Building a customer feedback form takes an afternoon. Deciding when to ask, getting people to the form, and answering the replies takes as long as the form stays in use. Guides tend to spend their length on the first part, which is the part that was never difficult, and skip the two that determine whether the exercise produces anything.
This covers the build, but the order is deliberate. Timing and delivery come first, because a well worded form that nobody sees collects nothing, and follow up comes last, because it is the reason the form exists and the place most feedback programmes stop.
Timing does more than wording
A customer feedback form sent on a schedule performs worse than the same form sent on a trigger. Schedules ask people about a period. Triggers ask them about an event, which they can actually remember.
The triggers worth using are specific. When a support conversation is marked resolved, within a day. When an order is delivered, within two days. When a piece of work is handed over, immediately. When a subscription reaches its first month, once. Each of these has a defined moment and a specific thing to ask about.
Two timings look sensible and are not. Asking at the moment of purchase collects enthusiasm rather than experience, because nothing has happened yet. Asking every quarter collects an average of everything since the last time, which produces answers too general to act on and a declining response rate as people learn the form is routine.
The window matters as much as the trigger. Asking about a support conversation four days later gets fewer and vaguer answers than asking the next morning, because the detail has gone. Asking within the hour, while the issue may still be unresolved in the customer's mind, produces answers about frustration rather than about the outcome. Somewhere between a day and two days is the practical range for anything transactional.
One send, one reminder, then stop. A second reminder rarely changes the total meaningfully and reliably annoys the people who already answered, which costs more than the extra responses are worth.
How people actually reach the form
There are four routes and they are not interchangeable.
A link in an email. The default, and it works because the trigger and the delivery are the same act. The thing that kills it is asking too much before the form opens. A first question rendered directly in the email, so that clicking an answer both records it and opens the form, performs better than a bare link, because the commitment is made before the page loads.
A link at the end of a completed interaction. Added to the closing message of a support conversation or the handover note on a piece of work. Response rates are high here because the person is already engaged, and the cost is that only people who reached the end of a conversation get asked.
A prompt on the site or in the product. Good for volume and poor for context, since whoever answers may not be the person whose experience you were asking about. Useful as a standing channel rather than as a measurement.
A code printed on something physical. Receipts, packaging, a card in the box, a sign at a counter. Response rates are low in absolute terms and the answers tend to be from people with something definite to say, which is not the same as representative but is often more useful.
For anything transactional, the first two do most of the work. The important property of both is that the form knows who is answering without asking, because the link was sent to a specific person. That single detail changes what the form can look like.
The build
Six fields, one required. This is for a transactional feedback form, sent after something happened.
- Did it get sorted? Yes, partly or no. First, because it is the easiest question in the form and because the answer determines whether anyone needs to act.
- How was the handling? A 1 to 5 rating with both ends labelled in plain words rather than numbers.
- How is the thing itself? The same 1 to 5 scale, asked separately. Handling and substance fail independently and get fixed by different people, and a form that merges them produces a number nobody can act on.
- What would have made it better? One open text box, labelled specifically. Generic invitations such as "any comments" get left blank several times more often than a specific question does.
- Would a reply be welcome? Yes or no. This separates the responses that need an answer from the ones being recorded, which is what makes the queue manageable.
- Email address. Prefilled and hidden if the link was personalised, shown and optional otherwise.
Everything else is optional or absent. No phone number, no company field, no job title, no matrix of adjectives to rate. Each additional field costs completions, and on a phone the cost is steeper than it looks.
Two build details are worth getting right on the first pass. Show one question per screen for anything over four questions, since most of these forms are opened on a phone and a long single page is where people stop. And do not make the open text field required, because a required paragraph turns a thirty second form into an abandoned one.
Knowing who answered
The single biggest difference between a feedback form that produces action and one that produces charts is whether each response is attached to a person.
When the link is personalised, the answer arrives already connected to the customer, the order, and the conversation it refers to. That means a two out of five can be opened, read in context, and replied to by name within the hour. When the form is anonymous or generic, the same two out of five is a data point and nothing more, because there is nobody to go back to.
There is a real trade off and it is worth stating plainly. Identified responses are slightly more polite than anonymous ones. Anonymous responses are marginally more candid and cannot be acted on individually. For transactional feedback the identified version wins clearly, because the value is in the recovery conversation rather than in the score. For broad attitude surveys about pay, management or culture, the reverse holds.
One mechanical point makes identified feedback compound. If responses are keyed on the email address, the same person answering three forms over a year stays one person with a history, rather than three unrelated rows. That history is what turns the fourth piece of feedback from an isolated complaint into a pattern somebody can see.
