Calculating Customer Lifetime Value

Choose the economic question

Lifetime value can support acquisition bids, service tiers, retention programs, or valuation. Each use may require a different scope and precision. Define the customer entity, start event, horizon, and whether the output is historical realized value or forecast expected value. Do not compare an expected forecast for one segment with realized revenue for another.

Decide whether acquisition cost is outside the value measure or subtracted within it. A common operating convention calculates customer contribution before acquisition, then compares that value with customer acquisition cost. Label the convention. Include refunds, discounts, payment fees, servicing, fulfillment, and other variable costs that materially follow customer activity.

Build cash flow from cohorts

For each period after acquisition, estimate the probability a customer remains active and the expected contribution while active. Multiply them to obtain expected period contribution. Cohort tables can supply retention and spend patterns. Use mature groups where possible, and separate segments with structurally different contracts, margins, or renewal cycles.

For noncontractual purchases, define inactivity carefully because absence of a recent order does not prove permanent churn. A finite historical window may produce a practical value estimate without claiming a precise customer death date. Contractual businesses can use explicit cancellation and renewal events, while still accounting for pauses and reactivation.

Treat the future honestly

Choose a finite forecast horizon supported by business history or fit an appropriate survival model with diagnostics. The shortcut value equals margin divided by churn relies on stable, memoryless churn and an infinite horizon, assumptions that often fail. It also becomes extremely sensitive when estimated churn is small or rounded.

Discount future contribution when timing matters, using a rate approved for the decision. Explain whether values are nominal or real and how inflation or price changes enter. Do not insert optimistic expansion without evidence. Base, downside, and upside scenarios are often more decision-useful than one precise-looking number.

Validate before allocating spend

Backtest forecasts against older cohorts by pretending the model was built at an earlier date. Compare predicted cumulative contribution with what later occurred. Inspect calibration by segment and acquisition source. A model that gets the average right while overvaluing one channel can still cause harmful budget choices.

Report value alongside acquisition cost, payback time, uncertainty, and cohort size. Consider selection effects: customers acquired through a promotion may have lower retention and margin than organic customers. Refresh the model when pricing, costs, product, or channel mix changes, and preserve prior versions so changed forecasts can be explained.

Estimate value for a meal subscription cohort

A meal service wants to raise advertising bids using the first eight weeks of customer data, but fulfillment costs and discounts vary over time.

1. Define the customer as a household subscription and exclude acquisition spend from contribution while retaining discounts, refunds, food, packaging, delivery, and payment costs.
2. Calculate weekly retained probability and contribution per active household for comparable historical cohorts.
3. Forecast through week fifty-two using a base, downside, and upside retention path instead of extending week-eight behavior indefinitely.
4. Discount expected weekly contribution and compare the resulting range with channel acquisition cost and payback time.
5. Backtest the method on last year's cohorts and reduce bids for channels where predicted value was consistently overstated.

Lifetime value model brief

Document these inputs beside every reported customer value estimate.

• Decision and entity: use case, customer definition, acquisition event, segment, and forecast date.
• Economics: revenue, discounts, refunds, variable cost categories, contribution convention, and acquisition-cost treatment.
• Behavior: cohort source, retention or purchase model, reactivation rule, spend path, and observation maturity.
• Future assumptions: horizon, discount rate, price or cost changes, scenarios, and terminal treatment.
• Validation: backtest periods, calibration by segment, uncertainty range, owner, and refresh trigger.

Common mistakes

• Calling gross revenue lifetime value while ignoring fulfillment, support, refunds, and payment costs.
• Dividing margin by one recent churn rate without stating the infinite-horizon assumptions behind the shortcut.
• Using a blended value to justify acquisition spend for a channel whose customers have different economics.

Try one

Sources

• NIST process and production control handbookNIST statistical methods for monitoring processes, time series, and change.
• NIST exploratory data analysis handbookNIST methods for exploring data, checking assumptions, and revealing structure.