Meet the User Halfway: Search Shouldn't Be Passive-Aggressive
I bet that if a coworker treated you the way most search engines do, you'd punch him in the nose.
Request: "Can you get me the TPS forms, please?"
Reply: "We don't have any TPS forms. But if you'd like, I can get you10,000 documents that use the letters ‘TPS.'"
Request: "May I have the TPS reports, please?"
Reply: "Oh, sure! Here you are."
And yet, this kind of nonsense is just what we've come to expect from search engines. They return completely different results depending on the specific words in the query; they return piles of content when you want one simple answer; and they make you play guessing games, rather than helping you get to the information you need.
You should expect better from "intelligent search." Your users certainly do.
Search with meaning, not just words: I recently entered two nearly identical searches into a support site: "[Product] is running slowly," and "[Product] performance problems." That sounds like the same general question to me, but I got completely different answers-some reasonable, some odd, but none of the results in the first few pages were the same.
Vendors sometimes talk about synonym lists, as if simply substituting one word for another will make searches work well. Or, sometimes, knowledgebase authors have to guess what query keywords to stick inside their documents. The reality is, search needs to return results based on what queries mean, not just the words they contain.
Fortunately, search for enterprise service and support is focused on a specific topic domain-we're not searching the whole Internet-so meaning-based search can work. Users should be able to enter "running slowly" or "taking too long" and know that they'll get consistent results about performance issues-which means search behaves like a helpful colleague, not a passive-aggressive pest.
Return precise results for high-value queries: Sometimes, searchers are researching, looking for lots of potentially relevant information. And sometimes, there is no single answer to a question.
But often, there really is a right answer. For questions like, What storage devices are compatible with this software? or How do I troubleshoot error 37?, the knowledgebase usually holds a definitive answer.
Queries about these definitive answers should be treated as special cases. Rather than just returning a normal results list, the system should take the user directly to the correct answer. Then, if the answer isn't what was desired, the system can send the user to the most probable results list. In some cases, the system needs to ask a few more questions to select that answer; if so, the user should be guided easily through that process.
By the same token, common queries should work well, whether or not there is a single answer. For the most frequent questions customers ask, results should be tuned to move highly relevant content to the top of the results list, while irrelevant content is eliminated. Of course, subject experts shouldn't spend hours tuning search results for every query, but making sure the top 10 or 20 search queries work really well provides tremendous benefit.
Guide users to the answers: Let's say a search query doesn't deliver the information the user needs. What's next?
Most of the time, it's "game over." The system smugly waits for another search. Users get no help in refining or improving the search, so they add words, subtract words and replace words to see if they can come up with a query that works better. They don't even get a hint about what their options are, or what refinements would be likely to return better answers.
This is maddeningly frustrating-and it's completely avoidable. If the system knows what documents are available, and what they mean, it can also make some suggestions for improving the query. These options guide users through the process of refining their queries, zeroing in on the information they need.
There's no magic. But technology can help: "Intelligent search" is a great concept... but most of us right now would settle for search that isn't actively rude.
The three best practices discussed here-searching by meaning, returning a definitive answer when available and guiding users through the process of search-can't be 100% fully automated. (Even IBM's "Jeopardy"-playing Watson software said Toronto was a US city.) If a vendor tells you all this happens without human intervention, hold on to your wallet.
But the right KM technology makes it easy for service and support professionals to implement these best practices with a minimum of effort, generating tremendous return on the investment in their time.
So if your search struggles with these issues, put your search engine in therapy: make sure it meets your users halfway.
Best Practices for Search Experience
If Your Users...
Describe the same issue or question in different words
Sometimes ask questions that have a definitive best answer
Struggle to refine their queries
Make Sure the Search...
Translates customer words to their underlying concepts
Presents the answer immediately to the user
Provides specific suggestions for improvement
Otherwise...
Searches that mean the same thing may return wildly different results
Customers will grow frustrated rummaging through search results
They may just give up without finding what they need
Consona Knowledge Management is a fully realized KM application built for the enterprise, KCS Verified, and especially designed to meet the needs of those with complex service and support queries. More than 1,000 customers worldwide use Consona CRM solutions to manage process efficiencies, drive revenue and enable exceptional cross-channel customer experiences. Visit crm.consona.com.