This is the extended version of a presentation Helen Lippell gave at Enterprise Search Europe in London in April 2014. It describes a holistic 5 step methodology for triaging a search application that people are complaining about, and offers a prescription for starting to make things better.
Market Analysis in the 5 Largest Economic Countries in Southeast Asia.pdf
Make Enterprise Search Less Broken
1. Enterprise Search -
How to triage problems
quickly and prescribe the
right medicine
Helen Lippell Enterprise Search Europe, 30 April 2014
2. Slide 2
Search and discovery
•Search is important - plenty of demand for scalable,
powerful tools
•Search is evolving
• Big Data (or is it the less-snappy “Lots of Different Data”)
• Structured data and semantics (not just searching free text
for keywords)
• Search tech underpins other applications e.g. BI and analytics
• Rise of open-source
• But, for many, search within the enterprise still isn’t what it
could be - Sometimes people just want to find their company’s
holiday form
3. Slide 3
Agenda
•Methodology for triaging search problems
–With an emphasis on workability
–Laying foundations for bigger improvements
•Cumulative experience of years in the trenches, both
as in insider and an outsider
4. Slide 4
“Search is rubbish!”
• Has there been an outbreak of “Something must be
done”itis?
• Is this true?
• What’s not working?
• Who is saying this?
• Users
• Senior management
• Other stakeholders
• Noisy stakeholders
• You?
5. Slide 5
Start from where you are
• Who owns Search?
• …If anyone?
• E.g. IT/ Comms / Editorial
• Who owns the data (and where is it)?
• Who owns the technology?
• Who, if anyone, is accountable?
• Useful to understand how the land lies when looking
for support and focus
6.
7. Slide 7
How bad a state is your search in?
If your search is known colloquially among users as the Random Page
Generator, it needs some TLC
8. Slide 8
1. User needs
2. Content
3. User interface
4. Technology
5. People and process
9. •It’s a holistic approach
•Which areas get most of your attention will depend
on your own time, interests and who else you can call
on
•Some recommendations might seem straightforward
or obvious – but if search was easy there wouldn’t
still be user dissatisfaction…
Slide 9
10. 1. User needs
• Arguably the most important area to look at
• Data, data and more data
• Get query logs
• As much as you can to start with
• Analyse them
• Repeat…with a sustainable process ((read “Search
Analytics for your Site”)
• Determine user intent
• Group common task searches together to aid
prioritisation of content review
• Aim is to derive insights to shape and complement
everything else I’ll cover
Slide 10
11. 2. Content
• Content is king (but you might be in a republic)
– Search experience function of the quality of the content that is indexed
• Do a content audit
– Duplicates / Redundant, Outdated, Trivial / Bad metadata
• Start with a sample, even if only small
• Content howlers can be great for powerful anecdotal
evidence of problems:
• “lorem ipsum”
• “title goes here”
• “deletethis”
• “Arial bold 14”
• “Helen’s test content” (embarrassing)
Slide 11
12. Metadata
• A “Minimum Viable Product” to aim for should be
good, unique page titles and descriptions/summaries
• Also consider:
– Tagging for different user groups
– Last modified date (and other date fields)
– Owner/author/creator
– Subject tags (whether from a taxonomy or user-
generated)
Slide 12
13. Content strategy and information architecture
• Search is one part of a wider information ecosystem
– It shouldn’t look like the tube map on a strike day
• Review:
• Overall IA
• Other tools like A-Z
• How easy it is to access search from elsewhere
• Are there different search functions?
Slide 13
14. 3. User interface
• No need to reinvent the wheel, lots of prior art in search UI
design thinking
• Focus on quick wins
– May be limited capability and appetite for redesign projects
and template changes
• Seek out guerilla feedback
– Informal research fine for triage
• Eliminate page clutter
– Default settings? Long-forgotten feature requests?
• Review information displayed for each result
• Facets and filters
• Aim is to build confidence!
Slide 14
15. 4. Technology
• How easy is it to tinker with the platform?
• Has it been configured well? At all?
– Configuration must suit what is being indexed eg document
length, format
• Is anything documented?
– What features are switched on? What were the out-of-the-box
settings?
• Is there a test environment? If not, get one
• Review features and defaults e.g. Boolean searching,
synonyms, indexing, spellcheck
• Benchmark, and repeat often
Slide 15
16. 5. People and process
• Hearts and minds
• Ongoing quality management
• Accountability
• User research
• Dedicated people resource
• Regular editorial and tech tasks
• Co-ordination across teams
• Support and guidance to content authors
• Stakeholder workshops
• Listen to moans
• …but don’t allow a big grumblefest
• Elicit unmet needs, offer reassurance
Slide 16
17. Slide 17
Prescription
• You can do triage relatively quickly
–Prioritise areas for action
–Identify recommendations – and be prepared to follow
them through
• Delineate quick wins and longer term strategy
• Shout about successes
• Iterate
• Strategy and culture change take longer
• Ideally, Search should have a dedicated team
• If Search is very political, consider outside help