"Any Other Comments?" May Be the Most Ignored Question Online. Ask This Instead
In one Dutch panel, only 3.6% of surveys got an answer to the final 'any remarks?' box. Why the catch-all question fails, and 6 sharper questions to ask.
The short version
- The final comment box is mostly left empty: in two Dutch probability panels, only 3.6% and 5.7% of completed surveys carried a final comment (Schonlau, 2015, 2,550 comments categorised).
- Catch-all wording does worst: across 92 open questions on Pew's American Trends Panel, the highest skip rate, 52%, belonged to an 'are there any other measures' question. The most specific one was skipped by 3%.
- Telling people why an answer matters raised item response from 57.7% to 76.0% in a randomised experiment with 1,348 respondents (Kumar Chaudhary and Israel, 2016).
- Replace the catch-all with one narrow, story-shaped question, ask it next to the topic it belongs to, and say who reads the answers.
Fewer than 4 in 100 completed surveys came back with anything in the final box. That was the rate in the Dutch LISS panel, where respondents were asked at the end whether they had any remarks. If your survey ends with an "any other comments" question, this is the likely fate of that question, and we think the wording is the main reason.
96 empty boxes out of 100. The respondents are not the problem. The question is.
The catch-all closing question persists because it feels safe. It costs nothing to add, and it seems to promise that nothing important was missed. The research on open-ended survey questions says the opposite: vague prompts get skipped the most, and late prompts get the thinnest answers. Below is the evidence, then 6 replacement questions you can paste in today.
Instead of "Any other comments?", ask one narrow question tied to what the respondent just did, such as "What nearly stopped you from signing up?" or "What is the one thing we should change first?" Place it beside the related questions, not at the end, keep it optional, and tell people who reads the answers.
What actually happens to the "any other comments" box
Matthias Schonlau, a statistician at the University of Waterloo, did something most survey owners never do. He read the final comments. His 2015 study took random samples from two probability-based web panels in the Netherlands: the LISS panel and a Dutch immigrant panel. The closing question asked, in Dutch, whether respondents had any remarks about the questionnaire.
In the LISS panel, 3.6% of surveys contained a final comment. In the immigrant panel, the figure was 5.7%. Schonlau opens the paper by noting that such comments are usually ignored, partly because open answers are hard to analyse. So the box is mostly empty, and what does land in it is rarely read.
- Left the final box empty96.4%
- Wrote a final comment3.6%
The few who do answer are not a random slice
The comments that did arrive skewed critical. Only 1.8% of comments were positive in both panels, while 43.5% and 47.7% were negative. The rest were neutral. Among the negative ones, complaints about unclear and difficult questions were common.
That pattern matches what Pew Research Center sees on its American Trends Panel. Pew's 2023 analysis of open-ended nonresponse found that women, younger adults, Hispanic and Black adults and people with less formal education were less likely to answer open questions. The authors warn that open-ended results may disproportionately reflect the people most willing to respond.
Put those together and the final box has two problems. It collects very little, and what it collects comes from an unusual group. We would not report "what customers said in the comments" as if it described customers in general.
What the comments are good for
Schonlau also found something encouraging. Nearly all the comments were thoughtful, and many pointed at specific questionnaire problems. Comments about unclear questions were about 2.7 times more common in the immigrant panel, which suggests the questionnaire needed more pretesting for that audience. That is a real use. It just is not the use most people imagine when they add the box.
Why the catch-all question gets skipped
The "any other comments" box fails for three separate reasons. Each one points to its own fix, which is why rewording alone rarely solves it.
It asks people to invent the topic
A specific question hands the respondent a topic. A catch-all asks them to search their memory for something worth saying, decide whether it qualifies, then write it. That is the hardest kind of open question to answer.
Pew's 2021 analysis by Dorene Asare-Marfo shows how much burden matters. She looked at 92 open-ended questions across more than 40 panel surveys. Closed questions were skipped about 1% to 2% of the time. Open questions averaged about 18%, ranging from 3% to just over 50%. Questions coded as high cognitive burden had median skip rates about 3 times those of low-burden ones.
The worst performer is worth reading in full. It asked whether there were "any other measures" respondents thought would do a lot to reduce inequality between Black and White people. It was not a closing comment box, and the topic was demanding. Still, its shape is the catch-all shape: think of something we did not list. Half the panel did not.
It arrives when people are finished
The comment box nearly always sits on the last page. By then the respondent has already spent their effort. Mirta Galesic and Michael Bosnjak tested this directly in a web survey experiment published in Public Opinion Quarterly in 2009. They randomised the order of question blocks and found that answers to questions placed later were faster, shorter and more uniform.
So the final box gets the worst of both. It is the hardest question in the survey, asked at the moment of lowest effort.
It signals that nobody will read it
This one is our interpretation, not a measured finding. "Any other comments?" is a phrase people have seen at the bottom of countless forms. It reads as a formality. Respondents have no reason to think a person will act on what they write.
