Search is the enemy of IA, redux
by Pete Bell
Pete Bell
My original post on this topic generated a lot of interest that made me want to unpack the question, “Is search the enemy of IA?”
# # #
Back in the early aughties, Jared Spool and his User Interface Engineering team were testing retail websites when they noticed a strange result. Their methodology certainly wasn’t at fault. As always, they followed meticulous protocols. For example, they gave each shopper hard cash for each purchase, and pre-interviewed each individual to uncover a genuine shopping goal (Oh, you’re going on a cruise this winter? So do you need a new bathing suit?), all so they would behave as much like real shoppers as is possible in a lab.
Each shopper attempted tasks on a dozen or so different sites, and at the end of the tests, the shoppers evaluated various aspects of their experience. One line item asked about the quality of the search, and the clear winner was bananarepublic.com. But here’s the curious part: that site was the only one without a search box. It only had navigation.
This was a completely unexpected outcome. UIE hadn’t designed the test to compare search and navigation — that was an accident. The shoppers simply had the most success finding what they wanted at bananarepublic.com, and when they were later asked about the quality of search on the site, they misremembered that they had found what they wanted through a search box.
[Update from Jared via email: "It wasn't clear whether the people who rated search as high were thinking they had used a search capability on the site, or that they equated the word 'search' with finding what they wanted and not the specific element of technology. Later on, we realized that laymen can't tell us what the search engine in a site is or when they use it. They use search, find, and navigate interchangeably." ]
So what can we learn from that result? First, a little more context about the state of online retail in the late 90s/ early aughties, since a lot has changed since then.
1) Quite a few top 100 online retail sites did not yet have search boxes. Search technology was new and expensive, and there were still retailers that just put PDFs of their catalogs online. Search was an investment they would stage after a more elementary one, like an ecommerce platform. In other words, bananarepublic.com hadn’t foregone search out of genius — they just hadn’t gotten around to it yet. In fact, not long after the test, they added their first search box.
2) Sites that did have search typically had bad search. And that wasn’t necessarily the fault of their search technology. Good search is an interdisciplinary affair that requires good design, good IA, good business processes, etc. And back then, all of that was still being invented for search. So the other sites being tested started from a low bar compared to today.
3) The field of IA was much further along than search was. (Yahoo! was still more popular than Google, and they expected you to find what you wanted by browsing.) Moreover, IA for online retail was quite mature. After all, customers already knew how to shop in Banana Republic at the mall, so it wasn’t a stretch to map their existing mental models to an online flow. And that laid the ground for some excellent navigation experiences. (I want to be careful not to collapse IA into navigation and usability — they are heavily related, but people with several different roles collaborate on them.)
So here’s my conclusion: good IA is better than good IA + bad search. That’s because if you give shoppers the choice between browsing and search — well, there actually is no choice for many visitors, because all they will notice is the search box, and the rest of the site might as well not exist.
Now, I heard some completely reasonable objections last time I posted: search can’t be the enemy of IA because search is a part of IA; or, search is not the end of the task — it’s one part of the task that could flow into faceted browsing, sorting, etc. All totally true, assuming that the search experience is good. But that turns out to be a big assumption. And this is my point: in reality, bad search is a black hole. For many visitors, it makes the rest of your site invisible, sucks people in, gives them a horrible experience, and causes them to give up and leave. And that’s what happened at bananarepublic.com’s competitors. [As Mauricio Freitas points out in “People browse by search,” it’s gotten even worse than that.]
The top 100 online retailers have graduated beyond bad search. But they are years ahead of most media sites, government sites, intranet search, business applications with search boxes, etc. That’s because it’s so easy to measure in dollars the results of improving the shopping experience that it became easy for retailers to justify the investment in inventing a better search experience. (And again, I’m referring to search broadly as technology, IA, usability, business processes, etc. Moreover, I’m not including web search in this discussion since it’s so different on so many facets that nearly everyone treats it as its own animal.)
Several years ago, one of the most senior technologists of a major US intelligence agency gave a talk at Discover, our annual user conference. The retailers packed the room, expecting a sneak peek at the future of search from Q’s James Bond lab. Instead, he told them he had traveled to the conference from DC to learn from them. The retailers were the innovators.
So that’s why bad search is the enemy of good IA. But part 2 to my last post was supposed to be the interesting part, because it was about the innovative parts of good search + good IA. Unfortunately, I bogged everyone done in my trollish title. So coming up, part 2 redux!
[Actually, there's one more important lesson to draw from the UIE test: "search" no longer means the search box. People say "search" to refer to all aspects of discovering what they want. I've seen technology analysts and vendors tear their hair out over defining search, since accepting this colloquial definition leaves them with a big loss in precision. For awhile, the going term was Information Access. But increasingly, people have given up the fight, and now search means both the search box as well as the bigger umbrella of features that relate to finding and discovery.]
on February 24, 2010 at 1:46 pm
Permalink
Nice set of posts. Don’t forget that keyword search was almost killed by overuse (overabuse) of keyword density. then Google re-taught people to keyword search after the likes of AltaVisita almost killed that instinct forever. Keyword search was so bad that Yahoo and its directory seemed like the future. Users had probably decided that search boxes sucked when that study was run.
Nowadays, I like the trend of users using more keywords in their keyword search phrases. It’s almost like narrowing in advance of getting the results themselves.