A churn spike coincides with a competitor launch and a pricing change, three people back three different explanations, and Maya looks for the observation that separates them instead of the one that supports hers
Three Stories, One Month — Volume I, Understanding Humans Before Products · Synthesis · Dilemma · Practitioner. A churn spike coincides with a competitor launch and a pricing change, three people back three different explanations, and Maya looks for the observation that separates them instead of the one that supports hers
01 · Opening
A churn spike coincides with a competitor launch and a pricing change, three people back three different explanations, and Maya looks for the observation that separates them instead of the one that supports hers
02 · The setup
March did three things at once, which is the least convenient number of things for March to have done.
March — Churn: 1.2% to 2.8%; Competitor launched: week 1; Our price rose 12%: week 2; Onboarding redesign shipped: week 2
03 · The setup
Sam: It is the price. Twelve percent is a lot and the timing is exact.
04 · The evidence
Priya: Their launch was all over the industry press that week.
Dev: And we changed the first screen every new account sees.
05 · The evidence
Three explanations, three competent people, and a month of data that fits all of them.
Maya thinks: Everyone has evidence. Nobody has evidence that rules the other two out.
06 · The evidence
Maya: Stop arguing about which is likeliest. What would we see if yours were true that we would not see if theirs were?
07 · The evidence
That question turns three opinions into three predictions, and predictions can be checked.
Priya: If it is the price, the churn should sit on accounts whose renewal fell in March.
08 · The evidence
Three predictions. One afternoon of queries, all against data they already had.
What each story predicts — Price — churn concentrated at renewal: check renewal dates; Competitor — churn across all tenures: check tenure spread; Onboarding — churn among March signups: check signup month; Data needed: none new
09 · The evidence
The answer is not the one the loudest person wanted and not the one Maya expected either.
Churned accounts in March — Renewal fell in March: 31%; Signed up in March: 6%; Tenure over two years: 78%; Mentioned competitor in exit note: 9%
10 · The evidence
Maya: Long-tenure accounts, mostly not at renewal, barely mentioning the competitor.
Dev: So it is none of our three.
11 · The evidence
When several explanations fit the same evidence, arguing about which is most plausible cannot settle it, because plausibility is exactly what they all share. The argument will be won by whoever is most senior or most certain, and neither of those is a source of truth.
The useful move is to ask each story what it predicts that the others do not, which converts opinions into checkable claims. Three people who cannot agree on a cause can usually agree on what they would expect to see.
The separating observation is often smaller than the argument it ends. Here it was one afternoon against data already collected, and it eliminated all three candidates — which is a real result, because a week of debate would have shipped one of them.
12 · You make the call
The story does not tell you first.
A. Weigh which is most plausible and act on it
B. Ask what each predicts that the others do not
C. Wait for another month of data
13 · What happened
The real cause turns out to be a silent failure in a two-year-old export path, found the following week by looking at what long-tenure accounts had in common.
Maya thinks: We nearly rolled back a price rise that was not doing anything.
14 · Complete
When explanations tie on plausibility, ask what each predicts that the others do not. Next: knowing when to stop investigating and just decide.