StudiesFintech · a consumer savings product

The customer most able to pay was the first to leave.

Modelled case · not a client resultSimulated customers, modelled on real people. Results are directional.
₹9,500A month: the lowest-income saver, and the most churn-resistant

Modelled case · a consumer savings product

8 simulated savers, one dot each.

8Simulated savers, on ₹9,500 to ₹48,000 a month
IndiaMarket
4 shocksMethod: each saver put through four shock events

Modelled case: a consumer savings product (anonymised). Not a client result.

What was real · what was modelled

What was real · what was modelled

Real inputs

  • The four real shock events (paper loss, mandate expiry, stipend delay, social trigger)
  • The income and savings profile of each buyer segment
  • The autopay mandate mechanics
  • The retention question the product team faced

Modelled outputs

  • Each saver’s churn risk under each shock
  • The day-level timing of disengagement
  • The share of churn that was friction- vs intent-driven
  • Which shock produced the widest impact

The question

Everyone assumed the customers most likely to drop out were the ones who could least afford to save.

A savings product wanted to know which of its autopay savers were about to lapse, so it could intervene. The going-in belief was the obvious one: lower income means thinner margins means higher churn risk. Eight simulated savers, spanning ₹9,500 to ₹48,000 a month, were put through four shock events (a small paper loss, a mandate expiry, a stipend delay, a social trigger) to see who broke and when.

Simulated customers

Simulated customers

Eight savers, each with a different reason the habit mattered.

The personas weren’t income brackets. Each carried a distinct saving identity (a survival discipline, a performance metric, a household duty), and that identity, not the balance, decided how each one absorbed a shock.

Meena Kumari
ASHA worker · ₹9,500 a month · most churn-resistant
Saved straight through a 52-day stipend delay by borrowing from her self-help group to keep the deposit going.
Vikram Nair
IT manager · ₹48,000 a month · churned on day 67
Five times the income. Disengaged the moment a ₹505 paper loss disrupted his performance identity.

The finding

Income didn’t predict churn. Saving identity did, and most churn was accidental.

₹9,500A month: the lowest-income saver was the most churn-resistant
Day 67The highest earner churned first, after a ₹505 paper loss
6 / 8Lapses were friction-driven, not intentional
Run your version →

Why it happened

The mandate expired before the intention to save did.

The product treated churn as a decision: someone choosing to stop. The simulation found the opposite: six of eight lapses were friction, not intent. A mandate quietly expired, or a small loss bruised a saver whose whole relationship with the product was about feeling successful, and the habit broke before anyone decided anything. Meena’s identity was survival discipline, so a shock made her save harder; Vikram’s was performance, so a trivial paper loss read as failure and he walked. In this simulation, the lever wasn’t a richer incentive for low-income savers; it was removing the friction that lets willing savers fall out by accident.

The rest of the explanation

A mandate quietly expired, or a small loss bruised a saver whose whole relationship with the product was about feeling successful, and the habit broke before anyone decided anything. Meena’s identity was survival discipline, so a shock made her save harder; Vikram’s was performance, so a trivial paper loss read as failure and he walked. In this simulation, the lever wasn’t a richer incentive for low-income savers; it was removing the friction that lets willing savers fall out by accident.

The decision it pointed to

Stop targeting retention spend by income tier. Re-sequence the mandate-renewal flow so expiry can’t silently drop a willing saver, and treat identity-disruption moments (a visible loss for a performance-motivated saver) as the real intervention trigger, not the low balance the dashboard flags.

Modelled recommendation

Confidence & limits

These are modelled outcomes, not measured ones. Every score and probability here is a simulated-population prediction, produced before any real-market test. The panel is eight savers across four shocks, built to isolate the identity mechanism, not to size a book. Treat the numbers as directional: the direction and size of the effect is the finding, not a forecast. Check the top ideas with your own customers before a decision commits to them. Simulatte can run those interviews and show where they agree with the model.

“The lever isn’t motivation. It’s friction — most of the people who left still wanted to save.”
From the study’s summary
All studies →

Run your version of this decision.

Bring the call you’re weighing. First answers from simulated customers in minutes, then check them with your own customers.