Key Takeaways
- Question order moves the numbers more than question wording does. Pew Research Center has documented contrast effects where a narrow question asked first pulls down the general rating that follows it. Put your one overall rating at the top, not after the detail questions.
- The number of points on your rating scale barely matters above three. A review of twelve studies found no reliable advantage for any particular count. Pick one and never change it, because comparability across rounds is what you lose.
- A closed list collects what you already suspected. In one Pew experiment, 58% picked an answer when it appeared on the list while only 35% volunteered it unprompted. You need one open field to catch the rest.
- A form with no named owner is a data-collection hobby. Decide before launch who reads each submission, inside what window, and what the reply says.
Most feedback form advice is a question library: pick twenty questions, arrange them under headings, send. That produces a form whose numbers nobody can compare to last quarter, because the scale changed, the order changed, and the one field that would have explained the score was cut for length. This guide treats the form as a measurement instrument.
What goes on a feedback form?
A working feedback form has six parts, and only one is mandatory. The pattern holds whether you are asking a client about a finished project, a patient about a visit, or an attendee about a workshop.
| Field | Type | Required | What it decides |
|---|---|---|---|
Overall rating | Single rating scale | Yes | The one number you trend over time |
FieldOverall rating TypeSingle rating scale RequiredYes What it decidesThe one number you trend over time | |||
Reason for the rating | Open text | No | Why the number moved |
FieldReason for the rating TypeOpen text RequiredNo What it decidesWhy the number moved | |||
Specific aspect ratings | 2 to 4 rating items on the same scale | No | Which part of delivery to fix |
FieldSpecific aspect ratings Type2 to 4 rating items on the same scale RequiredNo What it decidesWhich part of delivery to fix | |||
One forced-choice question | Single select, 4 to 5 options | No | Routing and segmentation |
FieldOne forced-choice question TypeSingle select, 4 to 5 options RequiredNo What it decidesRouting and segmentation | |||
Permission to follow up | Checkbox plus optional name field | No | Whether you can act on one account |
FieldPermission to follow up TypeCheckbox plus optional name field RequiredNo What it decidesWhether you can act on one account | |||
Permission to quote or publish | Checkbox | No | Whether the answer can become a public review |
FieldPermission to quote or publish TypeCheckbox RequiredNo What it decidesWhether the answer can become a public review | |||
Keep identity fields off the form where you can. If it arrived by a personal link you already know who sent it, and Nielsen Norman Group's guidance on rating scales notes that asking for identifiers raises social-desirability bias, the polite-answer problem that makes feedback useless.
Rule of thumb: one required question, everything else optional. Someone who abandons at a mandatory field you did not need gives you nothing, where the same person would have left you the overall score.
Why does question order change the score?
Because respondents treat a form as a conversation and avoid repeating themselves. Pew Research Center's notes on writing survey questions describe this as a contrast effect: in a 2003 study, 45% favoured legal agreements for same-sex couples when the question followed one about marriage, against 37% when it did not. The same mechanism applies to a feedback form. Ask about responsiveness, deadlines and billing first, and the overall rating that follows reads as "everything else", which is a different question from the one you meant to ask.
So: overall rating, then the open text, then the detail items, then anything administrative. That is the opposite of the warm-up advice you will see elsewhere, which puts easy specifics first to build momentum. Momentum is worth less than a number you can compare to last quarter's.
Two cheaper ordering details. Randomise the options inside any non-ordinal list, since Pew's methodology notes treat primacy and recency effects as real. Keep ordinal scales in their natural order, and never flip them halfway down to catch inattentive readers: Sauro and Lewis showed in 2011 that alternating positive and negative statements confuses respondents and researchers both.
Which rating scale should a feedback form use?
Any scale above three points, used consistently. That is less satisfying than the usual five-point Likert prescription and better supported. MeasuringU's review of the evidence on scale points covers twelve studies and lands on a blunt summary from Jim Lewis: points do not matter unless you use only three. Nunnally argued the same in 1967, that the reliability gains run out around eleven points.
Three points is the one real failure mode. It cannot separate a client who is quietly content from one about to churn, and a study of 1,737 participants comparing a three-point item against the original eleven-point likelihood-to-recommend question found that the extreme answers, the tens and the zeroes, are what predict behaviour.
