Product discovery and UX design: a buyer's guide / Start here
The signs your product has a UX problem
A usability problem is rarely reported directly. The evidence shows up in other places: support tickets that start with 'how do I', trial users who stop at the same step, customers who export everything to a spreadsheet and work there instead, sales demos that only work with an expert operating the product. Each signal has an ordinary explanation the team prefers. Together they identify the problem. The good news is that confirming a UX problem is cheap, and the returns on fixing a real one are among the best documented in software.
Published August 8, 2026. Editorial.
Key takeaways
- The evidence of a UX problem is already in your support queue, your funnel data, and your sales calls. It is rarely labeled as design.
- The strongest single signal is repetition: many users failing at the same place, asking the same question, or building the same workaround.
- Confirming the problem is cheap: watching five real users attempt the core task finds most of the usability problems in a design.
- Fixing a confirmed problem pays measurably. Usability redesigns in Nielsen Norman Group's surveys averaged large improvements in the KPIs they targeted.
Nobody files a ticket that says "your information architecture is wrong". They file a ticket that says "how do I re-send an invoice", and they file it every day, and each one gets answered politely and closed, and nobody ever sees the full pattern for what it is: a record of exactly where the product fails its users.
Here are the signals worth trusting, where to find each one, and how to confirm what they are telling you before spending money on a redesign.
The signals, in rough order of reliability
Support tickets that ask how, not whether. A "does your product do X" ticket is a marketing gap. A "how do I do X" ticket, when X is something the product does, is a design gap: the capability exists and the interface hid it. Pull a month of tickets and sort them. If the same how-do-I question appears again and again, you have found a screen that is not doing its job, and you know exactly which one.
Drop-off concentrated at one step. Funnel data shows clearly where people give up. A gradual loss of users through a long flow is normal. A sudden large drop at one step identifies the problem: whatever that screen asks, people cannot or will not answer it. The step itself tells you what to test first.
The spreadsheet that replaces the product. When users export your data and run their real work in a spreadsheet, they are telling you the product's model of their work is wrong. They wanted the data; they did not want your workflow. The same signal appears as browser bookmarks that deep-link past your navigation, or as the one power user in the customer's office who operates the product on behalf of everyone else.
Demos that need an expert. If sales cannot let a prospect touch the product, and onboarding requires a call, the interface cannot explain itself. Watch what the sales engineer says while operating it: every phrase like "so what this screen is really showing you" is an explanation for something the design failed to communicate.
Feature usage that never spreads. Analytics showing most users using a small part of the product can mean a focused product. It can also mean the rest is undiscoverable. The sign is the gap between what customers asked for and what they use: when people requested a capability, got it, and still do not use it, the problem is almost never demand.
The signals that are not UX problems
Two similar-looking problems waste redesign money regularly, so name them before spending. Low signups with good activation is a marketing and positioning problem: the people who arrive succeed, there are just not enough of them, and no interface change fixes having too few people at the first step. And high activation with high churn weeks later usually means the product solved a problem people do not have often enough, which is a discovery question about what you built, not a usability question about how. A redesign aimed at either of these will use up three months and change nothing, because the failure happens before users reach the screens.
Confirming it before you spend
Every signal above has an innocent explanation, which is exactly why teams argue about them for months. The way to end the argument is watching real users.
The method is small and public. Recruit a handful of actual users, give them the real task the funnel says they fail at, and watch without helping. Jakob Nielsen's long-standing finding is that about five users find most of the usability problems in a design, roughly 85% of them, which is why the confirmation step costs days, not months. The goal is to see the problem rather than measure it precisely, because a team that has watched three people fail the same way stops arguing about whether the problem exists.
Be strict about one thing: watch behavior, not opinions. Users are polite. They will say the product is fine and then fail the task in front of you. What they do is the data.
What a real fix looks like
A confirmed UX problem is rarely fixed by changing colors and styles, because the look was rarely the problem. The fix usually changes how the product works: reordering a flow around how users think, cutting choices that were only there because they were easy to add, making the important number easy to read and giving less space to the rest.
