UX feedback: what user research costs, and what your visitors will tell you for free

A page always looks obvious to the person who built it. Psychologists call this the false-consensus effect: we assume other people think and behave the way we do. Nielsen Norman Group turned it into the line every UX designer learns early, you are not the user. UX feedback, or user experience feedback if you prefer it spelled out, is how you replace that assumption with what real people do and say.

There are two ways to get it. You can watch people use your site, which is user research. Or you can make it easy for them to tell you, which is the job of a feedback widget. I'm a UX designer. I have built digital products for years, I hold Nielsen Norman Group's UX certification, and I took the Interaction Design Foundation's courses on user research and usability testing. I also make a feedback widget, Two Cents, so weigh what I say accordingly. Every figure below links to its source, so you can check it.

What user research costs

Usability testing got cheap, in theory, a long time ago. In 2000 Jakob Nielsen argued that five people find about 85% of a design's usability problems, and that three small rounds teach you more than one big study. Jeff Sauro of MeasuringU went through the maths and agrees, with a caveat. The 85% holds for problems that affect about a third of users. A problem that hits one visitor in ten needs eighteen people before you can count on seeing it.

Five people still cost something.

Time. Nielsen's estimate is 39 hours for a first test of a website with five users: planning, writing tasks, recruiting, running the sessions, analysing them and reporting. A practised team gets that down to about two working days. The number dates from 1998, it comes from 50 student teams, and Nielsen calls it an upper estimate. The steps haven't changed since.

Money. On User Interviews, a recruiting platform, pay-as-you-go recruiting starts at $49 a session, and the participant's incentive is not included. The same company's incentives report, drawn from nearly 20,000 research projects, recommends about $80 an hour for a one-to-one interview with a consumer. By my sum, five one-hour sessions come to roughly $650 before anyone on your side has spent an hour on them.

Or someone else's time. Nielsen Norman Group's own rapid user testing, for five to ten users, starts at $40,000.

I learned how research should be done while studying for my UX certifications: in a lab, or better still, where people really use the thing. Then I went back to work. In all my years of building products I have never worked anywhere that could pay for it, and I don't think I'm unusual. If someone offered me the chance tomorrow, I'd take it gladly.

So I'm not arguing against research. I'm saying it costs enough that most small teams do it rarely or never, when it should be a habit.

Why once in a while isn't enough

The practitioners who study product teams keep landing on frequency. Teresa Torres, who wrote Continuous Discovery Habits, asks teams to talk to customers at least weekly. Jared Spool found that the teams with the best designs were the ones where everybody, product managers and executives included, spent two hours every six weeks watching real users. He calls those exposure hours.

Most small sites are nowhere near either number. The site changes every week all the same, and between two studies nobody is watching.

Why your visitors stay silent

The people on your site today already know what's unclear, what's broken and what's missing. Most won't say so.

Nielsen described the pattern in 2006 as participation inequality: in an online community about 90% of people read and never contribute, 9% contribute now and then, and 1% write most of what you see. A study of more than 280 million reviews on 25 platforms, published in the Journal of Marketing Research in 2020, found the same thing from another side. People with extreme opinions are more likely to write, so the delighted and the furious are over-represented.

Silence is the default, so every bit of effort you add to giving feedback loses you people who had something to say. And what does arrive is not a fair sample of your visitors, which matters for how you read it.

Nielsen's advice for hearing from more than the 1% was short: make it easier to contribute.

What makes a feedback form usable

Easier how? Four rules come up again and again, and each has research behind it.

Don't interrupt. In Anna Kaley's list of ten popup problems for NN/g, two are about feedback: asking before people have done anything meaningful, and interrupting them in the middle of a critical task. Her alternative is a static, unobtrusive way to give feedback whenever the visitor wants to. Her later guidelines on feedback requests say the same and suggest a small feedback tab at the edge of the page. Google's Search Central documentation takes the same line: intrusive dialogs frustrate people and cost the site their trust.

Ask for little. Kathryn Whitenton's form recommendations start with "keep it short". Every field you remove lowers the effort, and she would allow one or two optional fields at most. Kaley puts a feedback survey at about a minute. The Baymard Institute, which studies checkouts, reports that 17% of US shoppers have abandoned an order because the checkout was too long or complicated, and that the average checkout shows 23.48 form elements where 12 to 14 would do. A shopper at checkout wants what's in the basket. A visitor giving you feedback wants nothing from you, so they'll put up with even less.

Don't ask for what you can capture. NN/g's EAS framework for forms goes eliminate, automate, simplify. The second step says not to ask people for what the system already knows or can work out. A browser knows which page it's on, how big the screen is and which language its owner reads. Nobody should have to type that.

Make it readable. WCAG asks for a contrast of 4.5 to 1 between text and its background at level AA. In WebAIM's 2026 survey of the top million home pages, 83.9% had low-contrast text. People don't fill in a form they can't read.

How I built Two Cents on those rules

Two Cents started with my own site. While building cyclingkits.cc, a directory of cycling clothing brands, I wanted a quick inbox for what visitors had to tell me. Email asked too much of them. They had to open a mail client, write something and send it under their own name, and most people don't bother. A small box they could click, type into and send looked like far less work. Every idea I had for the widget after that went one of two ways: make it easy for the visitor to leave their two cents, or make it easy for me to understand the message in its context.

Each of these is in the product today.

• Nothing pops up uninvited. There is a button in the corner, or a link you put on the page it concerns, and the visitor decides to open it.
• One sentence is enough. The email field is optional, and you can switch it off.
• The context comes with the message. The page, the browser, the screen size, the language and how long the page took to load arrive with every message. A bug report adds the console errors and the requests that failed, and a screenshot if the visitor attaches one.
• What a visitor typed survives closing the panel, so a stray tap doesn't cost them their message.
• The text meets WCAG AA contrast in the light and dark themes, whatever accent colour you choose.
• It speaks the visitor's language, picked from their browser: English, French, Dutch, German or Spanish.
• It sets no cookies.

My own evidence is small and I'll give it as it is. On cyclingkits.cc, 35 messages arrived between 13 and 21 September 2026. One of them pointed out that the country list wasn't sorted. No analytics tool would have shown me that. Messages still come in, and some visitors leave their email address, so I can write back.

If you're comparing widgets, I listed eight in Best website feedback tools in 2026, and put mine side by side with two of them: Feedback Fish and Ybug.

What a widget can't tell you

A feedback widget doesn't replace research, mine included. Nielsen's first rule of usability is to watch what people do and to be wary of what they say they do, or would do. Christian Rohrer's map of research methods files customer feedback as coming from a self-selected sample.

Both are right. A message through a widget is what somebody said, and that somebody chose to speak. It tells you where something is wrong, in their words, on the page where it happened. It can't tell you how many people hit the same thing in silence, or what they would do with your redesign. For that you still need to watch five people.

The widget is the closest I have come to doing user research myself, and I know it isn't the same thing. If you can afford both, use both. A widget is there every day, and its messages show you which page to test next. The test then shows you what to do about it.

Two Cents is free for one site and 25 messages a month, with no card.