The evidence that framing matters is solid, though. In a randomised experiment, Anil Kumar Chaudhary and Glenn Israel added an importance statement to an open question about improving a public service. Item response rose from 57.7% without the statement to 76.0% with it, and answers grew from 20.6 to 27.2 words. The size of the answer box made no difference to whether people answered.
Six questions to ask instead
Each replacement below fixes at least one of the three problems. None of them is magic. The goal is to give respondents a topic, a reason and a moment.
| Instead of | Ask | Why it works |
|---|---|---|
| "Any other comments?" | "What nearly stopped you from signing up today?" | Names a specific moment the respondent remembers |
| "Anything else to add?" | "What is the one thing we should change first?" | Asks for one item, which lowers the burden |
| "Any feedback on this survey?" | "Which question here was hardest to answer, and why?" | Targets what final comments mostly contain anyway |
| "Any other thoughts?" (after a rating) | "You gave us a 6. What would have made it an 8?" | Builds on an answer they just gave |
| "Comments (optional)" | "Our product team reads every answer to this question. What should they hear?" | States who reads it, which raised response in experiments |
| "Is there anything else?" | "What did we not ask about that we should have?" | Keeps a catch-all, but frames it as a gap to fill |
Ask about a moment, not about everything
"What nearly stopped you from signing up today?" works because it points at an event. Memory retrieves events far more easily than general opinions. It also tends to produce a story with a cause, which is what you wanted from the open box in the first place. For more prompts in this style, see our guide to voice of customer surveys.
Ask for one thing
Pew's data showed one-word and phrase prompts were skipped less than multiple-sentence prompts. "The one thing we should change first" asks for a single item and implies a ranking, which forces a choice. You lose breadth and gain an answer you can count.
Follow the rating with its reason
A rating tells you where someone stands. The follow-up tells you why. Asking "what would have made it an 8?" turns a vague box into a specific gap the respondent already has in mind. We cover the wording for net promoter scores in the NPS follow-up question that gets real answers.
Be careful with stacking follow-ups everywhere. In a 2009 web experiment, Holland and Christian found that interactive probes improved answers mainly for respondents interested in the topic, and nonresponse to the probe was even higher than to the original question. Use one follow-up, on the rating that matters most.
Say who reads the answers, and mean it
The importance statement only works if it is true. "Our product team reads every answer" is a promise. If nobody reads them, do not write it. If somebody does, saying so is the cheapest response-rate gain in this post.
Jolene Smyth, Don Dillman and colleagues found the same direction in a 2009 series of web survey experiments with university students. Any introduction that made response quality salient improved answers, for early and late responders alike.
Move the question, not just the wording
Rewording fixes the topic problem and the reason problem. It does not fix timing. If the question still sits after 25 others, it still gets the tired version of your respondent.
Place each open question directly after the closed questions on the same subject. Ask "What would have made it an 8?" right after the rating, while the reason is fresh. In a one-question-at-a-time form, this also means each open prompt fills the screen alone, instead of sitting at the bottom of a long page.
If you only need the follow-up from some respondents, use branching. Show the "what went wrong" question to people who rated you low, and a different one to people who rated you high. Our post on conditional logic in forms shows how to set that up.
Shorter surveys help too. In the same Galesic and Bosnjak experiment, the longer the stated survey length, the fewer people started and finished. Every closed question you cut gives the open one a fresher reader. We look at where people quit in our breakdown of form drop-off by question type.
When "any other comments" is the right question
We are not arguing for deleting the box everywhere. There are cases where an open catch-all earns its place, and the research supports some of them.
| Situation | Keep the catch-all? | Reason |
|---|---|---|
| Long-running panel or repeat survey | Yes | Comments flag unclear questions you can fix before the next wave |
| Pilot or pretest of a new questionnaire | Yes | You want to find problems you did not anticipate |
| Complaint or incident report where anything may be relevant | Yes | The open field is the point of the form |
| One-off customer satisfaction survey | No, replace it | A specific final question collects more and clearer answers |
| Short event or product feedback form | No, replace it | People are leaving; ask the single thing you most need |
If you keep it, give it a job. "Was anything in this survey confusing?" is a catch-all with a purpose. Schonlau's data suggest that is what many commenters want to tell you anyway.
Let people say it instead of typing it
Part of the burden of any open question is mechanical. On a phone, typing a considered paragraph is slow, and most surveys are now answered on phones. We look at that problem in detail in why open-ended survey answers come back thin.
One option is to let respondents speak their answer. In a randomised experiment, spoken answers were about 2.5 times longer than typed ones, but a voice-only design also lost far more respondents. The numbers, and the design that keeps the gain, are in our analysis of voice survey answers.
Our own view: a closing question is a good place to offer voice, because it is the question where you most want a story. Offer it as a choice, never as the only way to answer. In Would You Rather, the voice answer field has a switch that lets respondents type instead. It is off by default, so turn it on for this kind of question.
