Satisfaction Is Not Retention

In 2018 I wrote a goal in my journal: destroy NPS. I meant it. Not because the survey is annoying. Because satisfaction does not predict retention, and I can prove it.

Start with the test that matters. The standard “proof” that NPS drives retention compares a company’s aggregate score to its aggregate performance. That is the wrong test. The valid test compares each customer’s score to that customer’s lifespan. I ran that test across the behavior of more than 130,000 customers. The correlation between NPS and retention came out at r ≈ 0.01. Zero. Promoters, passives, detractors — there is no statistically significant difference in average lifespan between the groups. The entire gap amounts to a few days of customer life over four years.

The correlation between NPS and retention is r ≈ 0.01. Zero. Promoters, passives, detractors — no statistically significant difference in lifespan. The whole gap is a few days of customer life over four years.

So why does everyone believe the opposite? Because the company-level correlation is real — and spurious. Companies that are good at producing high satisfaction also tend to be good at other things, like delivering results. It is the results, not the satisfaction, that drive retention. NPS does not cause retention. It would be just as good a guess that retention causes NPS — and even that is unlikely, since the relationship is so weak either way.

The direct evidence is worse than a null result. Companies spent tens of billions on customer experience. Over five years, satisfaction went up 13% — and retention went down 28%. Happier, and leaving faster. That is not noise around a working strategy. That is a strategy pointed at the wrong target.

Now the finding that should end the argument for good: the customers who had bad experiences stay longer. Twice as long, on average. Customers who complain and have issues are more loyal. This sounds impossible until you see the mechanism: negative experiences are a common byproduct of the process of achieving results — and results are why customers stay. A customer pushing your product hard enough to hit its limits is a customer getting somewhere. The account with no complaints is likely one that isn’t using the product enough to discover its flaws — or to achieve any compelling result.

Customers who have bad experiences stay twice as long. Negative experiences are a byproduct of the process of achieving results — and results are why customers stay.

There is one real signal buried in your surveys, and it isn’t the score. Customers who don’t respond at all have significantly less than half the lifespan of those who do. Engagement correlates with lifespan. The number the respondents write down does not.

Why does satisfaction fail this badly? Because of what satisfaction is. Satisfaction is a fraction: all the things you have, divided by all the things you want. There are only two ways to raise it — give the customer more, or get them to want less. And fractions are more sensitive to the denominator, so the fastest way to move the score is to shrink what the customer expects, not grow what you deliver.

Run your best customers through that arithmetic. High-functioning customers always have more — but they also always want much more. By the math of the fraction, they are your least satisfied. I have watched one client be thrilled with a modest result because they were measuring against a baseline of zero, while another client getting three times the result was unmoved. Satisfaction tracks the baseline, not the value delivered.

Satisfaction is a fraction: what you have divided by what you want. Your best customers always want more, so they are always less satisfied. Optimize for satisfaction and you select the customers who expect the least.

Put the two arguments together and satisfaction goes from useless to dangerous. First: it doesn’t cause retention — the correlation is zero. Second: among high-value customers, it runs inverse to value. We don’t really want customers who can be satisfied easily, because they are lower value and lower retention. So a company that optimizes for satisfaction isn’t standing still. It is selecting for its lowest-value, lowest-retention customers — the ones who expect the least. That is the trap.

And even if none of that were true, NPS fails one more way: it can’t be actioned. A score with no attached reason gives you nothing to change. “What problem are you solving for whom, and how do they measure their results” is a question you can act on. A number from one to ten is not.

Here is the whole problem in two sentences. A satisfaction survey asks the customer how they feel about you. Retention measures whether you succeeded. Neither one measures whether the customer succeeded — which is the only number that was ever going to predict anything.

Stop asking customers how they feel about you. Start measuring whether they’re getting the result they came for. One of those numbers tells you something. The other one is a mirror pointed at your own reputation.