Most customer feedback surveys fail in the same two ways. They are too long, so nobody finishes them. And they ask everyone the same questions, so the answers say nothing useful about anyone in particular.
Here is how to build one that avoids both.
Start with the decision, not the questions
Before you write a single question, finish this sentence: “When the results come in, we will decide whether to ___.”
If you cannot finish it, you are not running a feedback survey. You are collecting data because collecting data feels productive. A survey with no decision attached produces a spreadsheet nobody opens.
Once you have the decision, the questions almost write themselves. You need the smallest set of answers that would change it.
Ask one number first
Open with a single rating question. It is the one thing every respondent will answer, so it should cost nothing:
- Satisfaction (CSAT): “How would you rate your experience?” on a 1–5 scale.
- Recommendation (NPS): “How likely are you to recommend us to a colleague?” on a 0–10 scale.
Either works. What matters is that you pick one and never change the wording, because the value of the number is entirely in comparing it to the last one. Rewrite the question and you have restarted the measurement from zero.
Use a linear scale here rather than a set of radio buttons. It is a single tap, and the answer is a number you can average.
Then branch on the answer
This is the part that separates a useful survey from a generic one. A customer who rates you 5 and a customer who rates you 2 need completely different second questions:
- The 5 should be asked what specifically worked, and whether you can quote them.
- The 2 should be asked what went wrong, and whether they want someone to call.
Asking both groups “any other comments?” gets you silence from the first and a wasted opportunity with the second.
In Questionful, conditional logic runs per answer choice, so each option can route to its own section. One caveat worth knowing up front: a linear scale cannot route on its own. The pattern that works is to put a short choice question next to the scale (three bands, “Great / Fine / Not good”) and branch on that.
Keep it to five questions
Not five sections. Five questions, including the rating.
Response rate falls sharply after about ninety seconds, and every question you add pushes more people over that line. If you have twelve things you want to know, you do not have a feedback form. You have a research survey, and it needs a different audience, an incentive, and a longer intro explaining why.
A shape that works:
- The rating (scale)
- The band (choice. This is your branching question)
- One open question, different per band
- One closed question you can aggregate
- “Would you like someone to get back to you?” (choice)
That is it. Our customer feedback form template is exactly this shape if you want a starting point.
Send it while the memory is fresh
Timing beats wording. The same survey sent an hour after a support call and a week after it will produce measurably different response rates and noticeably different specificity.
Send it:
- Immediately after a support conversation closes
- Within a day of a delivery or an onboarding session
- On the same day as an event, not the following week
The only exception is a relationship survey, a quarterly customer satisfaction survey, where you deliberately want the average impression rather than the last interaction.
Ask for identity last, if at all
The email question is where people abandon. If you genuinely need to follow up, put it at the end and mark it optional. If you do not, leave it out entirely: anonymous surveys get higher response rates, and Questionful never requires respondents to sign in, so an anonymous form is genuinely anonymous.
A good middle ground: end with “Would you like someone to get back to you?” and route the yes to a short section that asks for contact details. Everyone else finishes without giving one.
Make the low scores reach a human today
A bad rating has a shelf life. Discovering it in a monthly report is discovering it too late.
Turn on email notifications so the submission reaches a person immediately, up to ten addresses per form. If you want it to reach a system rather than an inbox, point the form at a webhook and let Zapier, Make, or your own service raise a ticket for anything below your threshold.
Close the loop, visibly
The single biggest predictor of your next survey’s response rate is whether anything visibly happened after the last one.
Publish something. A short note, a changelog entry, a line in your newsletter: “You told us X, so we changed Y.” Questionful’s opt-in public response summary can show the aggregate for closed-set questions if you want to be more transparent than that. Open text and uploads are never published, so nobody gets quoted by accident.
Skip this step and you can watch your response rate halve, survey by survey, until you conclude that “customers don’t fill in surveys.” They do. They stop filling in surveys that go nowhere.
The short version
- Name the decision the survey is meant to inform.
- One rating question, worded identically every time.
- Branch the follow-up on a band, not on the scale itself.
- Five questions total.
- Send it within a day.
- Identity last, and optional.
- Alert on the bad ones the same day.
- Tell people what changed.
If that sounds like a form you would want, the customer feedback form template already has the questions and the branching in place, and the customer feedback use case walks through the whole workflow, from the first rating to the response landing in your stack.