Your rating went up. Your bookings went down. Both can be true.

A star rating is the most-watched number in most service businesses and one of the least informative. It moves slowly, it is dominated by history, and it compresses everything anybody said about you into one figure with no shape. It is perfectly possible for it to rise for a year while the experience it summarises gets worse.

Three ways an average misleads

Volume weighting

A business with 900 reviews at 4.6 needs a very large number of bad months to move the headline figure at all. The average is mostly a record of who you were three years ago. Every recent customer is diluted by a crowd of old ones, which is comfortable and useless.

Recency blindness

The relevant question is never what is the average. It is what has the last ninety days looked like compared with the ninety before that. A business sliding from 4.7 to 4.1 in recent reviews while the lifetime average holds at 4.6 is a business with a problem that its dashboard is actively hiding.

Mix blindness

A 4.0 made of consistent fours describes a reliable, unremarkable business. A 4.0 made of fives and ones describes a business that is excellent for one kind of customer and a disaster for another. These require opposite responses, and the average cannot tell them apart.

Read the spread before the mean

The single most useful reputation number nobody looks at is the difference between your best-rated platform and your worst. When that gap approaches a full star, it is almost never random. It usually means one of three things, and all three are actionable.

• Different customers. The audience on one platform is buying something different — price-led rather than service-led, or booking a different product entirely.
• Different moments. One platform captures people at the point of booking, another after delivery. A gap between them is a gap between what you promised and what arrived.
• Different prompting. You ask happy customers for a review on one platform and nowhere else. This inflates one number and tells you nothing, which is worse than not asking at all because it looks like data.

Concentration is a risk, not an achievement

If the large majority of your public feedback sits on a single platform, you are not reading your customers. You are reading the subset of them who use one site. That is fine until the platform changes its ranking, its filter, or its review-solicitation rules — at which point a business with all its reputation in one place discovers it has no reputation anywhere else.

It also distorts the analysis. A theme that dominates a single platform looks like a company-wide problem, and a theme that is genuinely company-wide looks small because the other platforms were never read.

What to track instead

1. Recent trend, not lifetime average. Last quarter against the quarter before, per platform.
2. Theme frequency. How many times a specific complaint appeared, counted — not sensed.
3. Platform spread of each theme. One platform is noise; three unrelated ones is a fact about the business.
4. The gap between your best and worst platform. Above roughly a full star, go and find out why.
5. Concentration. What share of everything said about you sits in one place.

None of these require a tool to compute. They require somebody to read the same sources on the same schedule and count the same way each time — which is the part that almost never survives contact with a busy quarter, and the reason this is work worth putting on a cadence.