9 Lifetime Value Model — Methodology
9.1 How Much Is a Customer Worth Over Time?
To answer this question, the analysis uses three different models. Each takes a different approach to the same underlying data, producing a range from pessimistic to optimistic. Using all three together gives a more honest picture than any single number.
The chain model — the base case. This is the primary model used throughout the report. It takes a typical order value, multiplies it by how often a customer buys, and projects that forward using observed return rates. Every input comes directly from the transaction data. UK three-year LTV: £1.24K per customer. International: £1.59K.
The steady-state model — the optimistic ceiling. This assumes the business eventually reaches a stable long-run retention rate — a reasonable assumption for a mature business, but one that cannot be confirmed from 13 months of data. It produces the highest LTV figures and should be treated as an upper bound, not a planning number.
The probability-weighted model — the pessimistic floor. This weights each possible future order by the probability that the customer will actually reach it. It accounts for the fact that many customers drop off after their first or second order, producing a lower expected value. UK probability-weighted three-year LTV: roughly £355. International: roughly £574.
Anonymous customers are excluded from all three models — it is impossible to link multiple orders to a single customer when there is no identity attached.
| Segment | Customer Count | Median AOV | Median Reorder (days) | 1→2 Retention | Steady Retention | LTV — 1 Year | LTV — 2 Years | LTV — 3 Years |
|---|---|---|---|---|---|---|---|---|
| UK | 3,914 | £299 | 52 | 65.46% | 74.04% | £597 | £988 | £1.24K |
| International | 414 | £391 | 52 | 63.29% | 69.29% | £783 | £1.28K | £1.59K |
The three models value the same customer base under different retention assumptions. The chain model is the recommended base case for acquisition modeling. The probability-weighted model sets the floor — reality coming in below it would signal a fundamental change in the business. The steady-state model represents the ceiling, achievable only if retention stabilizes at long-run equilibrium.
How sensitive are these numbers? Every projection rests on four uncertain inputs: observed retention rate, median order value, the assumed program improvement percentage, and the observed reorder interval. None is a precise number. A five-percentage-point change in UK retention shifts the three-year per-customer LTV by a meaningful amount. Treat all figures as directionally reliable within a plus-or-minus 10% band. If a decision changes at the edge of that band, the decision needs more data.
| Model | Description | UK 3-Year LTV | International 3-Year LTV | When to use |
|---|---|---|---|---|
| Probability-weighted | Expected value weighted by probability of reaching each order | ~£355 | ~£574 | Conservative floor — accounts for non-return probability at each step |
| Chain model (conservative base) | Sequential retention applied to median AOV and reorder interval | £1.24K | £1.59K | Recommended for acquisition modeling — uses observed retention chain |
| Steady-state (ceiling) | Long-run retention rate applied at steady state | ~£2,990 | ~£3,680 | Ceiling only — assumes business reaches stable long-run retention |
9.2 Model Inputs and Limitations
All scenarios derived from this model use the chain-model 3-year portfolio as the base. Anonymous segment revenue (£1.51M annually) is excluded from forward modeling — it cannot be reliably projected without CustomerID resolution. Year 1 revenue is approximated from the current annual gross figure adjusted for retention dynamics. These are transaction-data-only projections — they do not incorporate cost structure, margin, working capital, or supplier dynamics. Primary-source verification of audited financials is required before any of these figures can be used in a formal valuation model.
Sensitivity to retention. Each scenario figure rests on four uncertain inputs: observed retention, observed median order value, an assumed program improvement percentage, and the observed reorder interval. None is a precise number. A five-percentage-point change in UK 1-to-2 retention shifts the three-year per-customer LTV by a meaningful amount. Treat all figures as directionally reliable within a plus-or-minus 10% band.
The four uncertain inputs:
- 1→2 retention is the largest lever. The day-30 follow-up program is designed to move this input — so the uncertainty is partially under the business’s control. A 5pp improvement moves the model toward the optimistic estimate.
- Median AOV is stable across the analysis period. The risk is downward: exchange rate movements can reduce international purchasing power without any change in list price. Pessimistic scenarios should account for a 10% AOV decline.
- Reorder interval is directly observable from order dates and carries the least measurement uncertainty. The risk is upward: if international logistics slow down, the interval widens and fewer orders fit per year.
- Program improvement assumptions (10pp retention improvement, 50% recovery, 10% reactivation) are assumptions, not observations. Each program should be measured against its assumption within 90 days of launch and the model updated with actual performance.
Lead with the base estimate. If the pessimistic portfolio value still exceeds program implementation cost, the investment is justified even under conservative assumptions.
Author: Shawn Phillips | Lailara LLC
← Wallet Share & H1 vs H2 Momentum | Lifetime Value & Forward Portfolio →