Labelling is similarly oversold. A review of seventeen published studies found little evidence that labelling every point beats labelling only the ends, and MeasuringU's own test with 415 respondents found no statistical difference on five- and seven-point scales. Label every point when the content is objective, as in daily, weekly, monthly. Label the two ends when you are asking for a judgement.
Keep a neutral midpoint. It draws between 10% and 20% of answers, roughly the share who genuinely have no view, and forcing them off the fence invents data. Add a separate "not applicable" option so the genuinely neutral and the genuinely unaffected stop sharing a bucket.
The decision that actually matters: whichever scale you choose, freeze it. Changing from five points to ten between rounds means your two quarters are not comparable, and rescaling after the fact does not fully repair it.
A client feedback form template you can copy
Here is the form in order, answer options written out. It takes under ninety seconds, and every question after the first is optional.
1. Overall, how would you rate working with us on this project?
Five-point scale, ends labelled: 1 = Poor, 5 = Excellent. Neutral midpoint present, unlabelled.
2. What is the main reason for that rating?
Open text, no character minimum. This is the field that earns the form its keep.
3. Please rate each of the following. (same five-point scale, with a Not applicable option on each row)
- Clarity of communication
- Meeting agreed deadlines
- Quality of the delivered work
- Value for the fee
4. Which part of the engagement was most useful to you?
Single select: the planning and scoping stage · the delivery work itself · the handover and documentation · the follow-up support · something else (open text)
5. What should we have done differently?
Open text. Phrased as a direct request, not as "do you have any suggestions", which invites a no.
6. Would you work with us again?
Forced choice: Yes · No · Not sure. Pew's guidance prefers forced-choice formats over agree-and-disagree statements because they limit acquiescence bias, the tendency to nod along.
7. May we contact you about this feedback?
Checkbox, plus an optional name and email field that only appears when it is ticked.
8. May we publish part of your answer as a review?
Checkbox, with the attribution options spelled out: full name and company · first name only · anonymous.
Note what is absent: no "how did you hear about us", no demographics, no fifteen-statement matrix, no recommendation score stapled next to the overall rating. Two scores measuring near-identical things is how a form gets long without getting more informative. For a wider question bank, our round-ups of open-ended question examples, multiple-choice question examples and frequently used survey questions cover the alternatives.
Skip the form entirely if you have fewer than about fifteen clients a quarter. A fifteen-minute call tells you more at that volume, and the form's one advantage, comparable answers at scale, has nothing to work with.
How many questions can a feedback form hold?
Fewer than you want, and the cost is front-loaded. SurveyMonkey's analysis of drop-off across 100,000 surveys found abandonment rises fastest with each question added up to fifteen, after which each further question up to thirty-five costs less than the first fifteen did. The expensive questions are the early ones.
Survicate's study of 21,000 surveys puts completion for one-to-five-question forms at 77.2%, with the steepest decline just above that. The template above has eight items, one a four-row rating block, which sits where most openers still finish.
Watch the matrix grids. A block of rating rows reads as one question in your form builder and as four questions to the person filling it in. Count rows, not questions, when you estimate length.
What happens after someone submits the feedback form?
Decide three things before the form goes live: where the answer lands, who owns it, and what the reply says. Skip any and the form becomes a collection exercise that trains clients that answering changes nothing.
Route on the score. A low overall rating should notify a named person immediately rather than wait for a weekly digest, since the window to rescue an unhappy client runs in days. A high rating with the publish box ticked goes to whoever manages your review profiles. Everything else can batch.
Keep the reply window short. Every submission that came with contact permission gets a human reply inside two working days, even when that reply is only an acknowledgement and a note of what changed. It is what holds up the next round's response rate.
Then settle the legal basis. Feedback about a named engagement is personal data, and Article 6 of the GDPR requires a lawful basis before you process it. For existing clients, legitimate interests is the usual route, with a documented balancing test and a retention period you keep to. Publishing an answer as a public review is a separate decision, which is what the form's final checkbox collects. Our guides to feedback programme design and running surveys and unsolicited feedback together cover the process side.
What should you build the feedback form in?
