Industries, Fintech & neobanks
Decisions that respect your CFO and your compliance team.
Subscription-tier upgrades, retention nudges, and pricing personalisation, with audit-grade causal proof.
Neobanks and investing apps monetise through tiered subscriptions. They obsess over D30 retention and upgrade rates; they distrust their A/B tests because conversion is noisy and attribution is sparse post-ATT. A five-percent lift in subscription retention compounds into eight-figure annual revenue. Doubly-robust off-policy evaluation is the credibility layer the CFO demands.
What agents handle
Workflow
Tier-upgrade prompt
Before
Same upsell modal to every free-tier user
After
CATE-targeted: shown only to users with positive predicted uplift
Result
+22% upgrade rate on targeted cohort
Workflow
Dormant-account nudge
Before
Email blast to all users inactive 14+ days
After
Bandit picks channel + copy + timing per user
Result
+8% reactivation, half the send volume
Workflow
Churn-save offer
Before
Fixed 50% discount offered at cancel flow
After
Propensity-targeted discount with full DR readout
Result
Revenue per save up 31% on the same offer budget
Workflow
Feature-discovery surfacing
Before
Generic onboarding tour, same to all
After
Slate-ranking bandit per user profile and device
Result
+14% feature adoption D30
The challenges we solve
- Regulatory caution on personalisation, every change needs an audit trail
- CFO scepticism of growth-team A/B claims; the boardroom expects holdout-validated numbers
- High cost of false-positive UI changes, a bad paywall change is a quarterly miss
- Subscription cohorts churn and renew on monthly clocks that confound naive A/B tests
- Per-environment isolation matters: test keys must never touch live data
Append-only decision logs
Every assignment is immutable. The propensity logged at serve time is the propensity OPE consumes. No revisionism, compliance review can replay the full decision history.
Per-tenant isolation
Admin and org roles enforced at the dependency layer, not the query layer. Live and test API keys partition cleanly. Cross-tenant access is structurally impossible.
Revocable API keys per environment
Scoped (decide-only / decide+reward / full) and instantly revocable. SHA-256 at rest. Plaintext shown once at issuance.
Worked audit example
What a doubly-robust re-analysis surfaced for Lyra.
Series-B neobank · ~$18M ARR · MENA + EU · 482,711 free-tier accounts · case ID LYRA-2026Q1-AUDIT-002
What their team reported
Premium upgrade prompt timing test: week-2 (+4.4%) vs week-4 (4.1%) on the upgrade rate. p = 0.02. Team called it "winner, ship week-2" and rolled to 100% of free-tier users.
What our re-analysis found
Doubly-robust per-cohort: +11.7% lift on balance-high × spend-regular (the engine), but −8.2% on balance-low × spend-irregular, a 14% slice being prompted into downgrade/cancel. The t-test averaged it away.
Recommendation · projected annualised impact
Personalise the prompt: send to the positive-uplift cohorts (~58% of base), suppress for the negative cell. Lifts net upgrade rate from +6.4% to +11.7% and preserves retention on the irregular-spend cohort.
+$1.4M / yr ARRSame shape we'll send back on your last A/B test, in three business days.
Read the full audit PDF →Audit your last paywall test — free.
One CSV, one experiment config. Same-shape readout back in three business days.