OK, I know it’s a bit of a provocative (and frivolous) title, but it was either that or “Anti-patterns for Search”, which is somewhat less transparent… Anyway, the point is that although this site focuses on sharing best practices for search usability and user-centred design in general, we sometimes learn more by studying examples of the opposite, i.e. poor practice. The thing about UX design is that when executed well it fades into the background, letting us get on with the task at hand and not worry about the machinery of the interaction. It isn’t until we encounter flawed design that we are jolted out of our flow and forced to make choices that don’t seem to fit with either our expectations or the natural course of our activity.
So to that end, I am seeking examples of flawed design in the world of search. Have you any stories you’d like to share? Any examples of ‘search that sucks’? I can certainly think of a few which would make good subject matter. But before climbing on my own soapbox, I’d like to hear what other have to say.
If it’s something you’re happy to share publicly, feel free to post it here. If it’s a little more sensitive, just send me an email and I’ll make sure your confidentiality is maintained.
If I get enough replies I’ll post a summary here, along with any design insights and lessons learned.
Information Interaction 
One of the things that drives me crazy is an insistence that users must drill to get any facets. Amazon is most guilty of this. You can’t even sort without drilling to some potentially very nebulous category, and you have no facets (not even simple high-coverage ones like brand or price) to help clarify when you get some weird polysemy.
Couple that with the fact that there _is_ actually bad data in Amazon because they sell on behalf of others who curate their data like a 5 year old “curates” a hamburger.
You’d think Jeff Bezos would have noticed by now =)
Ah, but that is also an outcome of our good friend the ‘clarify & refine’ pattern:
https://isquared.wordpress.com/2013/01/22/designing-search-part-6-interacting-with-results/
Pretty much all heterogeneous catalogue sites have this problem, and most solve it this way. In their defence they do execute it using just a single category (‘Department’) menu, so any drilling required is fairly lightweight and easily discoverable. But hey, who am I to defend Amazon 🙂
In a perfect world you would have already presented the right filters — at least for the more popular queries worth giving extra “best guess” treatment. Something like _Amazon’s_ Mechanical Turk could help with that 🙂
C.A.R. only works well when you have good data and not an overwhelming amount of polysemy. Amazon doesn’t have good data because some of their data aren’t curated by them. eBay doesn’t even more so — because _none_ of their data are curated by them. You often get a really bushy noisy taxonomy for that reason alone.
eBay is a grand master at building taxonomies, and they’ve been polishing them for years. They carefully make all accessory categories siblings not children to cope with a related problem I call the “accessory problem.” But C.A.R. won’t work for them despite all this polishing.
Lots of polysemy across all products severely limits the user’s ability to clarify based on the good taxonomy. Millions of users putting random stuff in wrong categories somewhat reliably. Polysemy^2 to begin with. Nightmares.
eBay and a few others nail it, though. They also have more data to glean this from than God, but try a few queries. Now try it on Amazon. eBay does a lot of “best guessing” and they’re very good at guessing.
And it doesn’t explain why I have to sort and can’t use other facets with good coverage like price to make the category tree, in turn, less noisy. I’m not sure all users would know to do that, and the risk is that they might not do what you want to yield more facets — but nothing is perfect, right?
Given that Amazon invented Mechanical Turk it’s pretty unforgivable they don’t do more. And I’d at least show some other basic facets like price, and allow for sorting.
Thoughts?
I call it the “diverse category” problem, that is the more product categories the e-tailer carries, the harder it is to give good search results. You don’t have enough context from the customer to steer them where they want to go.
For example if you only sold shoes, it’s easier. You have a couple of facets that narrow your offering down right away. Expand your offering to include apparel and things start getting harder: size may include 10 1/2, XL, and dress size 8. I worked on a site where we tried to figure out what to do when the user searched for “dress”; did they mean women’s dresses? girl’s dresses? men’s dress shirts? etc.
Amazon is probably the most degenerate case for this. Not that they are perfect, it’s a hard problem to solve.
Actually the issue you refer to of showing facets earlier in the dialogue does crop up fairly regularly – I recall an electronics distributor I worked with a few years ago that wanted to do exactly that, by presenting ‘lowest common denominator’ facets right off the bat, and not wait for category disambiguation.
But there were 2 gotchas. First, it did produce some wacky results: a query for ‘silver’ or something else fairly generic just brought back a crazy selection from across the inventory, with no logical ordering or structure, and no common attributes for sorting (hence Amazon;s strategy on this) and no way to filter other than on things like delivery time or price or brand.
