Collecting and answering reviews without making it a daily job

Reviews are the cheapest proof of trust you can get. Most stores still either never ask or ask at the wrong moment.

We run three of our own brands, nano, mate and MUSTAX, with two people. Collecting and answering reviews used to be one of those jobs that always slid to tomorrow. Today most of it runs without us. Here is the setup, including the places where we deliberately put a human in front.

Why timing decides almost everything

Most stores send the review request after a fixed number of days. Seven days after the order, whatever happened in between.

That is the core of the problem. Seven days after the order can mean the parcel arrived five days ago and the customer has settled into the product. It can also mean the parcel is still sitting with the carrier. In the second case you are asking someone for an opinion on a product they have never seen. That email does active damage.

At nano the request only goes out once the delivery event has come in from the carrier. That is a single extra condition and the strongest lever in the whole setup.

The second condition is how long the product needs to be used. You do not judge a toothbrush on the day it arrives. So we leave a few days between delivery and request, and the gap differs by product.

The triggers we set

Concretely the flow looks like this:

The three stop rules matter more than they look. Asking someone for a review while they are waiting for an answer to their problem is the surest route to a one-star rating. It happened to us exactly that way before the rule was in place.

Connecting these events is data work. The shipping status sits with the carrier, the ticket in the helpdesk, the order in the store. On our side n8n pulls these events together and then decides. You need the same data flow to automate WISMO requests. Build it once, use it twice.

The channel: email is not always the answer

Email is the default and it works. The channel should fit the customer, and your habit is a bad reason to pick one.

• Email is the workhorse channel. Good for detailed reviews with text and photos.
• The packaging insert costs almost nothing and reaches the people who never open your emails. A short link or a code, nothing more.
• The order status page in your store is the underrated spot. Anyone checking there is already interested in the product.
• Messenger works very well with younger audiences and needs its own opt-in. Without that: hands off.

What we avoid: messaging the same person across three channels. One request, one channel, one reminder after a week. Then it stops.

And keep it short. A review request with five questions and a mandatory login gets the response rate it deserves. One click on a star rating, then an optional text field. That is the whole thing.

Answering with AI: drafts yes, sending no

Answering is the second half, and it gets forgotten more often than collecting. Every future customer reads your replies. An unanswered critical review is the review itself plus the message "these people do not care".

Our setup has three levels:

Positive reviews without text. Four or five stars, no comment. Here the automation replies directly with a short, varying thank you. No human looks at it. The risk is close to zero.

Positive reviews with text. The AI drafts a reply that picks up the point the customer made. A human skims it and approves. That takes seconds per review.

Critical reviews. Three stars and below, or any text with a problem in it. Here the AI drafts a raw version at most, and a human writes the answer. Always.

That is the same pattern we run in support: the machine sorts and drafts, the human decides on anything with emotion in it. Our AI answers around 65 percent of support requests automatically, and the remaining 35 percent are exactly why that works. Our reasoning for keeping the line this strict is in AI in customer service: what actually works.

Two things you have to give the AI, otherwise it gets embarrassing: your tone of voice with real examples, and a list of things it must never promise. Refunds, exceptions to the return window, delivery commitments. That boundary belongs in the instructions, not in your hopes.

How to handle bad reviews

This is where the store that treats reviews as marketing parts ways with the store that reads them as feedback.

How we handle a critical review:

1. Respond within 24 hours. Later reads as damage control.

2. Keep the public reply short, factual, no justifications. What went wrong, what you are doing, how the customer can reach you. Three sentences are enough.

3. The fix happens in private. Refunds and exceptions do not belong in a public thread.

4. Follow up once it is solved. Some people change the review on their own. Never ask for it, never offer anything in exchange.

5. Write down the patterns. Three reviews about the same point are a product or process defect, and treating them as a support case wastes the signal.

Point 5 is where the real value sits. At mate, repeated remarks about unclear instructions led us to redo the insert. After that, questions on that topic dropped noticeably. The reviews were only the messenger.

When criticism circles around returns, the problem is usually the path to the return rather than the return itself. A self-service returns portal removes that friction at the root.

The hard line: no fakes, no filtering

This is the section where there is nothing to weigh up.

Faking reviews is off limits. No bought reviews, no self-written ones under invented names, no incentives tied to positive ratings. That is illegal in the EU, and it goes well beyond being tacky. It also gets caught reliably, because platforms check properly now.

Filtering is equally off limits. I mean the popular pattern of asking about satisfaction internally first and only sending the happy customers to the public review page. That distorts the public rating, and it is legally attackable too.

Incentives only if they are neutral. A voucher for every review, independent of the star rating, is allowed and has to be disclosed. A voucher for good reviews is not.

Delete only for real violations. Insults, obvious spam, someone else's order. A justified bad review stays up.

The practical reason beyond the legal one: nobody believes a wall of five-star ratings any more. An average of 4.5 with visible critical voices and good replies underneath sells better than a flawless facade. Trust grows out of visible friction, and a perfect surface kills it.

What the setup does not do

Three limits.

Automation does not make a bad product good. If your average sits at 3.2, that is not a communication task. Your customers are telling you something very clear about the product.

The volume has a ceiling. Even with perfect timing, only a small share of your buyers leaves a review. Anyone promising rates far above that is usually talking about bought reviews.

Reply drafts need maintenance. After a few months every answer sounds the same, because the same phrases keep coming back. Read twenty published replies in one sitting once a quarter. If they read like a loop, the briefing needs work.

And as everywhere: without measurement you do not know whether it works. Requests sent, reviews received, response time on critical reviews. Three numbers are enough. Which other metrics in customer contact are worth tracking is in customer service KPIs.

Conclusion

Automating reviews means automating the mechanics and keeping the judgment with you. You never hand a machine the decision about what a review means.

The mechanics are: ask after delivery instead of by calendar, stop on open problems, pick one channel instead of three, have positive replies drafted for you, write the critical ones yourself. That is built in a day and then runs for years.

The judgment stays yours: which piece of criticism is a one-off and which one is a pattern. You should not want to hand that part over, because it is the actual return on the whole exercise.

If you want to sort out the support around it, the guide to customer service automation is the right next step. And if you want to know whether your carrier even delivers the events you need: talk to us at Flowhouse.