A Problem With Search Forms

Golly – it’s about time I wrote down something about user experience design, seeing as this is what this blog is suppose to be about.

I’ve been doing some work for a site re-design, starting with user testing 24 people over two weeks. We asked them (a wide demographic) to use some currently live sites to see how they got on with them. Some people tested the client’s current site, others one of their competitors. There was only one task in the half hour or so we gave them to do this: order a product. We also showed them a couple of Web 2.0-style funky Ajax interfaces to see how they got on with things like dynamic search and asynchronous interactions – for that is what is what we are planning to do for the redesign.

The nature of the sites under test was that they are all predicated on quite complex search forms. Most, including our clients, have about ten to twelve pretty heterogeneous input fields (dates, ranges, etc. using dropdowns, checkboxes, and other things) that the user has the option to fill out to get the product they want to look at, and hopefully buy.

One of the things that really stood out was how few people actually found what they wanted. In fact, the success rate (as defined by whether the user could get to the end of the order process without giving up or being assisted) was appalling. Our client’s site scored about 35% – the others as low as 20%. So about 80% would have either given up, or got so confused as to have no choice but to give up. Yet these web sites account for about half of the companies’ business! All have offline retail channels as well.

We theorised various reasons for the failures, but one of the most significant was that users generally filled out most, if not all of the search form – optional fields and all. This is understandable: in life if we’re given a form (or something that looks like a form), the convention is to fill it in. Doing so generally means you stand a better chance of getting what you want. Leaving out information is not a good thing because you may get the wrong, or no, results: leave your name off the job application and you’re screwed, leave the date off the cheque and it bounces…

This logic does not work for search and retrieval systems. On the user tests, we saw people filling in the forms and constructing narrow searches which therefore produced either no results, or forced the system to provide inappropriate substitutions which were typically not understood by users – they didn’t see the message at the top of the screen that told them the results had been substituted.* This (and some other problems with the forms) then led to a multitude of issue. Hence the failure rates mentioned earlier.

There would seem to be two possible solutions to this. One is to make the search form simpler or otherwise discourage narrow searches (at least at first), the other is to use faceted search, for example like Kelkoo does. In fact, we intend to combine both: the former technique for the initial search form on the home page, followed by faceted search on the results page. The trick will be to get the balance right and allow a flexible enough search with appropriate familiarity (users didn’t get on with wizzy controls like sliders, incidentally) while avoiding presenting the user with zero results messages, or relying on equally damaging substitutions.

We’ll have to see how this pans out, but one thing’s for certain: it’s going to be tough to make the situation worse.

* One middle-aged user responded to this by narrowing her search still further in an attempt to get results. Not surprisingly, she sat back after about three complete failures and announced that she thought the site just didn’t have what she wanted. In fact it did – she just needed to remove a couple of optional constraints.