Second, they did not want to compete on price, (or at least to be seen as such), so were not comfortable introducing price as a critierion so early in the dialogue. That just left delivery time (which for many components just wasn’t relevant anyway) & brand. As I recall, in the end they felt this was all becoming a bit of a guessing game, and went back to encouraging some sort of clarification strategy instead.
OK, how about something less controversial. I go to a web site and I search for “brown shoes.” I see pages and pages of shoes that are available in everything _except_ brown, but have photos of the brown variation on the product page. This one is pretty pervasive. It’s a solvable problem, but few seem interested. Looks like good merchandising to me if you show what I’m looking for.
You get things like this:
http://www1.macys.com/search/index.ognc?SearchTarget=*&cm_sp=navigation-_-top_nav-_-search&Keyword=+brown+shoes&KEYWORD_GO_BUTTON.x=-946&KEYWORD_GO_BUTTON.y=-78&KEYWORD_GO_BUTTON=KEYWORD_GO_BUTTON
Ouch, I didn’t even expect to see this other problem. Apparently someone set up a synonym shoes pumps. But even ignoring that, I don’t see brown shoes, do I? More colors available? Now I have to pogo-stick. And I’d better use the back button, because breadcrumbs rarely persist facet selections.
Sorry Macys, I know your category navigation is different and totally awesome. The only reason I brought you up is to compare and contrast using the same site and the same products. See:
http://www1.macys.com/shop/shoes/all-womens-shoes?id=56233&edge=hybrid#!fn=COLOR_NORMAL%3DBrown%26sortBy%3DORIGINAL%26productsPerPage%3D40&!qvp=iqvp
Search related to music fall into the category you describe. In my view, it is because music is a complex domain which people query for different reasons. So, to give an example you can be looking for a song because: you want to hear it, you want to play it, or you want to get basic facts (dates, album, etc…), and these are, in my view, the three main reasons, sure there will be much material available, sadly the vast majority of that material will be replication and plagiarism of a few sources. Worse, if the reason you are looking for a song is more obscure, perhaps you are curious to know what instruments are being played, or if there is a specific musical phrase in it, or basically any unusual reason, it becomes mission impossible 1) because of the amount of material which literally litters the web regarding the main three reasons mentioned above, and 2) because your mainstream search is not able to grasp and target those unusual intents.
Thanks for stopping by, Gery. I’m a bit of a muso myself so feel your pain 🙂 Did you have a particular site in mind, or are you referring to web search generally?
I got an email from “Rakuten” today, actually. Looking at their site I’d say they’re an example of many things I personally dislike in search. Where do I start?
No faceted search for their “stores.” Everything is a “please wait,” often for 8 seconds or more. _Both_ sliders and multi-select for price? Department facet that isn’t hierarchical + weird hierarchical category. Out of stock items are ranked highly, not removed or kicked to bottom by the ranker. You have to click “update” before selections are submitted. Facet counts are not updated, so if you select across facets you will inevitably reach dead ends.
When you do reach a dead end it looks like they just clear the category filter hoping “maybe that will work?”
Lots to study here.
Indeed. That department hierarchy (if that’s what it is) is quite something. How does it relate to Category? Bit wierd also the way they change the facet ordering (by frequency for some facets, but by alpha for others). I think they need to prioritise those facets, e.g. by stage of the user journey, and show the lower priority ones closed by default, etc. But as you say that’s just the start, really.
As an analyst and consultant, I frequently seek information on products and vendors on vendors’ websites. The information I am seeking is often very specific. After an attempt to find what I am looking for using navigation, I move on to the search box. With 30+ years of search experience, I have a broad and deep arsenal of search models (syntax and commands) to draw upon.
When I am seeking information on enterprise search products or about a vendor’s customers or partners, and cannot figure out where or whether that exists on their web site, it makes me deeply skeptical about the company and their product. When I brought this to the attention of one of the top tier search vendors several years ago, they had a number of excuses: lack of marketing budget for their own site, length of time to develop a good interface, lack of human resources to build a good content repository. What really got to me was their insistence that their search engine was easy to deploy and out of the box you could expect stellar results. My conclusion, “except for your own organization.”
Buyers beware of the claims and look closely at what you get out of the box and what it takes to implement and deliver a really great “demo” version offered during “test drives.” You can see really bad search interface examples by just visiting search product sites.