The three reviews that nearly closed my shop — and the afternoon that saved it

How it started

We are six people and a warehouse unit. We sell lamps, throws, ceramics, the kind of things people buy as gifts in November. Roughly a third of the year's revenue lands between the second week of November and Christmas, and almost all of it comes through paid search and social. If the ads stop converting, there is no business.

On a Tuesday in early November I got the usual notification: a new review. One star. “Absolute scam, took my money and never sent anything, avoid.” I searched the name. No order. I searched the email domain they had used for the review account. Nothing. The next day there was another one, different name, same complaint, almost the same wording. By Friday there were three.

Before that week we had 24 reviews and a rating of 4.3, which on Trustpilot is the bottom edge of “Excellent”. Three one-stars on a profile that size took us to 3.9. Overnight the label under our name said “Great”, which sounds fine until you realise it is the band below the one your competitors are in.

What it actually cost

The number I watched was not the rating. It was the conversion rate on our paid traffic. In the two weeks the reviews were up it fell by 38 per cent against the same fortnight the year before, with more traffic and a better-looking site. The ads were working. People were clicking, reading “scam” in the first review on the page, and leaving.

I know that is what happened because I did it myself. I opened our profile in a private window and the three reviews were the first three things on the page, sorted by date. If I had been a stranger about to spend eighty pounds on a lamp from a shop I had never heard of, I would have left too.

At our margins, that fortnight was the difference between a normal Christmas and not making January's payroll. I had already ordered the stock.

What I got wrong first

I replied to all three reviews, publicly, the same evening. I was polite, but I was clearly angry, and I said we had no record of them as customers and that the reviews were fake. I thought I was setting the record straight. What I had actually done was put the word “fake” underneath the word “scam” and turn a review into an argument. Nobody reads an argument between a shop and a stranger and comes away reassured.

Then I had a message from a company offering to “push the negatives down” with new five-star reviews, forty of them, for a fee. I came closer to saying yes than I would like to admit. What stopped me was reading the platform's guidelines, which say in plain words that this is the kind of activity that gets a warning banner put above your profile. A banner would have finished us.

Finally I reported the three reviews through the normal button. I chose the reason that seemed closest, wrote a sentence in the box, and waited. One came back within a day: not enough information to take action. The other two sat in review while my conversion rate sat on the floor.

The afternoon that changed it

A friend who runs a bigger shop told me, fairly bluntly, to stop treating it as a customer-service problem and start treating it as an evidence problem. The platform does not remove a review because the business is upset. It removes a review because there is a ground for removal and the ground is demonstrated.

So I spent an afternoon building a file. For each review: the date the account was created (all three within the same four days), the fact that it was the account's only review, a search of our order system by name, email domain and postcode showing no match, and the near-identical wording across the three. Written up, that pattern is obvious. In three separate report forms, each a sentence long, it is invisible.

I will be honest about the last part: I did not file it myself. I handed the file to a firm that does this every day, because I had a warehouse to run and I had already got it wrong once. They reframed it as one case rather than three reports, filed it through the right route, and all three reviews were confirmed removed, each with a written confirmation from the platform. The score went back to 4.3. The label went back to “Excellent”.

What happened next

Conversion came back within about ten days. Not instantly, because the people who had seen the reviews did not come back, but the new traffic behaved like the previous year's traffic. We had our normal Christmas. January's payroll was fine.

The reviews had been up for sixteen days in total. Nine of those were me replying, hesitating and filing reports with no evidence attached. I think about that a lot.