Before you ask, Sherlock Holmes isn’t standing in for the designer, he’s the user.

God I hate designing filters.

I hate designing filters (also known as faceted search) not because I don’t know how to craft kickass filters, but because to be really effective at it, I need three things I almost never get offered by the person asking me to add filters:

  • Time to find out what filters users are currently using
  • Research to tell me what users would like to filter on but can’t
  • The number of items that could be in the unfiltered results

Time is important because depending on the cyclical nature of the business, the filters customers use during holiday shopping season might not be the same ones they use during the spring doldrums. (Also true: if in fields like financial, the filters customers use when the market is up might not be the ones they use when the market is down.) If no one is using a specific filter set over the course of a year, it’s worth asking why it’s there… but you’ve got to have a year’s worth of data (or the cyclical time field appropriate to the business) to analyze.

Note that “time” also implies “someone has realized we’re going to need this and put the appropriate hooks into the appropriate components so the web metrics software of choice is actually capturing the filters being used on a regular basis by session so I know what peeps are choosing in the filters and in what combinations.” Which really only happens if you’re rebuilding an existing flow and the previous team liked building metrics hooks.

If I don’t know what filters are useful, I might rip out the rare useful ones. (I don’t have the data to prove it because it’s locked up at my last job, but we found in repeated analysis and tests that the majority of people don’t filter at all.)

Research is important because there’s a weird thing I’ve noticed about people: everyone thinks that everyone else uses the same search techniques and mental map as they do. Just like most people wipe their butts the way they were taught and don’t realize half the population does it a different way (strangely pretty ok for work link), most people either were taught or stumbled into a few ways to search and use them most of the time.

Recently I had to explain to a younger designer how Boolean search works. Developers over a certain age were gobsmacked; all we had when the internet was invented was Boolean search.

I regularly stump my developers with “but what if the user doesn’t know what the thing is named?”, I think because the bulk of a developer’s searches are for named functions, features, etc. of programming languages and “There was a dog book out a few months ago with a blue cover, do you know what it’s called?” isn’t their normal pattern.

Because they are so different, the user’s search behavior patterns impact not only what I allow  filter but the interactions of the filters themselves.  So when the only information I have about filters is that the project manager or the quality engineer or the developer thinks we need a Boolean faceted search that includes seventy six facets, I get suspicious.

And data about the data set is important because I have lost count of the number of times I’ve been asked to add filters to a table or data set that — for the vast vast majority of customers — will have four items in it.

Four.

Which the team acknowledged would all be visible at once.

And, in some cases didn’t need to be compared or used together.

Eventually they’d come around (through some discreet questioning) to the idea that filters might be overkill.

So to review: the following three things make designing filters better.

  • Time to find out what filters users are currently using (or a previous design team that thought ahead and made sure that data was being captured).
  • Research to tell me what users would like to filter on but can’t (because we don’t all search the same way).
  • The number of items that could be in the unfiltered results (to determine if this filter thing is needed at all).

And without them, no matter how well I think I’ve designed my filters, what I’m really asking my users to do is figure out how to gauge the usefulness of the data we’ve provided based on my own conjecture, gut instinct, and the instincts and conjecture of my team.

The result too often is that my users land in a place where the filters are useless, and the results doubly so.

Side note: this isn’t what I set out to write about. But we’ll get there. 

Two is the loneliest number

Last post, we talked about a website that offered express shipping for $12 even though it would arrive the same day as free shipping.  The lesson there was “check your edge cases.”

In that same vein, today we offer up this gem:

A screenshot of the banner at the top of my Doordash login. It brags that they have "no delivery fees from over 2 restaurants near you" in attempt to sell me their DashPass service.

Now admittedly, I live in the sticks, 30+ miles outside of Philadelphia, where most of the restaurants are mom-and-pop shops in strip malls, and “ethnic” means “Taco Bell”. 1)For ethical reasons I prefer GrubHub over Door Dash. (Not that their hands are clean either.) However, GrubHub offers me 1/3 of the number of food options that Door Dash does. And sometimes you need chicken korma.

So on one hand, not having a lot of “free delivery” options is not a surprise.

On the other hand, offering me “no delivery fees from over two restaurants near you” (emphasis mine) shifts their message from “hey look at these options!” to “you poor thing, bless your heart.”2)The real kicker is that sixteen DashPass restaurants will deliver to me. That’s actually pretty significant. And if it said “there are over 15 restaurants in your area…” they might have gotten my attention.

In making the DashPass sound unavailable, Door Dash made their own product sound, well, unavailable, and uninviting.  I’m not sure what the threshold for “yes, this sounds inviting” is (that’s what usability testing is for) but I do know that “two” wasn’t it.

When you check your edge cases and their messaging, consider how your user affected by the edges will feel. (Can this be read as “My brand is an asshole”? Is that your goal?) Adjust your content accordingly. 3)Oh, and make sure your app can count.

Not all of your customers are worth courting, but all of them deserve to feel like they matter.

Notes   [ + ]

1.For ethical reasons I prefer GrubHub over Door Dash. (Not that their hands are clean either.) However, GrubHub offers me 1/3 of the number of food options that Door Dash does. And sometimes you need chicken korma.
2.The real kicker is that sixteen DashPass restaurants will deliver to me. That’s actually pretty significant. And if it said “there are over 15 restaurants in your area…” they might have gotten my attention.
3.Oh, and make sure your app can count.

A light post for a light week

Screenshot of the shipping section of a form, where ground shipping will arrive June 12, or for thirteen dollars more expedited shipping will arrive the same date

Nothing says “We tested the edge cases” like offering me free regular shipping that arrives on June 12 for free or expedited shipping that costs $12.95 and also arrives on June 12.

True story: the state of Pennsylvania has shifted to a print-your-own-registration-forms format for car registration. (Your car must be registered for it to be legally drivable.)

When they used to mail your registration, they charged $2 for a second copy to be stuck in the envelope.

Now, they still offer the $2 fee for a second copy… but you’re downloading a PDF and printing it yourself. (You can print as many copies as you have printer toner and paper.) And if you do pay the $2, they give you the same PDF they gave you the first time.

So it’s really become a “are you paying attention?” tax on drivers.

Form asks how many duplicate registration cards you would like at a cost of $2 each, and then warns that the duplicates will e included in the document at the end of the transaction. No duplicates will be sent in the mail. So you're spending $2 per duplicate to have extra length added to the pdf file when you could just print the pdf file multiple times.
Or I could just print two copies of the PDF

Test your edge cases and your processes, folks.