One-question surveys: why one question works, and where to put it

A one-question survey is a survey with a single question. The two best known are both scores. Fred Reichheld made one famous in a 2003 Harvard Business Review article, The One Number You Need to Grow. Bain & Company's Net Promoter site words it "How likely are you to recommend us to a friend or colleague?", on a scale from zero to ten. The other comes from usability testing. Jeff Sauro's Single Ease Question asks "Overall, how difficult or easy was the task to complete?" on seven points, right after someone has tried a task.

This article is about a third kind: one open question, asked on the page it concerns, that people answer in their own words. I'm a UX designer and I make a feedback widget, Two Cents, that is built around that kind of question, so weigh what I say accordingly. Every figure below links to its source.

Why one question

Every question you add loses you people. SurveyMonkey looked at 100,000 of its surveys, 2,000 at each length from one to fifty questions, deployed between 2009 and 2010. The more questions a survey had, the more respondents dropped out before the end, and the sharpest rise came with each question up to fifteen.

Nielsen Norman Group's guidelines on asking for feedback, written by Anna Kaley in 2023, put a feedback survey at about a minute. They add that surveys of two or three minutes, which the authors saw often, are too long. Kaley's other rule is about timing: "task, then ask". The request should come once the person has done something, and it should be about what they just did.

One question is the far end of that advice. It takes seconds, and it can only be about one thing, so it had better be about the page it's on.

Open or closed

A closed question gives the visitor choices: a scale, yes or no, a list. An open question gives them a field.

Closed questions give you a number you can chart, and they're easy to answer. Open ones give you what you didn't think to ask. Maria Rosala puts it this way in NN/g's article on open-ended questions: "The greatest benefit of open-ended questions is that they allow you to find more than you anticipate."

A score also needs its reason. Sauro's article recommends asking people why whenever they rate a task below five out of seven. It also reports that people who failed a task still rated it extremely easy about 14% of the time. A number on its own can point you the wrong way.

Open questions get skipped more, and some much more

Pew Research Center measured the cost. Dorene Asare-Marfo went through 92 open-ended questions from more than 40 of Pew's online surveys, fielded between 2018 and 2021. Closed questions there usually go unanswered by 1 to 2% of respondents. Open ones went unanswered by about 18% on average, from 3% for the easiest to 52% for the hardest.

What made the difference was effort. The questions that took the longest to answer were skipped about three times as often as the easiest ones. Questions that asked for a single word were skipped less than those that asked for a few sentences, and every kind was skipped more on a phone than on a computer.

Pew's respondents are a panel of people who agreed to take surveys, which visitors to your site are not. Still, the advice carries over: an open question works best when the answer is already in the visitor's head and fits in a few words. "Which brand are we missing?" asks for a name. "What do you think of our site?" asks for an essay.

Where to put it

On the page it is about, and nowhere else. "Is anything unclear?" means nothing in a popup on the home page. Under a pricing table, the same question gets you the sentence your plans forgot to say. On an empty search result, "Which one were you looking for?" gets you the thing the visitor just failed to find.

I sorted thirty such questions by the page they belong on in Website feedback questions, with the rules I use for wording them.

The question should also wait to be opened. A link or a line of text on the page, which the visitor who has something to say clicks, follows Kaley's advice better than a popup that interrupts everyone else. The other way to follow it is to ask when something has just happened: once a sign-up is finished, or when someone cancels. That is the moment Kaley's "task, then ask" describes, and it only interrupts the people who just did the thing you're asking about.

What one question brought in on my site

cyclingkits.cc is my directory of cycling clothing brands. It asks two questions. One asks visitors to add a brand that's missing. The other asks them to report a mistake on a brand's page.

By mid-October 2026 the first had brought in about forty answers, and a new one arrives most days. The second has had none. One site proves nothing, but the split fits what Pew found. Naming a brand you looked for takes a word. Reporting a mistake means reading a page closely enough to be sure it's wrong.

What it can't tell you

The answers come from the people who chose to answer. Forty answers told me which brands were missing. They didn't tell me how many visitors looked for a brand, found nothing and left without a word. So I read each answer as a lead, not as a count. Why visitors stay silent, and what you still have to learn by watching them, is in my article on UX feedback.

Asking one question with Two Cents

In Two Cents, a question you write yourself gets a key, and any link or button on your site opens it. It can also sit in the widget's menu on every page, or be the only thing the widget offers, and it can ask for stars as well as words. Your own code can also open a question when something happens on the page, such as a finished sign-up or a cancelled order:

twocents('open', 'where-stuck');

The answer arrives with the page it was sent from. Nothing opens by itself, on a timer or when someone moves to leave: a visitor opens the question, or your code does. How it works is on the custom questions page.

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