ask you to pass on all of the information
you want to index via a data feed. This process involves setting up an FTP transfer of your
information in XML, CSV, or RSS format every 24 hours.
Aside from being cumbersome to configure and maintain, data feeds often leave your search
index out of sync with your live website offerings, since website changes are only communicat-
ed to the provider when the FTP transfer goes through.
API integration: Build database connection between website and search engine
An API will accomplish the same as a data feed, but without the annoying lag — information is
passed on instantaneously to your search engine. API integrations also allow complete control
over your engine schema, which allows you to more precisely control how your search experi-
ence will work. If someone on your team is experienced with API integrations, then this might
be the way to go.
Web crawler: Website information automatically extracted from HTML
If you don’t want the hassle of dealing with a full-scale API integration, or you don’t have the
technical expertise to do so, you can get the same results using a web crawler.
When working with a web crawler, you simply provide a URL and the web crawler automatically
extracts website information from the HTML. In the case of Elastic App Search, you can control
exactly what information is extracted by specifying website information with custom Elastic
<meta> tags, which speak directly to the web crawler and allow you to create a custom search
engine index.
11
Implementation
Do I need to buy servers?
Once you’ve indexed your content, you’re ready for the next step in the implementation phase:
resolving server and operational issues.
There are two reasons to monitor your servers if you self-manage your index:
1. To ensure that you have enough servers to handle traffic
2. To resolve any speed issues that might arise from customers searching your knowledge
base in other parts of the world
One of the many advantages of cloud-based search is that these issues are entirely handled by
your provider. On the other hand, if you’re building your own search, you’ll need to buy servers
in data centers around the world to serve international audiences at speed. Additionally, you’ll
need to ensure you have the correct amount of servers to handle your fluctuating search
engine and website traffic.
How long will the frontend work take?
In this step, you’ll take what you’ve planned and work with designers and frontend engineers to
put everything in place for the user interface.
Allocate at least one week of work time for designers and frontend engineers to put this into
place. As with all phases, be sure to add in some buffer, as tasks often end up taking longer
than anticipated.
12
Testing
How do search results look for my most
important queries?
In order to ensure your search is working properly, you should run tests of how your search
handles your top 50100 queries. The key here is analyzing relevance and asking yourself: Do
these search results match my expectations? Are these the most helpful answers for this
search? If not, you should customize the results for these queries until they match expectations.
How does my search handle common
language issues?
In addition to relevance, your team should investigate how your search handles common
language issues such as misspellings and phrase matching. Below are some of the most
common situations to test.
Misspellings corrects “templaet” to “template”
Bigrams recognizes equivalence of “e mail” and “email”
Dashes & spacing recognizes equivalence of “i phone” and “iphone”
Stemming recognizes equivalence of “starter” and “started”
Phrase matching distinguishes between “iOS 9 apps” and iOS 8 apps”
13
Maintenance
What is required to maintain my index structure?
Once you’ve determined how your index will be structured, you’ll need to have every piece of
content created moving forward conform to that indexing structure.
If you have an open source solution, stay on top of updates by keeping in close touch with
software developers. Keep in mind that one engineer could spend almost a full week making
sure their search works with software updates.
If you’re using a web crawler, maintenance will be a lot simpler. When you add <meta> tags to
your content, changes are made on the template level. Once your site is recrawled, that
content is automatically indexed. Finally, ensure that engineers have mapped your CMS to the
correct indexing structure.
14
Maintenance
How can we optimize our search over time?
To optimize your search and case deflection rates over time, be sure to keep track of KPIs that
matter most. These could be clickthrough rate, search exit rate, percentage of searches with
no results, or other metrics. The key is not to get bogged down with metrics, but to choose
carefully those that demonstrate the most impact and continually monitor how these affect
your inbound ticket volume.
Finally, optimize your search by continuously customizing your search engine.
Tune your algorithm to control what Customize results for individual queries
impacts the order of search results. to feature the most important articles.
FIELD
My payments are failing
popularity
Failed payments may be happening because...
Changing payments information
To change your payment information, log in to...
IMPACT 2.5
FIELD What payments are accepted?
We accept credit/debit cards, PayPal, bitcoin...
average rating
Can I make payments late?
IMPACT 5 Unfortunately, We cannot accept late payments...
15
About Elastic
Elastic is a search company. As the creators of the Elastic Stack (Elasticsearch, Kibana,
Beats, and Logstash), Elastic builds self-managed and SaaS offerings that make data
usable in real time and at scale for use cases like application search, site search, enter-
prise search, logging, APM, metrics, security, business analytics, and many more.
Founded in 2012, Elastic is a distributed company with Elasticians working in countries
around the world. Learn more at [Link].
