How to reply to negative reviews across a chain: rules you would not be ashamed to show a guest

Why the «we are sorry» template hurts

The standard reply reads: «Thank you for your feedback, we are very sorry, we will take it into account». It contains not a single word about the guest's situation, so it reads like an auto-responder. Worse, it confirms the complaint by silence: the guest wrote about cold soup and a rude waiter, the reply neither disputes nor checks, so it must all be true.

The opposite extreme is retelling. The reply repeats the complaint in its own words: «we are sorry the soup was cold and the waiter was rude». Now the complaint is printed twice, the second time in your voice. That is what the next guest remembers.

Seven rules for a reply

First: do not retell the negative. Apologise for the situation and for the guest's experience without listing what exactly was bad. Second: promise a concrete action you will actually take — we will check, we will go through it with the team, we will follow up. Do not promise compensation publicly unless you are ready to give it to everyone.

Third: give a contact and an invitation into a private channel. A phone number, a messenger or an email in the reply is the exit from the public thread, and it has to be visible. Fourth: reply in the language of the review, without mixing languages and without a machine tone. Fifth: address by name only when the name is beyond doubt, and never inflect it if there is any uncertainty.

Sixth: the voice of the team, not of a manager. «We» instead of «I», the register of an educated person, no exclamation marks and no emoji. Seventh: shorter than the review. A reply longer than the complaint reads as an excuse.

An example: before and after

The review: Google, three stars, a delivery order, «waited over an hour, the rolls arrived warm, the operator did not answer». Before: «Hello! We are very sorry that you waited over an hour, the rolls arrived warm and the operator did not answer. We will definitely fix everything!» The complaint is printed twice, no action named, no contact.

After: «Good afternoon. Thank you for writing, and we are sorry for this experience with your order. We will go through it with the venue and the delivery service and make sure it does not happen again. If you are willing, message us — the number is in the profile: we would like to understand the details and make it right». Three sentences, one action, one exit into a private channel, and not one repetition of the complaint.

The guest's language and register

In a chain across several markets the reply language is not politeness but a signal that the venue is local. A Romanian review gets a Romanian reply, a Ukrainian one a Ukrainian reply, and languages do not mix inside one reply even in dish names. Greeting and address follow the norm of the language, not a calque from another one.

A separate rule for the Ukrainian market that we keep in the scripts: one fixed greeting, the name never inflected if there is the slightest doubt, contacts mandatory and visible in the first lines. The prompt for draft replies is written in the market's language, because a model given a task in one language answers with its intonation.

The speed norm and who presses the button

At one location replying the same evening is easy. At twenty-five without a system replies are written by whoever has time, and the norm turns into «someday». The working norm is no more than three hours from the review appearing to a published reply, approval included.

That is achievable only when the draft is prepared automatically by approved rules and the human does one action: publishes with a button or edits with one message in a messenger. Next to the reply a private channel with the guest should open, so a small complaint does not become a public rating.

What to measure instead of the average rating

A chain's average rating hides the problem: one venue at 3.4 drowns among twenty-four at 4.6. Watch the loyalty index per location and its change against last month, the recurrence of themes — delivery, kitchen, staff — and the share of reviews answered within the norm.

Those three figures turn reviews from a reputation chore into a management signal: where the problem is systemic, where it started, and whether it was answered in time.