Customer feedback buttons: what a smiley press does and does not tell you

You have seen them. Four faces on a stand by the exit of an airport security lane, or screwed to the wall outside a supermarket restroom. Dark green delighted, light green fine, orange not great, red angry. You press one on the way past without breaking stride.

They work. That is worth saying first, because the rest of this piece is about their limits and it would be easy to read as dismissal. A row of buttons by a door collects feedback at a rate almost nothing else matches, from people who would never fill in a survey, and it does it in about one second with no phone, no app, and no typing. As a way of turning footfall into a number, it is very good.

The question is what that number is for, and this is where a lot of businesses buy the wrong tool. So: what a button press actually contains, what it structurally cannot contain, and how to tell which of those you need.

What you genuinely get

You get volume, and volume buys you three real things.

A baseline. After a few weeks you know what normal looks like, which means you can recognize abnormal. Without a baseline every complaint is an anecdote and every good day is luck.

Time resolution. This is the underrated one. Because presses are timestamped and plentiful, you can see the shape of a day and a week. Thursday evenings are worse than Thursday mornings. Satisfaction falls off after four in the afternoon. The first hour after opening is your best hour. That pattern is genuinely hard to see from the inside and genuinely actionable when you can.

Place resolution, if you put out more than one unit. The door by the car park scores worse than the door by the high street. The restroom on the second floor scores worse than the one on the ground. Comparing two locations against each other is often more useful than either number alone.

None of that requires anyone to write anything, which is exactly why it works on people in a hurry.

What a press cannot contain

Here is the structural limit, and no amount of analytics fixes it: a button press records a verdict and discards the reason.

Say your Thursday evening dip is real and consistent. You now know something true and you still do not know what to do, because the dip has at least a dozen plausible causes and the data cannot distinguish between them. A queue forms when the second till closes at six. One member of staff starts a shift at five. The delivery blocks the entrance. It gets dark and the car park lighting is out. The bathroom has not been checked since lunch. The special sells out.

Every one of those produces the same orange press. You have measured the problem precisely and learned nothing about its cause, which means your next step is to stand in the doorway on a Thursday evening and guess. Sometimes that works. It is a strange thing to have paid a subscription for.

The touchscreen versions try to close this gap by asking a follow-up: you pressed orange, was it the staff, the wait, the cleanliness, the product. That helps, and it helps less than it sounds, because you wrote the list. You will find out which of your six hypotheses is most popular. You will not find out about the seventh thing that never occurred to you, which is usually the one worth knowing.

The thing nobody tells you about the numbers

Two quieter problems worth going in with your eyes open about.

Presses are not a sample of your customers, they are a sample of people willing to press a button, and that group skews to the extremes. Someone actively delighted or actively annoyed will detour to the stand. The large middle who thought it was fine walks past. So your score is not the average experience, it is the average of the people who felt something, which is a different and more volatile quantity. It also means small absolute changes can look dramatic.

And children press them. Anyone who has watched a kiosk in a supermarket entrance has seen a small child work along all four faces in order, twice. It is not fatal, and it does add noise you cannot filter, because the press carries no identity.

The wider pattern behind all of this is the one I keep coming back to: people do not tell businesses things, and the reasons are structural rather than personal. I wrote about that at length in why customers don't complain and just leave instead.

Measurement or diagnosis

The clean way to decide is to work out which problem you actually have, because these tools sit on opposite sides of it.

If your problem is measurement, buttons are the right buy. You suspect quality varies and cannot prove it. You have several sites or shifts and want them comparable. Somebody above you wants a trend line. You need to know whether last month was better than the month before. All of that is what continuous scoring is for.

If your problem is diagnosis, buttons will frustrate you. You know something is wrong, or you know customers are drifting away, and what you need is a sentence explaining why. Ten written notes will move you further than ten thousand presses, because each one names something. The counterpart article, feedback kiosks versus a printed QR sign, goes through that comparison properly.

My honest guess is that a single location usually has a diagnosis problem while believing it has a measurement problem. If you are one shop, you can already feel roughly how things are going. What you cannot see is why the regular who came every week for two years stopped coming, and no score will tell you.

What the hardware costs, roughly

Worth knowing before you get attached to the idea. This is a subscription category, not a purchase: the hardware is usually rented or bundled, and the recurring fee is the real cost.

Pricing transparency varies a lot. Some vendors publish a starting figure and some will not give you one without a call, and the ones that will not are generally the ones built for multi-site buyers. I went through what four of them actually publish, checked in August 2026, in what a HappyOrNot terminal actually costs, including why the figure everyone repeats online traces back to an unverifiable citation from 2018.

The part that catches people out is not the subscription, it is the physical reality. A unit needs somewhere to stand that customers pass but nobody trips over, and depending on the model it needs power. Somebody has to notice when it is unplugged, offline, or has been pushed behind a display. That job lands on whoever is nearest, which in a small business is you.

A cheaper way to find out which you need

If you are not sure whether you have a measurement problem or a diagnosis problem, there is a test that costs the price of printing and settles it.

Put up a sign with a QR code that goes to a two-question form, in the place you were going to put the kiosk. Leave it a month. Read what comes in. You will get far fewer responses than a button would collect, and if even a handful of them tell you something you did not know, you have your answer: your problem was diagnosis, and the notes are worth more to you than the score would have been. If instead you get very little and mostly want to know whether things are getting better or worse over time, buy the buttons.

You can build that test yourself with a Google Form and any free QR generator. It is also roughly what I sell, with the annoying parts handled: Knoted mails you the printed sign, emails you the moment a note arrives, sends a weekly digest, and puts a different code on each sign so you know which one it came from. Flat rate, 30-day money-back guarantee from the day the sign arrives, no questions asked, current rate on the pricing section.

The two are not really rivals. One counts, the other explains. If you have the budget and the footfall, running both is genuinely the best answer, and if you only get one, pick the one that matches the problem you actually have this month.

Common questions

Do smiley feedback buttons actually work?

Yes, at the thing they are for. They collect far more responses than any survey, from people who would never complete one, because pressing a button takes a second and needs no phone. What they give you is a score with good time and place resolution. What they do not give you is the reason behind the score, which is a structural limit rather than a flaw you can configure away.

Why can't I just add a follow-up question?

You can, and the touchscreen models do exactly that. It helps, but the options are ones you wrote, so you will learn which of your existing hypotheses is most popular rather than discovering the cause you had not thought of. That seventh unlisted reason is usually the valuable one, and only free text surfaces it.

Is the score representative of my customers?

Not exactly. It samples people willing to press a button, which skews toward those who felt something either way, while the large middle who thought it was fine walks past. Treat it as a trend to watch rather than an average experience, and be careful reading much into small movements. Children pressing all four faces in a row adds noise you cannot filter, since presses carry no identity.

Buttons or a QR code sign?

Depends whether your problem is measurement or diagnosis. If you need a trend line, comparable sites, or proof that quality varies, buttons. If you need to know why customers are drifting away, a written note beats thousands of presses because it names something. Single-location businesses usually think they need measurement and actually need diagnosis.