Enterprise search relevancy - in the eye of the user

There's a common misconception that the single factor in Google's relevancy algorithms is link structure (also described as PageRank). And since the enterprise environment doesn't typically have the same link structure found on the Web, the presumption goes, Google search must not work well in the enterprise environment.

Let's set the record straight. Google's enterprise search algorithms rely on hundreds of factors to determine relevancy. PageRank is one of a number of document quality variables that, combined with factors that measure how well the query matches each document, determine the right result for a user. By no means is PageRank the only (or dominant) factor in determining which results are the most relevant. Those of us on the quality engineering team have been coming up with new ideas since 1998; the best of them have ended up in our ranking algorithms.

But we're not comfortable with theoretical arguments, so don't take our word for it. We surveyed the customers of our yellow box, the Google Search Appliance, and here's what they had to say:

• About 75% of customers responding switched to Google from another search provider's product
• Over 50% of customers responding switched to Google due to poor relevancy from their prior search provider
• Over 90% would recommend the Google Search Appliance to another company like theirs

In case you missed it, some Raytheon search experts presented their findings on an enterprise search evaluation at the Semantic Technology Conference. They ultimately chose Google to provide integrated enterprise search across their desktop and network content repositories. Most impressive was the response from their user base who participated in the survey: 84% found the "right" answer (the answer they were looking for) in the top 3 results.

Let's be clear - there's plenty more to do. We're happy to see that a large majority of users are easily finding what they're looking for. It's just that we believe we can do better. Eighty-four percent still means 3 out of 20 queries aren't optimal, and we're not satisfied with that. But we also believe that asking each customer to write their own relevancy algorithms would amount to surrendering and saying "here, you try!" Instead, we're working closely with customers and partners to assess where and when our algorithms work best, and where they need improvement. We welcome you to join us and help improve enterprise search for everyone.

David Elworthy, Lead Engineer
Enterprise Search Quality Team