Savings Habit Decay & AutoPay Churn

Overview

Eight micro-savers. Four shocks. One counterintuitive finding.

Eight synthetic personas spanning ₹9,500 to ₹48,000 monthly income were exposed to four shock events designed to stress-test AutoPay retention across a realistic range of financial and social triggers. The lowest-income persona — a domestic worker on ₹12,000/mo — was the most churn-resistant. The highest-income persona — an IT manager on ₹48,000/mo — churned on day 67 after a ₹505 paper loss.

The difference was not financial capacity. It was the nature of the saving identity each persona had built. Meena Kumari, the ASHA worker on ₹9,500/mo, saved through a 52-day stipend delay by borrowing from a self-help group to maintain her deposit. Vikram Nair, earning five times more, disengaged the moment his performance identity was disrupted.

The study found six of eight churns were friction-driven, not intentional. Mandate expiry (S3) produced the widest single-shock impact — churn risk of 4/10 or higher for six of eight personas — most of whom had no intention to stop saving. The primary product failure was not in motivation. It was in the renewal flow.

Key Metrics

Personas

Shock events

Conversations

+41

pts · Best delta

Key Findings

Mandate expiry (S3) produced churn risk ≥4/10 for 6 of 8 personas — the widest single-shock impact
Income was the primary churn predictor for 0 of 8 personas
6 of 8 churns were friction-driven, not intentional — users wanted to continue
Flexible Pause (I5) was the only intervention with zero backfire cases across all segments
3 interventions actively accelerated churn for specific persona types

Recommendations

Four interventions. Ranked by evidence strength.

Immediate

Deploy Flexible Pause as a proactive default

Flexible Pause (I5) was the only intervention with zero backfire cases across all eight persona types. For friction-driven churners — Sunita, Dinesh, Meena — it reduced churn risk by an average of 5.2 points. It works because it preserves the saving identity while removing the compliance barrier.

Target segments: All — no persona type was harmed by its presence

Immediate

Instrument friction-churn vs. intent-churn separately

Six of eight churns in this study were friction-driven — the user intended to continue but could not complete the renewal or recovery flow. These are currently indistinguishable from intentional churn in server-side data. Separate instrumentation unlocks dramatically different intervention logic for each type.

Target segments: All — affects product analytics architecture, not user-facing

Short-term

Build persona-level intervention routing

Three interventions — Gamification (I4), Social Proof (I3) — actively accelerated churn for specific persona types (Pradeep, Vikram). The same message that retains a streak-driven saver (Arjun) damages trust for a SEBI-literate optimizer (Pradeep). Segment-level routing is necessary before any retention campaign can avoid net harm.

High-risk: Pradeep-type (accountants, financially literate); Vikram-type (performance-identity savers)

Short-term

Front-load identity-anchoring at onboarding

The highest-leverage finding in the study: Vikram Nair's Day 67 churn was architecturally predicted by his Day 1 onboarding. He had no named goal, no specific savings horizon, and no non-performance reason to save. Named goals at onboarding — a 10-second step — would have changed the entire behavioural architecture that led to churn.

Target segments: Performance-identity savers (Vikram-type); applies broadly at onboarding

Open Questions

What this study cannot answer.

01 What is the actual distribution of goal-specificity scores in live onboarding data? If a large share of current users have no named goal (like Vikram), the onboarding intervention has a much larger addressable surface than any retention campaign.
02 Can friction-churn and intent-churn be distinguished from server-side data alone? The simulation suggests they are behaviourally very different — but the question is whether server logs contain enough signal to classify them without a new instrumentation layer.
03 What is the 5–10 day post-churn re-engagement rate if a targeted message is sent within 24 hours of lapse? The simulation identified several personas (Sunita, Dinesh, Meena) who had strong residual intention to re-enroll — the window may be narrow but it exists.
04 How does peer network composition affect scam narrative spread velocity? Dinesh's entire trust architecture collapsed through a single WhatsApp forward from one peer. The platform has no relationship with that peer and no mechanism to intercept it.

Savings Habit Decay & AutoPay Churn

Eight micro-savers. Four shocks. One counterintuitive finding.

Eight personas. ₹9,500 to ₹48,000/mo. Widely different retention profiles.

Four shock events. Eight personas. What actually drives AutoPay churn.

Four interventions. Ranked by evidence strength.

What this study cannot answer.

Run your own simulation.