Compliance · 5 min read
Feedback Automation: Do Not Incentivise and Do Not Gate
Automating a feedback request is technically trivial. An order closes, a template goes out, replies land in an inbox. A competent person builds it in a morning.
Everything that makes it worth building or not worth building happens in the three decisions around that: who gets asked, what they are offered for answering, and where their answer is allowed to go. Get those wrong and you have not built a feedback system. You have built a machine that manufactures agreement with yourself.
Never pay for the answer
A discount, a coupon, entry into a draw, points, free shipping on the next order — any of these attached to a feedback request changes what comes back.
It is not that people lie for a coupon, though some do. It is that the incentive changes who answers. The people who will fill in a form for a reward are a different population from the people who would have answered anyway, and the reward pulls the sample toward whoever is most motivated by the reward. The result looks like more data and is worse data.
There is a second problem and it is not a matter of taste. In many jurisdictions, a review or testimonial obtained in exchange for something has disclosure obligations attached to it, and rules on incentivised reviews differ by country and by platform and get enforced. What applies to your business is a question for someone qualified to answer it, not for a blog post. The safe version is to not create the question.
An incentive does not buy you honesty. It buys you the opinions of people who wanted the thing you offered.
Never gate the unhappy ones
This is the pattern worth naming precisely, because it is sold as a feature and it has a friendly-sounding pitch.
The mechanic goes: send a message asking "how was your experience?" with two buttons. Happy customers get a link to the public review page. Unhappy customers get routed to a private complaint form instead. The vendor describes this as "protecting your rating while capturing feedback internally".
What it is, is filtering the public record. The published rating stops being a measure of what customers experienced and becomes a measure of what you chose to let through. Review platforms treat this as manipulation and act on it. Consumer protection regimes in several countries have taken a similar view. And the customers you routed away notice, because it is not subtle — they clicked the sad face and the review link never appeared.
What honest asking looks like
- Pick a group by an event, not by a mood. Everyone whose order was delivered last week. Everyone who finished a service appointment. The selection rule should be something you could publish.
- Ask once. One message, no reminder chain. A second ask is pressure, and pressure produces the answer that ends the conversation.
- Ask one open question. "What could we have done better?" outperforms a rating scale for a small business, because a number tells you nothing you can act on and a sentence usually tells you exactly what went wrong.
- Time it to the experience, not to the invoice. After the thing happened, while it is still remembered. A week later you are asking about a memory.
- Reply to complaints with a person. Always. An automated "we're sorry to hear that, your feedback is important to us" on top of a real problem is worse than silence.
- Publish nothing without asking. A reply in a private thread is not a testimonial. Consent to complain is not consent to appear on a homepage.
The messages you will actually get
Anyone expecting neat structured responses should look at what a real inbox returns.
A single word: good. A voice note. A photograph of the item with no text at all — and nothing reads images on any channel, so there is no rule that can classify it and no model that can answer it. A long paragraph in Hinglish about a delivery agent who was rude. A complaint about something you do not sell. And, reliably, a handful of people who reply with an entirely new question because you just opened a thread with them.
That last one is worth planning for. A feedback message reopens a conversation, and some of what comes back will be a fresh support issue. If it lands somewhere nobody watches, you have collected a complaint and ignored it in the same week.
On the platform mechanics
A feedback request sent to somebody who has not written to you recently is an outbound message, which on the Business Platform means an approved template, a category, and consent to receive it. Whether a satisfaction survey counts as a service message or a marketing one is Meta's determination and it is revised; the rules and the pricing for each category are documented by Meta and it is the only place worth checking on the day.
The cheaper and cleaner alternative is to ask inside a conversation the customer is already having with you. A customer who just messaged about their order is a customer you can ask, in reply, at no template cost and with no consent ambiguity.
Where this stands today
PostEngage does not ship WhatsApp. On Instagram it answers comments and DMs, and it has no survey feature, no rating collection, no review integration and no analytics — there is no dashboard here that scores customer sentiment.
If the underlying goal is testimonials rather than diagnosis, the Instagram post on collecting them honestly is the closer read. For where an incoming reply lands and who watches it, support automation is here, and consent mechanics for any outbound ask are here.
