A survey asks customers something rather than telling them something. Because it costs them attention, the design questions that matter are when you ask and how much you ask for.
NPS — the nought-to-ten recommendation question. Comparable across time and against other companies, and largely meaningless without the follow-up asking why.
Satisfaction — how happy someone is with something specific.
Rating — a simple score.
Multiple choice — options you define, which is what you want when you intend to count the answers.
Open text — the answers that actually tell you something, and the ones you have to read.
A survey can appear as a small prompt in the corner or take over the screen. The small one gets ignored more and interrupts less; the full one gets answered more and costs more goodwill.
Use the compact form for anything ongoing, and reserve the full screen for a moment where stopping is reasonable — after something was completed, not in the middle of it.
Timing decides your response rate far more than wording does. Ask after someone has done the thing you're asking about, on the page where it happened, and only once they've been using the product long enough to have an opinion.
One question, plus an optional follow-up, is the format that gets answered. Every additional question loses people, and the ones who abandon halfway leave you a partial response that's worse than none.
If you have five things to learn, run five short surveys at different moments rather than one long one.
Responses come with a summary, and individual answers can be pulled out for your own analysis. The free text is where the value is — the score tells you where you stand, and only the words tell you what to do about it.
It's also worth deciding in advance who reads them and how often. A survey nobody reviews still costs your customers the interruption.
