The review had three stars and said, among other kind things, one concrete sentence: "the shower water took forever to come out hot." Chema skimmed it one Sunday night, on his phone, among twenty other things, thought "oh well" and closed it. He never thought of it again.
Eleven months later, that same sentence — said in many ways, by many different people — had turned up in another fourteen reviews. "We waited five minutes for it to run hot." "The shower, so-so." "The water took a while." All from the same house, number two. None had set off any alarm, because each had landed loose, on a different night, on a different platform, among the noise of the rest.
Chema runs three country guesthouses in the Sierra de Gredos. A lovely business, the kind that's tended with care. Small: him, his wife, two cleaners and a lad for the maintenance. The houses filled up almost all year and yet, for a while now, the average rating on the platforms had dropped from a high nine to a low eight, slowly, with no clear reason. And dropping in rating, in his business, means dropping in bookings.
What the guests were already telling him
Chema's reviews lived spread across four places. The big booking platform, the internet map, another rural-stays website and the odd loose comment on social media. Between them all, around three hundred came in a year. Three hundred opinions from people who had slept in his houses and gone to the trouble of saying how it was.
Nobody read them all. Not through laziness: through impossibility. To read them all, gather them, and realise that fourteen people in a year mentioned the hot water in house two, someone would have had to sit down and go through the four platforms, one by one, copying onto a sheet of paper what kept repeating. Nobody did that, because between changing sheets, answering the phone and welcoming guests there's no gap to sit and play librarian of the reviews.
So Chema was living the worst of situations without knowing it: his customers were telling him, in full detail and for free, exactly what to fix to get back to a nine. And he didn't hear them. Not because they weren't speaking. Because they were speaking spread out, one at a time, and nobody added the voices together. The hot-water problem in house two could be solved with a new forty-euro heater. It had been costing him bookings for a year because he didn't know it existed.
What he asked for and what he needed
Chema wrote to me because someone had told him about "putting in an artificial intelligence to answer the reviews". That's what he thought he needed: something to reply, fast and politely, to people. And, while at it, he asked me whether the happy customers could be "nudged" to leave five stars, to lift the average on the back of good votes.
Answering reviews with a machine was solving the least of it. Replying is fine, it looks elegant, but replying "sorry about the water, we'll bear it in mind" to fourteen different people doesn't fix the water. It's good manners on top of a problem that's still there. And nudging the happy ones for five stars was covering his own eyes: lifting the rating without fixing what was lowering it is masking the fever without looking at why there's a fever.
What Chema needed wasn't to talk more. It was to listen better. To find out what his customers were telling him as a whole, what kept repeating, what was really going wrong and what was going so well it deserved showing off. The reply came afterwards; first he had to hear.
—So I don't answer the reviews?
—Answer them, of course. But first let's read them all together for once, and see what they've been telling you for a year.
How it works, no smoke
What we built is nothing out of this world, and it's worth telling without mystery. A system collects all the new reviews from the four platforms and gathers them in one place. That alone, having them together, is half the battle, because the problem began with them being scattered.
Then it reads them. Not one by one to reply to them: it reads them to understand what they're about. It files each review by topic — cleanliness, sleep, welcome, breakfast, hot water, noise, equipment — and looks at whether it's good, so-so or bad on each. A review that says "everything lovely, but the water took a while" it understands whole: good overall, bad on the water. It doesn't stop at the global score, which hides the detail.
And once every Monday, it sends Chema half a page. Not three hundred reviews: half a page. "This week, eight opinions. Most praised, again, the breakfast. For the fifth week running, the hot water in house two: three people mention it. A new complaint about the wifi in house three." What really matters, separated from the noise, on a sheet you read in two minutes with your coffee.
The honest thing is to say what it doesn't do. It doesn't fix the heater, that's the plumber's. It doesn't reply on its own, because Chema prefers to reply himself, in his tone, and he's right to. And it doesn't make things up: if one week nobody says anything odd, the summary is short and dull, which is exactly what you want to read. It doesn't invent problems to look useful. It only sees to it that no repeating voice goes without Chema hearing it.
What changed
The hot water in house two was fixed in the first week. Forty euros of heater and an afternoon from the maintenance lad. A problem that had been gnawing at his rating and his bookings for a year, and that was written, in black and white, across fourteen reviews nobody had pulled together.
But the summary didn't only bring complaints, and that was the pleasant surprise. Week after week, the most praised thing was the breakfast: bread from a village bakery, the jam his wife made. It was the houses' greatest treasure and it appeared nowhere on his website, because to Chema it seemed an unimportant little thing. We put it on the front page, with a photo. It turned out to be one of the things that made people book most.
In a few months, the average rating climbed back towards nine. Not by asking for stars, nor by replying faster. By fixing, one by one, the few concrete things customers had long been pointing out, and by showing off the ones he was already doing well without knowing it. The bookings followed, as they always follow the rating in that business.
Listening to what they're already telling you
There's something Chema told me, a little embarrassed, that is the heart of all this.
—They'd been telling me about the shower for a year. I wasn't reading it.
Almost no company has a problem of the customer not telling them things. Today the customer says them more than ever, in reviews, in surveys, in messages, in stars. The problem is one of hearing: those opinions arrive loose, spread across four places, one at a time, and nobody has time to gather them to discover that the same sentence shows up fourteen times. The point isn't that the information is missing. It's that it's broken into hundreds of pieces and nobody puts it back together.
Serving the customer well isn't only replying quickly. It's finding out what they've long been telling you and doing something about it. And that, when it's hundreds of scattered voices, today can be gathered and truly listened to: not so a machine speaks for you, but so you hear, at last, what your customers have spent a year trying to tell you.
The customer already tells you what to fix; the problem is hearing it
Opinions arrive loose, spread across four platforms, one at a time. Each complaint lands isolated on a different night and sets off no alarm, even though the same thing repeats fourteen times in a year. The information is there, but broken into pieces nobody puts back together.
Gathering every review in one place, reading them by topic and summarising each week what repeats — the bad to fix it and the good to show it off — doesn't reply for you: it lets you hear, at last, what your customers have long been telling you. Listen better, before talking more.
Do you know what your customers have spent months telling you?
If your reviews and opinions are spread across several platforms and nobody has time to gather them, something can be set up that pulls them together, reads them by topic and summarises each week what really repeats. We can look at it together.
Let's talk