Website user feedback: how to collect it on a site you run alone

A link that goes nowhere, a price someone couldn't find, a brand you forgot to list, a thank-you. That's the kind of thing visitors tell you about your site when you make it easy, and it comes from strangers on the live site, not from your team testing a new version.

Most of what you'll read about collecting it is written for product teams. Userback, which calls itself user feedback software for product teams, promises "one place for your product team" on its home page, and its plans are counted in seats (both read on 16 October 2026). This article is for people without one. Maybe you're a founder who also writes the code, or a freelancer looking after a few client sites, or two people in a small studio. I'm one of them. I run my own sites alone, I'm a UX designer, and I make a feedback widget, Two Cents, so weigh what I say with that in mind. The sources are linked.

Five ways to collect it

None of these replaces the others. They hear from different people, at different moments, and a site run by one person can afford all five.

A feedback button

A small button in a corner of every page, which opens a short form. It costs the visitor almost nothing, and it waits until they have something to say. But it mostly hears from people who were already annoyed enough to go looking for it.

Where to put one, what to write on it and how to add one are in Feedback button for your website. The widget I make is described on the feedback widget page.

A question on the page

A button waits for people to come to you. A question goes to them, on the page it's about: "Which brand are we missing?" under an empty search result, "Is anything unclear about our plans?" under a pricing table. It hears from visitors who had something to say but no reason to say it.

On cyclingkits.cc, my directory of cycling clothing brands, one question asks visitors to add a brand that's missing. By mid-October 2026 it had brought in about forty answers. A second question, asking visitors to report a mistake on a brand's page, has had none.

I sorted thirty questions by the page they belong on in Website feedback questions. How Two Cents opens one from a link or a button is on the custom questions page.

Email and the contact form

Keep them. A visitor who wants a conversation, a quote or a refund will look for your address, and should find it. They're a poor way to hear about a broken filter, though: writing an email from your own address, under your own name, is more than most people will do for a stranger's site. I went through what each option costs the visitor in Contact form or feedback widget.

Analytics, for the what

Analytics isn't feedback, but it belongs on this list because it tells you where to listen. In Nielsen Norman Group's comparison of qualitative and quantitative studies, Raluca Budiu sums up the difference in one row of a table: quantitative methods answer "How many and how much?", qualitative ones answer "Why?". Your analytics are the first kind. Say they show that half the visitors leave the pricing page without clicking anything. They can't show what those visitors were wondering.

So use them to choose where to ask. A page with an odd number on it is the page that gets a question, and the question asks about what the numbers can't see. Don't ask "Where did you get stuck?" when your funnel already says where.

Talking to people

Some visitors leave an email address with their message. That's an invitation to write back, and sometimes to ask one more question.

Ask about something that happened, not about what they might do. Jakob Nielsen's article on interviewing users warns that asking people to remember how they used something, or to guess how they will, gets answers that are "extraordinarily weak and often misleading". You "can't ask 'Would you use (potential future) feature X?'" because nobody can judge something they haven't seen. His advice is to ask about specific instances instead, such as the last time something went wrong. Rob Fitzpatrick's book The Mom Test takes on the same problem: how to talk to customers "when everyone is lying to you".

"Which page were you on when the filter stopped working?" gets you a fact. "Would you use a filter by country?" gets you a polite yes.

What to do with what arrives

Collecting is the easy half. Then someone has to read each message, decide what to do about it and maybe answer, and on a site you run alone that someone is you.

Read it at a fixed time

Messages one by one through the day pull you out of whatever you were doing. Once a day is enough for most small sites: a digest in the morning, ten minutes with a coffee. Bugs that stop people from buying are the exception, and those you want to hear about straight away.

Decide each message

Every message gets one of three outcomes: fixed, planned, or closed with a reason. A message left undecided gets read again next week, and the week after. A one-line note on each ("added on 14 October", "already in the menu, under Settings") saves you from deciding the same thing twice.

Group what repeats

After a few weeks, the same thing comes back in different words. Nielsen Norman Group's guide to affinity diagramming, by Rachel Krause and Kara Pernice, describes a way to sort it. You write each observation on a note, cluster the notes that belong together, and give each cluster a name. They prefer it as a team exercise, though you can do it on your own. With a few dozen messages and nobody else in the room, a short tag in your note does the same job.

Read a request as a symptom

A visitor who asks for a feature may be telling you they couldn't find it. In the same interviewing article, Nielsen recalls that Microsoft asked customers to suggest new features for Office 2007 before work on it began, and that "most of the requested 'new' commands already existed in Office 2003". The team concluded that its problem was discoverability. On a small site, "please add a search" sometimes means the search box is there and nobody sees it.

Answer the people who left an address

Say thank you, and say what changed. It costs you a minute, and it's the one moment someone learns their message was read. When a visitor on cyclingkits.cc leaves an address, I reply. Messages there pointed out things no analytics tool would have shown me, such as a country list that wasn't sorted. They led to filters and sorting, a layout reworked for large screens, and every relevant suggested brand added to the directory.

What it won't tell you

Feedback comes from the people who chose to send it. About forty answers told me which brands were missing from my directory. They didn't tell me how many visitors looked for a brand, found nothing and left without a word. So I read each message as a lead, never as a count. Why most visitors stay silent, and what you still learn by watching people use your site, is in my article on UX feedback.

On a small site it also arrives slowly, a message here and there, and some questions, like my "report a mistake", can stay empty for weeks. An empty question is information too: either nothing is wrong, or the question asks for more effort than visitors want to give.

And it costs your time. Replying, deciding and fixing take longer than installing anything. Collect what you can read.

Where Two Cents fits

Two Cents covers the first two ways on the list, and the reading. It puts a button and your own questions on any site with one script tag, and each message arrives with the page, the browser and the screen size. The inbox is built for the deciding part: done, planned or closed in one key, a private note on each message, a reply from your own mailbox for those who left an address, and an export in CSV or JSON. It's priced by the site.

It has no voting board and no public roadmap, and it isn't a tool for testing a staging site with your team. If that's what you need, I compared eight tools in Best website feedback tools in 2026.

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