The Renewal Math Nobody Runs
Net retention read 108%. Strip the top five accounts and it was 91%. The blended number hid which customers were carrying the base.
At one company I worked with, net retention read 108%. The number sounded healthy. The board referenced it three times in the quarterly review, and the CFO put it on slide six of the deck. It was an average across a customer base whose cohorts were behaving nothing alike.
Strip the top five accounts and our net retention was 91%. Those five had grown by a combined $2.4M over the year, while our other ninety-four accounts had shrunk by a combined $620K. The headline averaged the two and produced a figure that looked like expansion. The middle of the base was rotting.
The pattern is consistent at growth-stage SaaS companies. The top decile of customers, usually three to ten accounts, drives most of the visible expansion. Senior CSMs manage them, executives call them, and they respond to the attention. The middle cohort, the long tail of standard contracts, gets managed lighter and quietly contracts through downgrade renewals and small-seat churn. The bottom cohort churns at a predictable rate and rarely earns its own line.
A single retention number averages the three cohorts into one. The three shapes together are the truth. We were running 108% blended and 91% in the middle, which is a structural problem the top accounts were hiding. The hiding is temporary. The day one top account contracts, for any of a dozen unpredictable reasons, the headline drops far enough to expose the middle decay, and we'd have been steering against the wrong baseline for a year.
The decomposition is an hour of work. Split customers into top decile, middle eighty percent, and bottom decile by ARR. Calculate net retention for each. Report all three. The shape tells you where the company sits, the same way a clean dashboard can flatter a base that is quietly coming apart.
Most CFOs skip it because the blended number flatters and the breakdown costs a column. The breakdown is the version that tells the truth. It is also the same discipline as watching which cohort is renewing and which is bleeding. A board handed only the blended figure is steering on a metric tuned for the wrong audience, the one that wants comfort instead of visibility.