The line appears on the same slide in every growth deck: improve retention by five percent and profit rises anywhere from twenty-five to ninety-five percent. It gets a nod, a circle, a shared agreement that more of that is in order. Almost nobody asks where the ninety-five came from, or what separates it from the twenty-five.

The number traces to Frederick Reichheld's 1990 Harvard Business Review piece on zero defections. That work watched a handful of service firms and measured profit per customer, not profit at the company level. It also assumed, quietly, that keeping a customer is free. Byron Sharp has spent years calling that framing a thought experiment dressed as research. Bain's own 2001 brief walks the claim back to something narrower: in financial services, five percent more retention produces more than twenty-five percent more profit. That is a real finding, with a fence around it.

What gets quotedWhat the source actually said
Retention up five percent, profit up twenty-five to ninety-five percentA range drawn from a few service firms, per-customer economics
Acquiring costs five to twenty-five times retainingDirectionally useful, rarely measured cleanly inside one product
Loyal customers are always the profitable onesTenure-to-profit correlation sits between zero point two and zero point four five
Any five-percent lift is worth chasingOnly lifts among customers worth keeping actually pay

The peer-reviewed pushback is more useful than the myth. Gupta, Lehmann and Stuart, writing in the Journal of Marketing Research in 2004, modelled five firms and found that a one-percent retention gain raised customer value by three to seven percent. A one-percent gain in margin raised it by about one percent. A one-percent improvement in acquisition cost moved value by a rounding error, zero point zero two to zero point three percent. Read that slowly: retention earned roughly ten times the leverage of acquisition efficiency in those models, but the margin number was not nothing, and the acquisition number was not irrelevant. The honest statement is comparative, not absolute.

Then Reinartz and Kumar muddy the water further, in the best way. Across sixteen thousand customers at four firms, their 2002 Harvard Business Review work found the link between how long a customer stays and how much that customer earns sat between zero point two and zero point four five. About half of the customers labelled loyal were barely profitable at all. Long tenure and good economics overlap, but they are not the same customer. A growth team that copies the Reichheld slide without segmenting its base will happily spend money keeping people whose departure would improve the margin.

Here is the uncomfortable part for the retention-at-all-costs crowd. The famous ratio assumes the retained customer keeps buying at the same rate and costs nothing extra to keep. Both assumptions break in consumer products. Discounts, points, win-back offers and concierge support are real costs. When the incentive is generic, the product trains a price-sensitive cohort to wait for the next coupon, and the retained account arrives at a permanently lower average value. The metric improves. The business does not.

A hosted mini-program runtime does not resolve the argument, but it changes the cost side of it in a concrete way. Updates ship to the host app without a store review, so a team can adjust an offer or a flow for one segment and watch the effect within a day. The shared runtime carries login, payment and identity, so serving a small high-value cohort does not require building a separate surface for them. That makes the disciplined version of retention cheap to run: keep the segment worth keeping, stop subsidizing the rest, and measure profit per retained customer instead of the retention rate alone. The five-percent number survives contact with reality only when someone checks which five percent.

Sources