“The data is semantically rich,” said Lou Rosenfeld, information architecture consultant and founder of Rosenfeld Media. “Users are telling you want they want in their own words.”
Search analytics – an often under-utilized and free resource – allow you to “carry on a conversation with your customers” by listening to their needs and measuring how well your site meets those desires. You can also analyze search queries and results to improve UX design, site navigation, search performance, content strategy, and – as many online retailers have done with search analytics – product offerings. “If you just take the top 50 most frequent queries… and then throw that data into a spreadsheet and play with it, you’re going to get an unbelievable amount of insight in an hour,” he said.
Lou, who is lecturing about Search Analytics for Your Site this October at Web 2.0 Expo New York, recently spoke to us about how companies can give themselves the upper edge using site search analytics (SSA). (You can also read in-depth about the topic in his book of the same title.)
Case in point: the Financial Times uses SSA as a predictive tool to determine what they should be covering. The media company’s site analytics team monitors queries looking for spikes of personal and company names. “If there’s a discrepancy [between popular queries and recent coverage] they can bring it to the editorial team and they can see if there’s something breaking they don’t know about,” Lou said.
Meanwhile, online retailers have used SSA to build “desire paths.” When a retailer sees that a search result for “shirt” is frequently followed by “pants,” for instance, they can create a system to move users horizontally across their product line by listing suggestions on product pages based on how the user arrived. In this case, SSA is used to “predict where people want to find themselves next.”
(For a better understanding of what a good desire path is, imagine a nicely paved but unused pathway next to a worn path through grass walkers in the area would rather take – SSA helps you avoid wasting time on paths users don’t care for.)
Beyond the development team, all departments in an organization can improve performance using SSA. For instance, the content team can see if their pages match what users truly look for. Marketing can study the different needs of audience segments through what they want. And designers can use it to make gradual improvements to the site by observing how people maneuver through search.
In addition to the recent phone interview with Lou, he was kind enough to answer a few follow-up questions, found below. If you have a question about SSA, you can chat him up on Twitter or talk to him in person at Web 2.0 Expo New York this fall. (Use code BLG20 to save 20% on registration.)
Kaitlin: Do you recommend a particular search technology on a site?
Lou: Do you mean a particular search engine? No, choosing a specific engine is absolutely the worst place to begin improving search performance. It’s better to determine what your functional requirements are, and then choose the tool as part of the broader solution that best meets those requirements. Site search analytics is a great way to help determine which user-facing features your functional requirements should include.
Kaitlin: What are you using for search query analysis? Is it just Excel or are their other tools?
Lou: Most analytics applications (and some search engines) include a minimal set of generic reports, such as the most frequent queries that retrieve zero results. Those reports are a nice starting point, but you’ll really want to create reports that are tailored to your own organization’s needs. That’ll often mean working with raw data–which you may be able to export from your analytics app into your favorite database or spreadsheet so you can work with and analyze it directly. In the book, I show an example of how Netflix uses a humble Excel spreadsheet to do a powerful custom analysis of its data.
Kaitlin: Am I right in assuming that queries such as “contact info” or “faq page” are common? Are there other common queries across sites?
Lou: Sure; you might also encounter terms that prepopulate search fields (like, um, “search”), variants of the name of your organization (e.g., queries for “IBM” on the IBM site), or URLs (people mix up the address and search fields). Though one may think that these examples all suggest some particularly dumb users, a deeper look at the data–especially session data–might indicate that these are actually the fault of dumb designers.
Kaitlin: Your book talks about “best bets” aka “recommended links”. Can you explain what these are and what the purpose of a recommended link is?
Lou: Best bets are hand-selected results matched to common queries. We use them when we want to augment (or when we don’t trust) the results that our search engines are likely to retrieve. For example, your site’s search engine may retrieve many different contact information pages, each corresponding to its various business units, departments, and locations. Which one is best? Your search engine is just a robot–it won’t know. To make sure searchers had a fighting chance at retrieving what’s likely the most relevant information, you’d make your organization’s main contact page the best bet result for the query “contact information”.
Kaitlin: Many sites use auto-complete for search queries. Where should a site get their list of suggested completion terms?
Lou: Ideally, there are standard vocabularies in place that can used. For example, Netflix populates its list with useful terms it already owns: the names of movies and TV shows. Those terms can be complemented by common search queries. Either way, make sure you’re using terms that are not only likely to match your users’ queries, but that have been scrubbed–raw query data will include typos and the naughty and nasty stuff that you don’t want to expose to users.