Research library Customer value, acquisition and retention
Nobody Got More Loyal.
A cohort's retention rate climbs every year. Fader and Hardie showed why: the customers likely to leave go first, so the ones left were always the stayers.
Peter S. Fader and Bruce G. S. Hardie, 2007
Each year's retention rate, two customer segments
What they were looking at
In 2004 Michael Berry and Gordon Linoff published survival curves for two segments of a subscription business: “Regular” customers and “High End” customers. Of every 100 Regular customers, 63 were still active after one year, 38 after three and 17 after twelve. Their book argued, as Fader and Hardie quote it, that “parametric approaches do not work” for projecting curves like these.
Peter Fader of Wharton and Bruce Hardie of London Business School took the first seven years of that data and tried to forecast the next five.
What they found
The usual regressions missed badly. On the High End segment, the linear and exponential fits underestimated year-12 survival by 81% and 30%, and a quadratic overestimated it by 92%. Their own model, the shifted-beta-geometric, treats every customer as having their own fixed chance of leaving each year, and it fits in a spreadsheet. It “overestimates year 12 survival by only 4% and 2%.”
“The resulting predictions are almost too good to be true” Fader & Hardie, authors’ 2006 version, p. 9
The model works because of what it assumes about the people inside the curve:
“the observed phenomenon of increasing retention rates is simply a sorting effect in a heterogeneous population” p. 10
What it means
Look at the chart. The Regular segment’s retention rate climbs from 63% in the first year to about 94% by the twelfth. It’s tempting to read that as customers growing more loyal over time.
Nobody has to change for that line to rise. Every cohort is a mix of customers who were always likely to leave and customers who were always likely to stay. The leavers go first. Each year the survivors are a more loyal group than the year before, so the rate goes up, even if no individual customer’s odds ever move.
Where it shows up in a revenue team
- A blended churn rate mixes cohorts of different ages. A book full of new customers shows high churn; the same book a few years later shows low churn, with nothing changed but its age.
- “Retention improved this year” may only mean that last year’s likely leavers have already left.
- A growth spurt makes churn look worse, because it adds a wave of young customers who haven’t sorted yet.
- Compare like with like. Read churn by cohort, and by tier within the cohort: first-year retention against first-year retention.