If you’ve ever registered for an event or bought tickets online, then you’ve seen a countdown like this one:
Limiting a person’s checkout time isn’t the norm for most ecommerce sites, but in the ticketing world we’ve come to expect it. Sellers want to create urgency for buyers, and they don’t want to lose sales to people who could hold onto reservations indefinitely without making a purchase. As buyers, we want fairness. We want equal opportunity to snag a limited commodity, and so once an order has been started, everyone receives the same amount of time to make a decision and complete it.
But how much time? And how can designers of such experiences make sure they’re being fair to users while still meeting the requirement for urgency?
These were the questions on my mind after two recent checkout experiences with Eventbrite. During the first, while registering for an upcoming UX conference, I noticed that Eventbrite’s timer started at 8 minutes, and in a passing curiosity I wondered how the company’s designers arrived at that number. A short while later, while signing up for a local technology meetup, I again saw the timer began at 8 minutes … except this registration had only 4 fields:
By comparison, buying a ticket for the UX conference involved 22 fields, including the entry of credit card information. Surely if there was more for me to enter, I should’ve been given more time, right? I didn’t need all 480 seconds, but I probably didn’t represent the average user anyway. Was this default fair to slower typists and other people (such as disabled users) who might take longer?
Now, I don’t have any insight into Eventbrite’s design contexts or decisions, though I did learn that in its settings the company gives event managers the option of extending the default time limit, “if you’ve created a lot of custom questions for your attendees.” If Eventbrite were to share how it determined 8 minutes was the sweet spot, I’d be super interested in that case study. (Hey, I asked.)
But we don’t need Eventbrite — or even any users, really — to estimate whether 8 minutes was enough time for me to register for that UX conference.
Keystroke-level analysis is a nifty, simple method for determining how long we can expect average, skilled users to complete tasks. Also known as keystroke-level modeling, the approach is built on usability research that details how long it takes to execute the smallest of actions — a keystroke (0.2 seconds), a click (0.1 seconds), a moment to think (1.2 seconds). All we need to do is break down a task into its micro-moments and add everything up.
I ran a couple of calculations. In the first, I assumed the speed of an average skilled typist, and I made room for additional thinking and review time. I also used my own personal input, and I assumed that a skilled user would hit Tab to advance to the next form field. To my surprise, the completion of 22 fields added up to only 70.1 seconds.
In the second calculation, instead of hitting Tab my “skilled user” was assumed to reach for the mouse, click into the next field, and then reach back for the keyboard between every entry. That raised the total time to 1 minute 44 seconds.
I expected 8 minutes might be a squeeze, but lo and behold, it was plenty of time. Right? Eh … probably. Again, the research behind KLM assumes a user that knows exactly what she’s doing. It can adjust for more intensive tasks (for instance, entering my credit card number and checking it takes longer than typing my name), but that said, it’s a general benchmark with varying mileage for tasks that require extra thought or labor, or users with special needs. There are also contextual circumstances KLM doesn't necessarily account for, e.g. the time it takes to retrieve my credit card if I can remember where I left it.
Still, instead of wildly guessing, I now have some idea of what to expect if I were to continue with user testing. Measured against my own baseline at least, somebody could take between 4 and 7 times as long to complete the conference’s Eventbrite form and still have seconds if not minutes to spare. Eight minutes seems like a safe bet, and a balance between fairness and urgency.
(As an aside, I’m curious to know whether anybody is updating KLM research for mobile, voice, or touch interactions.)
My personal takeaways from this little experiment:
Keystroke-level modeling is a lesser-known asset for the UX toolbox, a easy way to estimate task times and make comparisons between alternative designs without needing users. It can be useful, as long as we’re mindful of its limitations.
Not all form fields are created equally. If I found myself designing a checkout process with a countdown, I’d give careful consideration to the level of challenge and the amount of information that my users need to process or input. Consider what range would be necessary to accommodate as many people as possible, whether “skilled” and fully abled or not.
I’d also keep in mind external factors. Let’s not assume that users are sitting in a quiet space without interruption, able to focus on the checkout with robotic focus.
And finally, if you’ve got access to people and data, use them! Counting keystrokes is no substitute for watching real people go through a task. In the case of the timed checkout, it would also be revealing to know from site analytics how many users were running out of time.
Too many words for such a small thing, perhaps. But to quote Charles Eames: “The details are not the details. They make the design.”