Imagine your website analytics tell you that 68% of people abandon the checkout process right at the shipping address step. That's a valuable data point, but it's also a dead end if that's all you have: it tells you where people drop off, but not why. Is the form confusing? Does that step take too long to load? Is there a required field people don't know how to fill in? Does the continue button not display properly on mobile? Analytics can pinpoint the symptom with pixel-level precision and still leave you completely blind about the cause. A usability test is the tool that fills that gap: instead of looking at numbers, you watch real people using your site and trying to accomplish something specific.
What a usability test actually is
A usability test involves sitting down (in person or on a video call) with a handful of people who represent your actual customer type, giving them a specific task they should be able to complete on your site ("find this t-shirt and add it to your cart", "find the clinic's contact phone number", "compare the price of these two services"), and silently observing how they work through it, without helping or correcting them along the way. What gets measured isn't just whether they complete the task, but where they hesitate, what they read twice, which button they look for and can't find, what they say out loud without being prompted (the "think aloud" technique delivers the most insight, because it reveals the logic a person follows while navigating, not just the final outcome).
Why analytics alone isn't enough
Web analytics is extraordinary at telling you the "what" and the "how much": how many people come in, how many drop off at each step, how much time they spend on each page. But it's practically blind to the "why", which is exactly the question you need answered to fix the problem. You can spend weeks changing a button colour or rewriting a piece of copy based on hypotheses about why people abandon, and get it right by luck or get it wrong without ever knowing, because you've never actually watched a real person run into the specific problem. A usability test with just five well-chosen people usually reveals, in a single one-hour session, friction points that months of data analysis wouldn't have identified with the same clarity.
Why five users (and not fifty)
There's a widely held idea in product design, backed by landmark usability research, that five well-selected users catch around 85% of a website or online store's most important usability problems. Adding more users beyond that brings diminishing returns: the sixth and seventh users tend to repeat the same problems you already saw with the first five, rather than surfacing new findings. This has a very practical implication for a small business: you don't need an enterprise-sized user-research budget to draw useful conclusions. Five real people, well chosen, are enough for a solid first test.
How to set up a simple, cheap one
The first step is defining concrete, realistic tasks, not open-ended questions like "what do you think of the site?". A well-framed task is specific and actionable: "imagine you want to book an appointment for next Thursday afternoon, go ahead and do it" is far more useful than "browse the site for a bit and tell me what you think". A concrete task lets you observe the actual decision-making process, not a generic opinion people tend to soften out of politeness.
The second step is recruiting participants. You don't need a specialised agency or a big budget: five people who resemble your actual customer (in age, in familiarity with the internet, in interest in your type of product) are enough, and they can often be recruited from second-degree acquaintances, existing customers willing to help in exchange for a small incentive, or even a simple call-out on social media.
The third step is the session itself: 30 to 45 minutes per person is usually enough, asking them to think aloud while trying to complete each task, without stepping in or correcting them even when you see them making a mistake (the temptation to help is what ruins a usability test the most, because it papers over exactly the problem you wanted to uncover). Recording the screen and audio (with their consent) lets you review afterwards, at your own pace, the moments with the most hesitation or confusion.
The fourth step is analysing the patterns that repeat across the five people, not each individual comment on its own. If a single person gets confused by something, it might just be their own quirk; if three out of five get confused at the exact same point, you've got a real problem worth fixing before launch, or before continuing to invest in traffic toward a site that carries that friction.
What kind of friction this catches that numbers don't show
A usability test tends to surface very specific problems analytics never points to directly: button copy that people interpret differently from what you intended (someone who won't click "Request information" because they think it implies committing to a purchase), expectations about where a specific piece of information should live that don't match where you actually placed it, checkout steps people find logical on desktop but confusing on mobile, and above all, moments of hesitation or distrust that never turn into a measurable drop-off but still create a worse experience than you think you're delivering.
Before launch and after launch: the two moments this pays off most
Usability testing adds value at two distinct moments. Before launching a new site or a major redesign, run on a clickable prototype (it doesn't need to be finished), it lets you catch and fix friction problems at minimal cost, well before those problems affect real sales. After launch, once analytics has already pointed you to a specific drop-off point (like the 68% example on the shipping form), it lets you understand the exact cause of that symptom and fix it precisely, instead of testing changes blindly one after another hoping one of them works.
Frequently asked questions
Do I need special tools to run a usability test?
Not for a simple version. A video call with screen-sharing and recording (most free tools already support this) and a clear list of tasks is enough. Specialised paid platforms exist for larger studies, but they're not necessary for a first test with five people.
How do I choose participants if I don't have a budget for professional recruiting?
Look for people who resemble your target customer in relevant ways (age, familiarity with technology, interest in your product), avoiding, where possible, coworkers or family members who already know your site too well. Real customers willing to help in exchange for a small incentive are usually the most accessible and most representative source.
How often should I repeat a usability test?
It's worth repeating every time you ship a major change (a redesign, a new checkout process, a key new feature) and, more generally, at least once a year even without major changes, because user expectations and the reference standards they compare your site against shift over time even if your site doesn't.