Nikita Boyarkin
← Volta Neobank

Extended · project 7 of 23

Three lifetime-value methods: historical, retention-curve and Gamma-Gamma. The order Power > Growth > Casual > Dormant is robust across all methods.

Situation
To set CAC ceilings and retention priorities we need an estimate of future value, not just past revenue
Task
Own a forward-looking value estimate rather than a backward-looking revenue total to set CAC ceilings and retention priorities
Action
Historical CLV from actuals, a retention curve with a power fit, and probabilistic Gamma-Gamma
Result
Power > Growth > Casual > Dormant holds across all three methods; Gamma-Gamma Power €5,166 vs Dormant €27.7 — a ~187× gap
Stack
Pythonpandas / NumPySciPy / Statsmodelsscikit-learnMatplotlib / Seabornuv + ruff
Source View on GitHub →
On this page
  1. Situation
  2. Task
  3. Actions
  4. Result
  5. Recommendations
  6. Documentation

Volta — CLV Modeling

Situation

To set CAC ceilings and retention priorities we need an estimate of future value, not just past revenue.

Task

I needed a forward-looking value estimate rather than a backward-looking revenue total, so I owned the basis for setting CAC ceilings and retention priorities.

Actions

  • Historical CLV (actuals), retention-curve (power-fit), probabilistic Gamma-Gamma.

Result

  • The order Power > Growth > Casual > Dormant is robust across all three methods.
  • Gamma-Gamma: Power €5,166 vs Dormant €27.7 — a ~187× gap.
  • Predictive methods run 2.9–5.9× above historical.

Recommendations

  • Use predictive CLV as the per-segment CAC ceiling.
  • Prioritize retention of Power/Growth (largest value at risk).
  • Don’t rely on historical CLV for forward-looking decisions.

Documentation

Charts

Source: github.com/NikitaBoyarkin/volta-banking — 22 projects; figures recomputed from the repo's own datasets (data/*.csv) via its analysis scripts. Funnel counts from data/volta_funnel_data.csv (10,000 users); A/B, retention, segmentation, churn, RFM, CLV, attribution, anomalies, spend, support, NPS, JTBD, unit economics, premium, KYC deep-dive, referral, assisted CAC, FX sourcing, premium offers, anchor CAC and dormant win-back follow the published project narrative (README + part pages).

CLV by three methods

Customer lifetime value by segment, estimated three ways: historical, retention-curve and Gamma-Gamma. The order Power > Growth > Casual > Dormant is robust.

0 2000 4000 6000 Power — Historical: 1678 1678 Growth — Historical: 238 238 Casual — Historical: 37 37 Dormant — Historical: 4.7 4.7 Power — Retention-curve: 5069 5069 Growth — Retention-curve: 644 644 Casual — Retention-curve: 107 107 Dormant — Retention-curve: 13.8 13.8 Power — Gamma-Gamma: 5166 5166 Growth — Gamma-Gamma: 694 694 Casual — Gamma-Gamma: 133 133 Dormant — Gamma-Gamma: 27.7 27.7 Power Growth Casual Dormant Segment CLV, € Historical Retention-curve Gamma-Gamma
Key takeaways
  • Power leads across all three methods: Gamma-Gamma €5,166 vs Dormant €27.7 — a ~187× gap.
  • Predictive methods run 2.9–5.9× above historical — the historical method understates future value.

Volta project map

Volta overview →

Core

Extended

Market & Jobs

RAT v2

Causal