Skip to content
woodfrog
← Deep dives
What drives what, and by how muchB2B SaaS and subscriptionsUse case, not a client result

What actually drives customer churn, and how fast?

Rank the factors behind churn, with the size of each effect and how soon a customer is at risk.

The decision it informs

Where to put the next unit of customer success effort, and which accounts to act on this month.

Why a dashboard cannot answer it

A dashboard shows who left. It cannot separate what caused it from what merely happened at the same time, or say how much one more support ticket, a missed onboarding step or a price rise changes the odds.

What we ask in the first week

The questions behind the question.

  • How do you define a lost customer, and when does the clock start?
  • Which actions does the team control: onboarding, support, pricing, account reviews?
  • What would you do differently if one factor mattered twice as much as another?
  • What have you already tried, and what did you conclude from it?
Data we would look at
Subscriptions and renewalsProduct usage eventsSupport tickets and response timesInvoices and price changesAccount notes from customer success
How we would analyse it

Methods, chosen for this question.

Survival analysisHow long customers stay, and how each factor changes the chance of leaving over time.
Driver modelThe effect of each factor with the others held fixed, with a range, not a single number.
Cohort comparisonCustomers who were alike except for one thing, to separate cause from coincidence as far as the data allows.
  1. 01Scope
  2. 02Immerse
  3. 03Analyse
  4. 04Test the evidence
  5. 05Read out and hand over
What you would get

Findings you can act on.

A ranked driver tableEach factor, the direction, the size of effect per unit change and a confidence range.
A risk windowHow early the signs appear, so the team acts while there is still time.
A what-if workbookChange support response time or onboarding completion and see the effect on retention.
How a driver ranking readsSchematic, no values
Each bar is the size of a factor's effect and the marker is the range around it. In a real deep dive both come from your data.
Scenarios you could then run

Ask what-if, on your own numbers.

  • What if first-response time halves?
  • What if every new account completes onboarding?
  • What if renewal prices rise on one segment only?
Try a what-if playground →
More investigations

Is this your question?

Tell us the decision you are trying to make. We will say whether a deep dive is the right way to answer it, and what it would take.

Build your own brief