Back to the basics of site searching

By David Walker (Google profile)

Until recently, Web site management defined itself through its ability to avoid the unmissable in pursuit of the unaffordable. Organisations conducted $200,000 graphic design exercises on sites with Fantasyland business plans, or spent $1 million to "personalise" shallow and uninformative content. There was more money to be made selling complicated software and the whimsies of former CD-ROM designers than there was in insisting that sites do the basics right.

Since the dot-com crash, though, the basics are receiving a little more attention. And some site managers paying renewed attention to a mid-2000 Forrester Research report. This document reminded sites that most of them make a hash of their on-site search facility - that little search box that lets you search through their own content.

On-site search is one of four key methods by which users find information. Usability guru Jakob Nielsen estimates more than half of all users are "search-dominant", using search as the main road to the content they want. But the evidence suggests that you should work hard to satisfy users through the first three methods (page-by-page navigation, featured content and tables of contents). Back in 1997, usability consultancy User Interface Engineering claimed that using a site's internal search engine actually cut users' chances of finding what they wanted. UIE's report was titled Why On-Site Searching Stinks.

On-site search has improved a little since 1997, but Why On-Site Searching Stinks still looks a pretty relevant document. Hence the title of the new Forrester study: Must Search Stink?

Right up-front, the Forrester study pinpoints a huge barrier to site search improvement: the gulf between the reality of current site search and the way e-commerce executives see it.

According to Forrester, e-commerce executives see search as crucial. Forty-five of fifty executives interviewed by Forrester describe it as "extremely important" or "very important". Many rate it as the top way to get to their site's content. They also believe they implement search well: two-thirds told Forrester their search is either "useful" or "very useful".

Now given that just three years ago UIE rated search as worse than useless, you might wonder whether these guys are kidding themselves. Sure enough, they are. When Forrester tested the wonderful on-site search facilities they'd implemented, hardly any received even a pass mark:

• They often didn't find the content that best matched what the user wanted.
• They usually failed to find synonyms (pages about "notebook" when the user asked for "laptop") and almost always broke down in the face of spelling mistakes ("notevook" instead of "notbook")
• They rarely put all the best content on the first page of results.
• They typically returned more irrelevant results than useful ones.
• They often provided vague page titles and descriptions that left users wondering which links would produce useful content.

So how has this gap opened between management belief and reality?

Forrester notes that the executives brought trouble on themselves: few spent much time choosing their search technology, overseeing its implementation or testing its results. (At least one was beyond redemption, complaining that his users refused to search using the "right" keywords.)

The available technology also limits a site-builder's options. Most of the available search solutions require too much expertise from users, return too many results and can't be easily made to understand where various pages sit in a site's information structure. Some solutions cost far more than you'd think - Forrester estimates up to $A500,000 for the software and its implementation within an existing system. Forrester asserts the cost can be worth it, but I doubt there's a site in Australia which could justify that expense.

But Forrester also offers two much cheaper partial fixes for today's sorry site searching. First, create site search interfaces that give users more confidence. Second, give your pages proper titles and "meta-information" which will describe the pages for the search engine. (This process should cost you somewhere between $5 and $12 per page; a few outside agencies will do it for you.)

To recast a Jakob Nielsen utterance, the average Web user is not on the page she wants to be on. She'll keep using search to try and get there. The least you can do is devote some effort to helping her.