Customer lifetime value is an average taken over people who have already been acquired, which makes it a description of the past wearing the costume of a forecast. The customers inside that average arrived through the cheapest channels first: word of mouth, search intent, an early cohort that went looking for the product. Every additional customer has to be found somewhere less efficient than the last, and the ones found there behave differently — persuaded rather than motivated, and quicker to leave. So the blended number holds steady while the marginal number falls, and a company can watch a healthy CLV all the way into insolvency.

Blue Apron is the clean case. Its IPO filing showed roughly $795 million in 2016 net revenue against about $144 million of marketing spend, a ratio that only works if the customers bought with that $144 million stay long enough to repay it. Outside analyses of its cohort data consistently found a majority of customers stopping within their first year. The company went public in June 2017 at $10 a share; Amazon announced its own meal-kit entry the same month, and the stock never recovered. Blue Apron was eventually sold to Wonder Group in 2023.

The fix is not a better CLV formula but a different denominator. Reichheld and Sasser's 1990 finding — that a five percent improvement in retention could raise profits by twenty-five to ninety-five percent — is usually read as an argument for loyalty programs. It reads better as a statement about leverage: retention compounds and acquisition does not, so a company buying growth is buying the one input with no compounding attached. What a CLV number cannot tell you is whether the next cohort will resemble the last, and by the time the average finally moves, several quarters of spending are already committed.