A feedback form builder, and how to get honest answers back
Most feedback forms produce a number nobody acts on. The fixes are unglamorous: pick one metric and stick to it, ask at the moment the experience is fresh, add exactly one open question, and route the bad scores to a human within the day. Here is how to design that, and what it costs to run.
Pick one metric, then stop changing it
Feedback programmes rarely fail because the wrong metric was chosen. They fail because the metric changed — a five-point scale became a ten-point one, the wording shifted, the timing moved — and eighteen months of trend data quietly became incomparable. Choose deliberately once, write down the exact wording, and treat it as fixed.
Whichever you pick, the number itself is close to worthless in isolation. A CSAT of 4.2 means nothing until you have last quarter's 4.0 to compare it to, and neither means anything until you read what people wrote underneath. The score tells you whether to look; the open text tells you where.
CSAT, NPS, CES — which one fits
Three standard measures, three different jobs. Using the wrong one produces a number that moves for reasons you can't act on.
CSAT — satisfaction with a specific thing
"How satisfied were you with..." on a 1–5 scale, asked immediately after a specific interaction: a support ticket, a delivery, an onboarding call. Best for measuring one moment, and the easiest to act on because the thing being rated is unambiguous. Report it as the percentage answering 4 or 5 rather than as a mean, which is less sensitive to the handful of respondents who use only the extremes.
NPS — likelihood to recommend
"How likely are you to recommend us?" on 0–10, treated as a relationship-level measure rather than a transactional one. Useful as a long-run trend and for benchmarking against an industry, notoriously blunt for diagnosis — a score can drop for reasons the question cannot reveal, which is why the follow-up "what drove that?" is doing most of the work. The dedicated NPS question type here is one of three that require a paid plan to publish.
CES — how much effort it took
"How easy was it to get your issue resolved?" This one is underused and often the most actionable of the three, because effort maps directly to a fixable process. If people report high effort, something specific was hard, and the open follow-up will usually name it. Particularly strong for support and self-service flows.
The one open question
Whatever score you use, follow it with a single open text field: "what's the main reason for your score?" This is where the actual information lives. Keep it optional, keep it to one, and make sure someone is genuinely reading the answers — an open question nobody reads is worse than no open question, because it implies a promise you aren't keeping.
Make feedback change something
A feedback form is only the collection step. These three make it a loop rather than an archive.
Ask while it's fresh
Response quality decays fast. Send within a day of the interaction you're asking about, not at the end of the month. A short survey delivered at the right moment outperforms a thorough one delivered late, both in response rate and in how specific the answers are.
Route the bad ones to a person
A low score is a service recovery opportunity with an expiry date. Wire the form to push responses where someone will see them — Google Sheets, Zoho CRM, Zapier, webhooks, or email notifications, all on the free plan, with Slack and HubSpot on Pro. Use tag rules to label responses at submission so the urgent ones are already sorted when they land.
Tell people what changed
The largest single driver of future response rates is evidence that last time mattered. A short note — you told us X, we changed Y — costs almost nothing and does more for participation than any incentive. Feedback programmes die when respondents conclude they are talking into a void, and they conclude that quickly.
Rules for writing feedback questions
Most bad feedback data is created at the drafting stage, not the analysis stage.
- One question about one thing — never quality and speed in the same sentence
- No adjectives that presuppose the answer ("how helpful was..." already assumes helpful)
- Keep the scale identical between waves, including the labels on the endpoints
- Label the endpoints explicitly so 1 and 5 mean the same thing to everyone
- Put the score first and the open question second, so the text explains the number
- Three to five questions total for a transactional survey — this is a hard ceiling, not a target
- Make everything except the core score optional
- Say who will read it, and whether it's anonymous — then honour that
- Never ask for feedback and a testimonial in the same form; they need different framings
Anonymity, incentives, and the two ways feedback gets skewed
Anonymity changes answers, especially internally. Employee feedback collected with a name attached is systematically more positive than the same questions asked anonymously, and people are good at working out which one they're filling in. If you promise anonymity, don't collect anything that would identify someone — a free-text field plus a small team is not anonymous no matter what the header says.
Incentives change who answers rather than what they say. Offering a prize draw raises response volume and pulls in people motivated by the prize, which is fine for volume and bad for representativeness. If you incentivise, do it consistently so at least the bias is stable across waves.
The bias you cannot design away is that people with strong opinions answer more. Your feedback form hears disproportionately from the delighted and the furious, and the large middle stays silent. That's tolerable if you remember it — treat the distribution as a signal of what can go right and wrong, not as a measure of how most customers feel.
Feedback templates
Structured to ask one score and one open question, which is the shape that works.
Star rating, NPS, and open feedback in under a minute.
A five-question satisfaction check after any customer interaction.
Overall rating, a matrix across event aspects, and next-year intent.
Post-appointment feedback on punctuality, the technician, and the fix.
Questions people ask
What's the difference between CSAT and NPS?
CSAT measures satisfaction with a specific interaction on a 1–5 scale and is asked right after it happens. NPS measures likelihood to recommend on 0–10 and is a relationship-level measure asked periodically. CSAT tells you whether a particular thing worked; NPS tells you how the overall relationship is trending. They answer different questions and are not substitutes.
Can I run NPS on the free plan?
The dedicated NPS question type requires a paid plan to publish — it's one of three, alongside e-signature and payment. Everything else needed for a feedback programme, including rating questions, conditional logic, tag rules, and integrations to Sheets, Zoho CRM, Zapier and webhooks, is on the free plan.
How many questions should a feedback form have?
Three to five for anything transactional. One score, one open follow-up, and at most a couple of context questions. Every additional question costs response rate, and the people who abandon are the busiest — so a long feedback form doesn't just get fewer answers, it gets less representative ones.
When should I send a feedback request?
Within a day of the experience you're asking about. Memory for specifics fades quickly, and a request that arrives weeks later gets a general impression rather than the detail you can act on. The exception is anything where the outcome takes time to become apparent, like onboarding — there, ask once early and once after the thing has had a chance to work.
Should feedback forms be anonymous?
For internal and employee feedback, usually yes — identified responses are measurably more positive than anonymous ones on the same questions. For customer feedback, an identified response is often more useful because it lets you follow up and fix the specific problem. Decide which matters more for the decision you're making, then be explicit with respondents about which one it is.
Keep reading
Better forms.
Better data.
Build your first form in under 60 seconds. Free forever for personal use, no credit card required.