A score out of ten does not tell you what to do on Monday

Somewhere in your business there is a number that is meant to tell you how customers feel. It might come from a survey that goes out after a job. It might be a rating out of ten, or a percentage, or a score that gets reported each quarter.

Say it went up two points. Good news, presumably.

Now try to answer the next question. Why did it go up, and what should anybody do differently on Monday morning?

If nobody can answer that, the number is not doing the job you are paying for it to do. And it is usually not because the survey was badly run.

What the number is made of

The most common of these asks how likely someone is to recommend you, on a scale of nought to ten. It is called Net Promoter Score and it is the default in most industries. The same objections apply to almost any satisfaction rating.

Notice what is being asked. Not whether they recommended you. Whether they think they might. It is a prediction of future behaviour, made by a person who is not thinking about it very hard, in a moment that costs them nothing.

Between thinking you might recommend a business, actually recommending it, and the relationship still existing in two years there are two large gaps. The score measures the first and is used to talk about the third.

Why it moves without anything else moving

This is the part that matters, and it is not an accusation.

These scores are extraordinarily sensitive to who you ask and when you ask them. Send the survey right after a job that went well and the number rises. Send it to your long-standing customers rather than your recent ones and it rises. Have the person who just solved someone's problem mention that a survey is coming and it rises.

None of that is dishonest. Every one of those choices is defensible on its own terms, and people make them for sensible reasons. But once anyone is responsible for the number going up, the cheapest way to move it will always be to adjust the asking rather than the experience. That is not a discipline problem to be solved with a policy. It is what happens whenever a number becomes a target.

Which leaves you with a figure that reliably goes up and an experience that may not have changed at all.

The half of it that is true

There is a real thing underneath, and it is worth saying, because the honest case against the number is stronger than the unfair one.

These scores do capture something. They capture memory, which is dominated by the best moment and the ending rather than by an average. That is genuinely the thing that decides whether someone returns.

So the score is not meaningless. It is a rough reading of memory. The trouble is that memory is exactly the thing you can adjust by changing when you ask, which is why the same property makes it both worth something and easy to move. And even at its most accurate it tells you the temperature without telling you where the draught is coming from.

Count things people did

The alternative is not a better survey. It is to count behaviour, which your business is already recording and which nobody can be polite about.

• Did they come back? Repeat rate, measured over a sensible window for your kind of work. The plainest loyalty measure there is.
• Did they actually refer anyone? Not whether they would. Whether a name arrived. Ask every new customer how they found you and write it down.
• How hard was it for them? If you are going to ask one question, ask how much effort it took to get sorted. Effort predicts leaving far better than enthusiasm predicts staying.
• Who left, and when? Most businesses cannot say how many customers they lost last year. It is the most expensive number nobody is tracking.

None of those need a platform. They need someone to write things down consistently for a quarter. And each one arrives attached to a specific customer and a specific job, which means you can go and look at what happened rather than speculating about what a score means.