There's a particular kind of confidence that a number carries when it has a decimal point in it. An NPS of 47 sounds like it was measured. A 4.28 star average sounds like someone counted carefully. The format does a lot of quiet work here. It suggests that behind the figure sits a process precise enough to justify resolving your customers' feelings to a hundredth of a point.
Usually it doesn't.
Take the 4.28. To land on that second decimal you'd need enough responses that adding one more barely moves it. Most small businesses aren't working with those volumes. If your average is built on 30 or 40 responses, the difference between 4.28 and 4.35 is one person who had a bad morning, or one regular who felt generous. The digit past the decimal isn't measuring your business. It's measuring who happened to fill in the form that week.
Run the arithmetic and it gets uncomfortable. With 25 star ratings, a single respondent switching from a 5 to a 2 drags the average down by more than a tenth of a point on its own. That's not a signal about the quality of your service. It's one person, and the maths amplifies them into what looks like a trend. When the number ticks from 4.4 to 4.3 and someone books a meeting about it, they're often reacting to statistical noise wearing the costume of a result.
NPS has its own version of this. The score is a subtraction: the percentage of promoters minus the percentage of detractors. On a small sample, both of those percentages jump around, and subtracting one from the other stacks the wobble. An NPS that reads 47 one month and 39 the next can be the same underlying reality sampled twice. But 47 and 39 look like a story. So a story gets told, usually one that flatters or panics whoever is reading it.
The scale underneath makes this worse, because it was coarse to begin with. When a customer gives you four stars, they haven't measured anything. They've rounded a fuzzy feeling to the nearest option you offered. Maybe they were a soft five who didn't want to seem gushing. Maybe they were a generous three. You have no way of knowing, because the star scale only has five stops on it. Then you take a pile of those rounded-off gut calls, average them, and report the result to two decimal places. You've manufactured precision that none of the individual answers ever had. The respondents resolved their experience to a whole number. The dashboard resolves it to a hundredth and hopes you won't ask where that came from.
I think the real damage isn't the inaccuracy itself. It's what false precision does to how you behave. A number that looks exact invites you to manage it like it's exact. You start watching the decimal. You set a goal of moving from 4.3 to 4.4 by the end of the quarter. You treat a two-point NPS dip as a problem to be solved rather than the sampling flutter it probably is. The apparent precision pulls your attention toward the number and away from the thing the number was supposed to point at.
And the number, however many decimals you dress it in, can't point very far. It can't tell you that the dip came from three separate people all mentioning the new opening hours. It can't tell you whether the promoters and the detractors are describing the same visit or two completely different businesses. Those are the things worth knowing, and they don't live in the average. They live in what people actually wrote, and in the pattern across enough responses that the pattern stops being one loud voice. This is the same reason a 4.2 rarely tells you where to start: the aggregate strips out exactly the detail you'd need to act.
None of this means metrics are useless. A score tracked the same way over a long enough stretch, read as a rough direction rather than a verdict, does tell you something. The trouble starts when the precision of the display gets mistaken for the precision of the underlying data. A weather forecast that said "63.4% chance of rain" wouldn't be more trustworthy than one that said "probably rain." It would just be pretending to know more than it does. Feedback scores do the same thing, and we mostly let them.
The more honest way to read one is to give it error bars in your head. When your average moves by a tenth, ask how many responses that tenth is standing on before you treat it as news. When your NPS swings, look at the raw count of detractors, not the headline number. If two extra grumpy customers can rewrite your quarter, the quarter wasn't really rewritten. If the same complaint shows up in fifteen open-text answers, that's worth more than any digit past the decimal, and no averaging step will ever surface it for you.
This is roughly why Qria leans on plain-language summaries across your feedback and your public reviews together rather than a single hero number: the useful part is usually the "why" sitting behind the score, and a summary can carry that where an average can't. If you're weighing up which score to build on in the first place, the trade-offs between NPS, CSAT, and the rest are worth reading before you commit to chasing any one of them.
A decimal point is easy to print. Earning it is a different matter, and most feedback scores never do. Round accordingly, and spend your attention on the sentences instead.
