Your reviews are a dataset, not a reputation problem.

There is a ritual performed nightly in this industry, usually after close, usually on a phone. The operator opens the review platforms and replies. Thank you for your feedback. We’re sorry to hear. We take this seriously. High praise gets a warm line; a two-star gets a measured apology and, occasionally, a small war. Then the phone goes dark and nothing in the building changes.

I performed this ritual for years. I was quite good at it. It achieved almost nothing, and it took me an embarrassingly long time to understand why: I was treating reviews as a reputation problem, when they are the cheapest operational dataset a venue will ever be handed.

Consider what a month of reviews actually contains. Not opinions — measurements, badly formatted. Somebody timed your Thursday service without being asked. Somebody audited the temperature of the terrace at 9pm. Somebody compared your sea bass to a competitor’s, unpaid. Two thousand words a month of unsolicited field research, and the industry’s standard response is to reply to each unit of it individually, politely, and then file it under feelings.

Read one review and you have an anecdote. Structure a hundred and you have signals no mystery diner could produce: which dish drags your food score, which shift generates the service complaints, whether "noise" appears on Fridays or every day, what people actually mean when they say value. The pattern is invisible review by review and obvious in aggregate — which is precisely why nobody sees it, because nobody reads their reviews in aggregate. They read them one at a time, at midnight, emotionally.

So at RAYN we ingest a venue’s public reviews and extract the structure: themes by dish, by shift, by aspect — food, service, noise, value — trend lines, and suggested replies in the venue’s own voice for the ritual that still has to happen. The interesting output isn’t the reply. It’s the Tuesday-morning question the data puts to the operator: this pattern exists — what changes on Thursday?

An admission, since this letter would be smug without one. We also built a system that tries to match public reviews to the actual guests who wrote them — connect the two-star to the table, the visit, the booking. Elegant idea. Its current success rate rounds to zero, because anonymous reviewers are, it turns out, anonymous. I keep it running out of stubbornness and it keeps teaching me the same lesson: the public review is the guest’s last resort, the thing they write after the visit where nobody asked. The venues that win this game aren’t the ones that match reviews to guests. They’re the ones that ask before the guest reaches for the megaphone — which is why the private post-visit question, sent the morning after, matters more than every public platform combined.

The reframe, then, for the next time you’re replying at midnight: the reply is customer service. The pattern is operations. Answer the review for the audience reading over your shoulder — but let the aggregate change what the building does, because "thank you for your feedback" was never meant to be the whole transaction.

You answered the review. Did anything in the building change?

Next week: the real cost of your booking platform — who owns the guest, what per-cover fees actually add up to, and why switching feels impossible when it isn’t.

— B

Disagree with any of this? Message me. I answer.

Bhrij Patel is the founder of RAYN, an AI-native hospitality intelligence platform, built after years operating restaurants from quick service to fine dining. For an honest, vendor-independent read on where AI fits your operation, ask about the AI Readiness Audit