An end-to-end analytics narrative for a fictional neobank — from onboarding
funnel to A/B test, cohort retention, a difference-in-differences causal test, segmentation,
and a Market & Jobs layer. Every chart below is rendered from the repository's own
committed CSVs. No install, no server, no JavaScript.
17
analytical projects
funnel → JTBD → causal
170+
tests · 98% coverage
ruff · mypy · CI green
+5.72pp
KYC activation lift
A/B, p<0.0001
+9.2pp
M3 retention lift
post-fix cohorts
+9.2pp
causal DiD ATT (M3)
diff-in-diff, parallel trends ✓
36.0% vs 4.8%
referral conversion
anchor vs 45+ gap
Onboarding funnel
Where do new users drop off between install and first transaction?
KYC completion is the critical step: the largest relative drop in the funnel. Registration loses the most users in absolute terms — two lenses, reported both.
KYC progress-bar A/B test
Does a progress bar in the KYC flow lift completion?
Treatment lifts completion by +5.72pp (p<0.0001, no SRM) — ship. This is the randomized claim the causal layer later strengthens.
Cohort retention
Did the fix hold up in long-term retention?
Post-fix cohorts retain better from M1 onward (+9.2pp at M3). The gap persists across the curve, not just at one month.
Causal layer — difference-in-differences
Was the retention lift caused by the fix, or just a shared time trend?
DiD ATT = +9.2pp: the treated flow rises at the cutoff while the comparison flow does not. Pre-trends flat, placebo null, covariates balanced.
Customer segmentation
Who are the users and where does revenue concentrate?
The top segment is 12% of users but ~41% of revenue — a classic concentration story that drives the per-segment monetization strategy.
Market & Jobs — premium upsell
Does the premium upsell transfer to every job segment?
The anchor segment converts at ~17% while Digital Newcomers 45+ sit near 2% — the value proposition does not transfer. Segment-specific offers, not one upsell.
Market & Jobs — referral conversion
Does referral scale to new segments?
Referral converts best in the anchor and collapses for 45+ and family budgeters — don't scale referral spend before segment-specific incentives.
Market & Jobs — segment sizes
How big is each job segment in the simulated base?
Five JTBD segments simulated at scale — the population the Market & Jobs layer reasons about.