Better ways to find, with faceted search

Taken from Ornament with Two Women Blindfolded by Sebald Beham (1527)
Taken from Ornament with Two Women Blindfolded by Sebald Beham (1527)

I like faceted search. I think eBay is my favourite search engine of all. But search facets have dropped out of fashion in interface design. If we got back into search facets, then interfaces of all kinds would be better.

I often remind myself that my usage patterns are not representative. I might even have that tattooed on the insides of my eyelids, just to be sure. I also know eBay isn’t where the cool kids hang out anymore but, as the bathwater drains away, we should remain mindful of the baby. In this case the bathwater is market share, and the baby is quick but also precise online search mechanisms. Or perhaps not even search, but find.

It’s funny that the term ‘search-engine’ stuck, given all the ways this functionality could have been described. And what you call things over-influences what they do—ask Carphone Warehouse. Okay sure, searching might describe what users do (or did back then), but it doesn’t describe what users want. It’s foolhardy to make sweeping generalisations about users but I think I’m on reasonably safe ground when I assert that users don’t want to be looking; they want to have found. The result of the search is where the dopamine is, not the search itself. There’s no thrill-of-the-chase. Anyway.

Faceted search is a good, solid, finding-stuff paradigm. Enter something that very broadly describes the thing you’re looking for, and then use facets to turn a huge number of hits into a small selection of precise, usable results. You could type in ‘car’ or pick the equivalent category, then choose age, number of seats, location, price range and so on, and fairly quickly you’ll have found a shortlist of cars fitting your exacting preferences. Faceted search is a really good way to buy shoes, since a user is interested in shoes only of a certain style and size, give or take.

Free-text searching, even using fairly hardcore search operators, doesn’t compare. I’m a hardened user of search operators, wrapping critical terms in quote-marks and so on. But this requires me to remember all this syntax, which I can’t, and popular search engines seem to adhere to it less strictly with every passing year. Google Shopping might be good for browsing, but is nearly unusable for anything precise, seeing as the only facets are condition, price, colour and seller. Amazon’s search, while more finely faceted, seems to be deliberately imprecise to encourage users to browse. I’d guess they’re measuring dwell-times over particular pictures, to gauge what to recommend later. But again, browsing isn’t finding.

Google Maps is infuriating for its lack of useful facets. It can plot intricately precise routes from A to B, but can only facet these routes to exclude tolls, motorways or ferries. I can’t avoid tight turns or blind corners, which would be useful in something bigger than a car, but maybe the dataset doesn’t exist. That said, what if I wanted to include something in my route? The cheapest fuel-stop, maybe, a nice bite to eat, or a scenic spot for a leg-stretch. The satnav I bought nearly two decades ago could do this. If I were minded to, I could research it all myself (by searching separately) and set intermediate destinations along my route, but the voice in my head says: you’re the search-engine, you figure it out.

There used to be lots of ways to find online stuff: not just search but directories, curated lists, webrings and whatnot. The web outgrew many of these approaches but, even so, there also used to be various search paradigms—like some specifically for queries in the form of questions. Now there seems to be only one style of search: jabbing in keywords and hoping for the best. It has contorted online content into a keyword-dominant vernacular: keyword-jabbing, and optimising for it, has made the contemporary web even more categorised than the directories of old. Funny how that came back around. It’s influencing how we behave.

A poignant, if a little NSFW, example:

Data on what we search for, pay for and click on is being used to predict our desires and funnel us bespoke(ish) porn.

At first blush, it might seem like this kind of micro-targeting would just turbo-boost the internet’s existing trajectory, making it even easier for people to find and embrace a diversity of bodies and fetishes. But there’s a fundamental shift here from a world in which we explore a passive sea of content to a world in which porn actively explores and prescribes itself to us. Because this shift stems from deep financial upheaval in the adult industry, the content pushed upon us will likely increasingly reflect what is most profitable, not what is most widely desirable. It could well become narrowing, or at least channelling, rather than broadening.

This approach is also how ‘natural’-language interfaces—the ones you bark at from across a room—have evolved. Alexa is precise only if you are able to ask for something that the machine can recognise as being near-unique, such as a particular song. It can’t have an ongoing conversation with you about taking a large set of search results and turning it into a handful of sensible choices.

More widespread faceted search would be an opportunity to take ubiquitous keyword-based searching and explode it into new, useful, powerful interfaces: both useable and precise. eBay has stuck to this model and it’s really good. LinkedIn uses it a lot for finding contacts, jobs and so forth, and it works well: finding something as specific as which of your contacts who know someone at Microsoft would be arduous without it. Holiday sites—hotels, flights and the like—also give users plenty of facets to turn all the world’s options into a manageable shortlist. Users are used to doing it; it’s just not available widely.

That’s how I’d want to use a natural-language search but Siri, or whichever, can’t handle it. I can’t even remember what I have to shout at my phone to use its natural-language doodah: that’s how little use I have for such a blunt implement. It’d be more interesting for these interfaces—all interfaces—to take the 1990s multi-dropdown facet model like eBay and think about how to bring this level of specificity to contemporary users’ lives. Simply put, facets reflect how people think: must-haves versus negotiables. Certainties versus unknowns. A find-engine would do the same.

Subscribe to doing.digital in your newsreader with this RSS feed. (what does this mean?)

Latest