Before the first send
Five things have to be true, and skipping them is how a first send becomes an apology.
A reply address that a person reads. Feedback requests sent from an unmonitored address get replies anyway, by people who ignore the form and just hit reply. Those replies are often the most useful ones received.
Bounce handling. Some addresses will fail. A bounce is a fact about the record, and treating it as one keeps the contact list from quietly rotting.
A named person on duty. Before the send, not after. Somebody has to be responsible for the responses that arrive in the first forty eight hours, which is when most of them arrive.
A test to a real inbox. Not a preview inside the tool. Send the real thing to a real address, open it on a phone, and complete the form as a customer would.
A decision about what a bad response triggers. A one or two out of five should cause something specific to happen, defined in advance. Deciding in the moment means the third one gets a slower response than the first.
Following up on every reply
This is the part that separates a form that changes things from a form that produces a monthly figure. The work is mechanical and the difficulty is entirely structural.
Responses need an owner, because the most common failure is not a missed response but a visible one that two people each assumed the other had handled. They need a status, so that read, replied and closed are distinguishable at a glance rather than reconstructed from memory. And the reply needs to be sendable from the same place the answers are, because switching to an email client to answer is the friction that turns a two minute reply into something for later.
Three practical accelerators make the volume manageable. Saved wording for the three or four situations that recur, so that a reply is edited rather than written. Sending to several people at once with each person's own name and answers dropped in, for the acknowledgements that do not need individual drafting. And a record of what was sent and whether it was opened, which is what prevents the same person being answered twice or nobody at all.
Tools organised around managing responses after they arrive treat owner, status and history as properties of the response rather than as columns somebody maintains by hand. The situations where that pays off are the recurring ones, where the same form runs every month and more than one person reads it. For a single person handling a handful of responses a week, an inbox and a spreadsheet genuinely are enough, and the honest test is whether the current setup can answer how many responses from last month received a reply.
What it costs, by the shape of demand
Feedback volume is spiky. Quiet for a month, then busy after a launch, a delivery push or an incident. That shape is the thing to price against, and tools meter very differently.
| Tool | Entry point | What is metered |
|---|---|---|
| Google Forms | Free with a Google account | Nothing meaningful, but responses stay in a spreadsheet |
| Tally | Free, unlimited forms and submissions under fair usage | Paid Pro at 24 USD a month for branding and collaboration |
| Typeform | Basic at 39 USD a month | 100 responses a month and one user, 1,000 and three users on Plus at 79 USD |
| SurveyMonkey | Team plans, minimum three users | Charged per user, with responses beyond the allowance billed individually |
| Airtable | Free plan available | Editors at 20 USD per user per month on Team billed annually, with form submissions not charged |
Figures are each company's published pricing as of September 2026. The pattern worth noticing is that response meters and seat meters fail in opposite situations. A response meter breaks in the month the form matters most. A seat meter set at one breaks the moment a colleague needs to help, which for customer feedback is immediately, since replying is the job.
What to change first
Move the send from a schedule to a trigger, and personalise the link so responses arrive attached to a person, because those two changes do more for response quality than any rewording. Then check what share of last month's responses got a reply, and if the answer is unknown, the gap is between where feedback lands and where replies are sent. Halict shows what a single response looks like when both are the same screen.
Q1. When is the best time to send a customer feedback form?
On a trigger rather than a schedule, within a day or two of the event. After a support conversation is resolved, after an order is delivered, or immediately after work is handed over. Asking at the moment of purchase collects expectation rather than experience, and quarterly sends produce answers too general to act on.
Q2. How many questions should a customer feedback form have?
Five or six, with only the email address or the first question required. Rate handling and the product separately, since they fail independently and get fixed by different people, and label the open text question specifically rather than asking for any comments. Anything longer is answered less often and read less carefully.
Q3. Should a customer feedback form be anonymous?
For transactional feedback, no. An identified response can be opened in context and replied to by name, which is where the value is, and it costs only a small amount of candour. Anonymity is the right choice for broad attitude surveys about pay or management, where follow up matters less than honesty.
Q4. How do you get more people to answer a feedback form?
Ask sooner, ask about one specific event, and shorten the form. Rendering the first question inside the email, so that clicking an answer both records it and opens the form, beats a bare link. One reminder helps and a second one rarely does while costing goodwill with people who already replied.
Q5. What should happen to a low score once it arrives?
Decide in advance and write it down, so the third low score is handled as quickly as the first. Typically that means an owner is assigned the same day and a reply goes out within one working day, from whoever has the context. The response also needs a status, otherwise two people either both reply or neither does.