That last pattern is common in analytics-heavy products. A platform can hold genuinely powerful analytics that everyday users cannot act on, because the screens show everything with equal emphasis. The fix starts by auditing which numbers people actually use, rebuilding the experience around that smaller set, and moving the rest behind a drill-down. Nothing about the underlying data changed. What changed is that users could finally see it.
And the returns on this class of fix are unusually well documented. Nielsen Norman Group's surveys of usability redesign projects found large average improvements in the KPIs the redesigns targeted: 83% in their recent survey, 135% in the earlier one. Your result will be your own, but the direction is clear. Removing the places users fail tends to show up in the numbers that depend on users succeeding.
Deciding what to do next
If the signals are ambiguous, buy the cheap confirmation first: a short round of watched sessions on the core task. If they are strong, the question is scope: one flow, or the product's whole model of the work. One flow is a focused engagement. A wrong model is closer to re-discovery, covered in what product discovery delivers. Either way, agree the metric before the redesign starts, so six months later the argument is not about taste. Choosing who does the work is its own decision, covered in how to choose a product design agency.
Most teams are in the same situation: the evidence arrives daily, spread across inboxes where no single person sees the whole pattern. Collect it once, watch five users, and the argument usually ends in an afternoon.
Common questions
How do I know if my product needs a UX redesign?
Look for repetition in the evidence you already have: support tickets asking how to do things the product already does, trial users dropping off at the same step, customers exporting to spreadsheets and working there, demos that fail without an expert operating the product. Then confirm cheaply by watching about five real users attempt the failing task. Repetition plus watched failure is a reliable diagnosis; any single signal alone is not.
How many users do you need for a usability test?
About five per round, per Jakob Nielsen's well-known finding that five users find roughly 85% of usability problems. The point is iteration, not sample size: run a small round, fix what it found, run another. Save large samples for measurement questions, like whether a redesign changed a conversion rate, not for finding the problems.
What metrics indicate poor UX?
The useful ones are specific: the share of support tickets that begin with 'how do I', drop-off concentrated at one funnel step, time for a new user to complete the core task, and the gap between features customers requested and features they use. Broad scores like NPS change for many reasons and are poor at locating a design problem even when they reflect one.
Is a UI refresh the same as fixing a UX problem?
No, and confusing them is expensive. A refresh changes how screens look. A UX fix changes whether people can do their work: reordering a flow, cutting choices, making the key information easy to read. If users fail at a task, restyling the screen they fail on produces a better-looking failure. The watched sessions that confirmed the problem will also tell you which kind of fix you need.
How long does it take to confirm a suspected UX problem?
A few days, not months. The method is recruiting a handful of real users, giving them the actual task the data suggests they fail at, and watching without helping. About five users find most of the usability problems in a design, so the confirmation step is fast on purpose: the goal is to see the problem clearly enough that a team stops arguing about whether it exists, not to measure it precisely.
How do I tell a UX problem apart from a marketing or positioning problem?
Low signups paired with good activation is usually a marketing problem: the people who arrive succeed, there are just not enough of them, and no interface change fixes having too few people at the first step. A UX problem looks different, drop-off concentrated at one step inside the product, or repeated how-do-I tickets about a feature that already exists. Watching real users attempt the task is what tells the two apart.
What happens if a confirmed UX problem is left unfixed?
The costs keep showing up in places nobody labels as design: support tickets asking how to do things the product already does, users rebuilding their workflow in a spreadsheet without telling you, sales demos that only work with an expert operating the product. These costs grow because each new user reaches the same failure point, and the evidence stays scattered across inboxes instead of being collected into one clear finding.
Can analytics alone confirm a UX problem, without watching users?
Analytics can show where a problem likely is, a drop-off step or a feature nobody touches, but it cannot explain why. Two products with the same large drop at one step can have opposite causes: a confusing form in one, an unwanted question in the other. Watching a handful of real users attempt the task is what turns an analytics pattern into an actual diagnosis, because you see the moment of failure rather than inferring it from a chart.
Related reading
The early warning signs your software project is in trouble
A software project rarely fails in one dramatic moment. It falls behind slowly, and the early signs are easy to explain away. Here are the ones to watch and what to do about each before it is too late.
Knowing what to build is now the most important decision
AI made writing software cheap. That moved the hard part earlier, to deciding what deserves to be built and having the discipline to cut the rest.