The limits of this evidence
The headline 3.6% comes from one study of two Dutch panels. Panel members answer surveys every month, and the closing question asked about the questionnaire itself, not about a product or service. A customer survey with a sharper closing prompt may see a very different rate. The same paper notes an employee survey where 40% wrote comments, which shows how much context matters. We have not opened that employee study ourselves.
Pew's 52% figure belongs to a single question on a demanding topic. It shows the catch-all shape at its worst. It does not prove that every "anything else?" question will be skipped by half your respondents.
The importance statement experiment used one question, one public service and a web and mail mix in Florida. The direction matches the Smyth study, but the size of the gain will not transfer exactly.
Finally, the replacement questions in our table are recommendations built from these mechanisms. Nobody has run a head-to-head experiment of "any other comments?" against each of them. Test on your own audience before you assume the gain.
Before your next send
Open your current survey and scroll to the bottom. If the last question is "Any other comments?", you have two choices. Replace it with one of the 6 questions above. Or keep it, and give it a purpose in its wording.
Then move it. Put the open question next to the topic it belongs to. Add one honest sentence about who reads the answers. Send to your usual list and compare two numbers with last time: the share of people who answered that question, and how many of the answers you could act on. If both rose, the old box was costing you more than it looked.
References
- Schonlau, M., What do web survey panel respondents answer when asked "Do you have any other comment?", Survey Methods: Insights from the Field, 2015. Used for the final comment rate, comment types and the 40% employee survey figure it cites.
- Asare-Marfo, D., Why do some open-ended survey questions result in higher item nonresponse rates than others?, Pew Research Center, 2021. Used for open vs closed nonresponse, burden effects and the 3% to 52% range.
- Dunn, A. and Gómez, V., Nonresponse rates on open-ended survey questions vary by demographic group, other factors, Pew Research Center, 2023. Used for who skips open questions.
- Kumar Chaudhary, A. and Israel, G. D., Influence of importance statements and box size on response rate and response quality of open-ended questions in web/mail mixed-mode surveys, Journal of Rural Social Sciences, 2016. Used for the importance statement and box size effects.
- Galesic, M. and Bosnjak, M., Effects of questionnaire length on participation and indicators of response quality in a web survey, Public Opinion Quarterly, 2009. Used for question position and stated length effects (abstract only).
- Holland, J. L. and Christian, L. M., The influence of topic interest and interactive probing on responses to open-ended questions in web surveys, Social Science Computer Review, 2009. Used for probe effects (abstract only).
- Smyth, J. D., Dillman, D. A., Christian, L. M. and McBride, M., Open-ended questions in web surveys, Public Opinion Quarterly, 2009. Used for the effect of motivating introductions (abstract only).
The weakest point in this source base: no study we found tests the generic closing comment box against specific replacements head to head, so the replacement advice rests on mechanisms shown in other questions.
Questions people ask
Should I end my survey with any other comments?
Only if you have no sharper question to ask. In two Dutch survey panels, just 3.6% and 5.7% of completed questionnaires carried a final comment. A specific closing question, such as asking what nearly stopped someone from signing up, gives people a topic instead of making them invent one. If you keep a catch-all, make it optional, short, and say who reads the answers.
What is a good final open-ended question for a survey?
A good final question is narrow, concrete and tied to something the respondent just did. Examples: 'What is the one thing we should change first?' or 'Which question in this survey was hardest to answer, and why?' Pew Research Center found that open questions with a lower cognitive burden were skipped far less often than demanding ones. Specific beats general almost every time.
Why do people skip open-ended survey questions?
Open questions take effort. Respondents must decide what to say, find the words and type them. Pew Research Center found item nonresponse averaged about 18% on its open-ended questions, against roughly 1% to 2% for closed ones. Skip rates rise with cognitive burden and on mobile devices, and answers get shorter when the question comes late in a long survey.
Where should open-ended questions go in a survey?
Put each open question right after the closed questions on the same topic, not in a block at the end. A 2009 experiment by Galesic and Bosnjak found that answers to questions placed later in a web questionnaire were faster, shorter and more uniform. Asking while the topic is fresh, and before fatigue sets in, gives the question its best chance.
How do you get more responses to open-ended questions?
Tell respondents why the answer matters. In a randomised experiment with Florida extension clients, adding an importance statement raised item response from 57.7% to 76.0%, whatever the box size. Also narrow the question, place it early, keep it optional, and offer a voice option for people who would rather talk than type on a phone keyboard.
Are final comments in surveys worth analysing?
Yes, if you have them. A study of Dutch panel comments found that nearly all were thoughtful, and many flagged unclear or difficult questions, which makes them a cheap pretest for the next wave. The catch is volume and bias: few people comment, and those who do lean negative. Read them as signals to investigate, not as a measure of opinion.
Drafted with AI from published research, then edited and approved by Apoorv Tamrkar.
Would you rather hear the answer?
Make a voice form in a minute. It’s free.
Create a form