Three capabilities, and they can come from three vendors: a builder with a rating field type, a sender that triggers off a finished job, and somewhere for answers to sit next to the client record.
meetergo covers all three in one subscription. Its form and funnel builder carries 25+ field types including star ratings, yes/no, text, file upload and an electronic signature field, with multi-page forms and conditional logic branching on EQUALS, CONTAINS, BLANK and GREATER THAN under nested AND/OR rules. Forms go out as an embed, a link, or a direct send by email or SMS to a named recipient, with opens and completions visible in the dashboard. Every submission fires a webhook to your CRM, automated workflows handle the send and the chase, and the client reviews and ratings app collects a review after a booking and publishes it once you approve. Responses stay on EU-hosted infrastructure, and the company behind it has no US parent, so no US jurisdiction reaches the data. The Free plan is free forever, Light is €9.90 a month, Suite €29.90 and Premium €49.90, with a 7-day trial on the paid tiers.
Two honest limits. meetergo's survey tool is built around option-based polls with live charts and Excel or PDF export, not rating-scale instruments, so a feedback form belongs in Forms rather than there. And while appointment data has a CSV export and submissions push out by webhook, a one-click export of form responses is not documented, so check it in your own account before you promise anyone a reporting workflow.
Dedicated survey platforms earn their fee once you need panel sampling, weighting or significance testing. Where there are only two or three questions and a booking already exists, questions attached to that booking remove the need for a separate form. For a field-by-field build of a general enquiry form, our guide to building a form online and the overview of online form creators go further.
What to check before you pick a form tool
- Is there a real rating field, or just a radio group? A radio group works, but you rebuild the same five options on every row and lose the matrix layout. Check that the scale is a field type.
- Can the send trigger off an event? A tool that only produces a shareable link puts timing back on a human, and timing is the variable with the largest effect on response rate.
- Where do responses live, and can you get them out? Ask for the storage location, the retention default and the export format, in writing. The anonymity checklist covers what metadata a platform keeps whether you asked for it or not, and our data protection overview sets out the hosting side.
- Does one scale hold across every form? If each owner picks their own, you have several instruments and no trend line. Template it centrally, the way a standardised form works.
Common mistakes
- Running two overall scores. A satisfaction rating and a recommendation question next to each other measure almost the same thing and cost twice the attention. Pick one as the number you trend, and let the other go.
- Making the comment field mandatory. It raises abandonment right after you have already collected the score, and the forced comments you do get are padding.
- Rewriting the questions each round. Pew fields identical wording when it wants to compare across waves. Fix questions 1 and 3, rotate only the optional extras.
- Reading averages without the response rate. A 4.7 average off a 6% response rate measures who likes you enough to answer. Report both numbers on the same line.
- Treating a low score as a data point. It is a notification. Route it to a person the day it arrives.
Build the form once, freeze the scale, let the trigger do the sending. One full round will tell you more about your delivery than a year of ad-hoc praise did.
Build your first feedback form free. No credit card required.
Booking + video conferencing in one tool.
Booking + video conferencing in one tool.
FAQs
What questions should a feedback form include?
One overall rating, one open field asking the reason for it, two to four specific aspect ratings on that same scale, and a permission checkbox. Everything beyond that is a trade against completion, so add a question only when you can name the decision it feeds.
Should a feedback form be anonymous?
Anonymous answers are more candid, and anonymity costs you the ability to fix a specific account. The practical middle is an anonymous submission with an optional "happy to discuss this" checkbox that reveals a name field when ticked, which a useful share of respondents tick once it is clearly voluntary.
Is a five-point or a ten-point scale better?
Neither is reliably better, as long as you stay above three points. The evidence shows diminishing returns past about eleven and no clear winner in between, so the real decision is picking one and keeping it unchanged across rounds.
How long should a feedback form be?
Short enough that a four-row rating block still counts as four questions in your estimate. Completion is highest in the one-to-five-question range and falls fastest up to about fifteen, so the eight-item template above is near the practical ceiling.
Do you need consent to send a client a feedback form?
You need a lawful basis, which is not the same as consent. For existing clients, legitimate interests usually applies, with a balancing test on file. Publishing an answer publicly is a separate purpose and does need explicit agreement.
