We collected 180 new reviews to bury one. People still quoted it back to us

The review

We run online courses for people changing careers. Our students are adults spending a meaningful amount of their own money, and they research us hard before they enrol. We had 61 reviews and a 4.2 when the review that this story is about arrived.

It said we had taken a payment, refused a refund, and ignored every email. It gave dates. It was well written and completely specific, which is exactly what makes a review persuasive. It was also untrue in the way that matters: the person had been refunded in full, eleven days after asking, and we had the bank record and the email thread to prove it. They had posted the review on day six and never updated it.

The plan: outnumber it

Our instinct was that this was a volume problem. One bad review in 61 was visible; one bad review in 300 would be noise. So we built a review-request flow into the end of every course module, we emailed every graduate from the last two years, and we asked. Properly, without incentives, which I am still glad about.

It worked, in the sense that it was meant to. In four months we collected 180 new reviews with an average just under 4.8. The overall rating went from 4.2 to 4.6. We were pleased. We put the number in our ads.

What did not change

Our sales team kept hearing the same thing on calls: “I saw a review saying you don't refund people.” Not a review. The review. Months after it was posted, with 180 newer reviews above it, prospects were still finding it and quoting it.

When I looked at how, it was obvious. Anyone weighing up a four-figure spend does not read the five-star reviews. They click the one-star filter, because that is where they expect to find the truth. And when they did, ours was the longest, most specific, most recent one-star review on the page. The 180 new reviews had changed the average. They had done nothing to the review.

A score is a summary. A review is a story. People buy, and refuse to buy, on stories.

The second approach

What finally resolved it was embarrassingly direct. The review stated that we had refused a refund. We had refunded. That is a false statement of fact about a transaction, and a false statement of fact is a ground for removal under the platform's own rules, independently of how we felt about it.

The case was the bank record showing the refund, the email thread showing the dates, and the review's own claims set against them line by line. We had a specialist check the file and file it, because our previous attempts at reporting things through the standard form had gone nowhere and I wanted it to be the last attempt. It was filed on a Monday and the review was removed and confirmed by the Thursday.

The 180 new reviews, of course, stayed. With the one review gone, the profile sat at 4.7 and, more to the point, the first thing in the one-star filter was now a mild complaint about video quality from 2024 that we had answered at the time.

What I would do differently

Both things, but in the other order. Deal with the review that should not be there first, because every week it stays up it is being quoted in someone's decision. Then build the review habit, because that is what protects you from the next one.

The habit is still running. We are at just over 400 reviews now. When a one-star review that is fair arrives, and they do, it sits among hundreds of others and a sensible prospect reads it as what it is. That is what the volume was for. It was never going to remove anything.