How to increase survey response rates: what actually works

What the research says

The best source on response rates is a systematic review by Cochrane, last updated in 2023. The researchers gathered randomised experiments in which one thing about a questionnaire was changed and the response rates compared. For online questionnaires they found 88 such experiments, covering more than thirty different strategies.

A selection of what raised response to online questionnaires:

Three caveats. The review was written with epidemiological research in mind, and the experiments come from all kinds of settings, not specifically customer or employee research. For half of the strategies, results varied considerably between experiments. And what works with one audience needn't work with another. Read the numbers as a direction, not a promise.

Ten things you can do tomorrow

1. Make it shorter. A shorter questionnaire clearly raised response in the experiments. Cut any question whose answer you wouldn't know what to do with. How to choose is covered in our employee survey questions and customer satisfaction survey questions.

2. Write a short invitation. Of all the strategies, a brief invitation letter had one of the largest effects. Three sentences: why you're asking, how long it takes, what happens with the answers.

3. Be honest about how long it takes. Say how long it will take, and say it honestly. An optimistic "just one minute" that turns out to be five costs trust. In the review, a longer time estimate was even associated with slightly more responses, although that effect wasn't significant.

4. Make it personal. A personalised questionnaire did better than a generic one in the experiments. Have the invitation come from someone people know, too, rather than from "noreply". In an anonymous survey, explain how that personal invitation squares with anonymity.

5. Explain why it matters. A topic that matters to people, and an explanation of what it achieves, both raised response. "We want to cut waiting times on the phone and would like to hear where it goes wrong for you" is more concrete than "Your opinion is very important to us".

6. Offer the results. Promising people the results worked as an incentive. With employees it's simply good manners: people who take part want to know what came of it.

7. Watch the word "survey". In the experiments, the word "survey" in the subject line lowered response. A subject line that says what it's about ("How did your delivery go?") is clearer anyway.

8. Give a closing date. A deadline raised response a little. It also stops people saving it "for later".

9. Send a reminder. In the review, email reminders raised response to paper questionnaires. Send one, only to people who haven't replied, and add something new rather than repeating yourself.

10. Show what you did with last time. This isn't in the review, but anyone who has run an employee survey knows what happens after a survey that led nowhere. "You said X, we did Y" is the best invitation to the next round.

With employees, trust is everything

In employee research, the biggest barrier is rarely time; it's trust. People who think their answers can be traced either stay away or give the answers they think are expected. In the Cochrane review, assuring confidentiality raised response to paper questionnaires. But a promise only works if it's true. What anonymity really takes is covered in anonymous employee surveys and the GDPR.

Two other things help: giving people time to complete it during working hours, and managers who back the survey without checking who took part.

With customers, timing is everything

Customers owe you nothing. So ask while the experience is fresh: shortly after the delivery, the call or the purchase. Don't send a survey after every contact; someone who gets three in a week won't open the fourth. And think carefully about cash incentives: they raise response, but a large one also attracts people who are mainly there for the reward.

What is a good response rate?

There's no percentage above which your results automatically become reliable. A 30% response from a representative group can tell you more than 60% where mainly the satisfied replied. So always ask two questions:

• Who's missing? Compare the people who responded with the whole group, by department, customer type or channel, for example. If an entire group is missing, you don't know what it thinks.
• Is it enough for the decision? For NPS or eNPS you can work out the margin of error for your number of responses, for instance with the eNPS calculator or the NPS calculator.

Does a conversation beat a questionnaire?

A conversation with an AI interviewer feels different from a form, and it's tempting to assume people will be keener to take part. The research is less clear-cut than you might hope. A study by Xiao and colleagues (2020) with around 600 participants found that a chatbot that followed up on open answers produced markedly higher engagement and better answers: more informative, relevant, specific and clear. A study by Zarouali and colleagues (2024), on the other hand, found no evidence that a chatbot is a better tool than an online questionnaire. The difference seems to lie in what the conversation does: following up on what someone says, or just asking the same questions in a chat window. More in What is a conversational survey?