Reading the e-Consultancy interview with Lou Rosenfeld on the importance of site search analytics, I was reminded of when I was product manager of intranet (or enterprise) search at the BBC.
It was back in 2002. People complained that search was broken, but we had neither quantitative nor qualitative data to analyse.
After consulting with my colleague Martin Belam, who at the time was looking after search on bbc.co.uk, I put in place a system for capturing search engine usage data.
In a nutshell, this included where people came from, what they searched for, and where they went.
We also conducted user research to gather qualitative data about the experience of using the intranet and what it was like to find things on it.
After only a few weeks, the usage data started to settle and patterns began to emerge.
We noticed that search term frequency had a long tail (though we didn’t know that’s what it was called at the time!).
The top 25 search terms accounted for half of all searches. The top 50 terms accounted for 75% of all searches made.
We tried searching for the top 10 terms, including “training”, “expenses”, “ariel” (the BBC in-house weekly newspaper) and “jobs”.
The results were worrying to say the least. Few of them returned the result expected on the first page of search results.
If we could find the most likely link (or links) for each of our top search terms, and return that as the top result, we should be able to guarantee providing most people with what they’re looking for.
So we set about building what became the Best Bets system. (How we did this I will write about another time, but needless to say it involved an SQL database, some asp code and some hacking in to the Microsoft Site Server system.)
We then set about working with the numerous intranet site managers to determine what the best link should be for each of the top 50 search terms. Further usage analysis, research and testing showed that this working. Success! People were clicking on the Best Bets result almost 95% of the time we had a search term match.
Over time we extended to cover the top 100 search terms, then 200.
Regular usage analysis showed us when new terms appeared and we took editorial decisions on whether they should have a Best Bet.
In spite of the success of the Best Bets system, we certainly couldn’t rest on our laurels. This was an improvement for sure, but as we well knew, it was only a part of the overall search and navigation user experience. There was plenty more work to be done!
Any search engine worth its salt these days comes with best bets functionality as standard. My advice is to take the time to study and understand how people are using your search engine. Test the top search terms yourself to see what the results look like.
Maintaining a best bets system is a relatively low cost exercise, and encourages regular monitoring of search usage and behaviour, which in turn will provide excellent information about your users’ expectations and your intranet’s findability performance.
A word of caution from experience: bear in mind that the formatting of the best bet in the search results page should be minimally different from the natural results, otherwise it risks ending up in the user’s blind spot and be regarded as a “Sponsored Link” rather than something genuinely useful.