Customer lifetime value for subscription businesses

Three corrections make the formula usable:

1. Use margin. Revenue overstates value by what it costs to serve the customer: at an 80% margin, a €29 plan brings in €23.20 a month.
2. Measure churn by tenure. If new customers leave faster than those who have stayed a year, one average gives both the same value, which is right for neither. A retention curve for each month’s sign-ups, called a cohort, shows whether they do.
3. Set a horizon. The formula counts every future month at full value; a value over three or five years, discounted, is the figure to set against what a customer costs to win.

Annual plans renew once a year, so their churn is a renewal rate. With 75% renewing, the average annual customer stays four years.

Each tier then gets its own question: which top-tier customers are at risk of leaving, which middle-tier customers are likely to upgrade, and what the bottom tier costs to serve. The tiers are checked month to month, because a segmentation that reshuffles every month cannot be planned against.

Where customers buy again rather than subscribe, the same ranking runs on how recently they bought, how often and how much: recency, frequency and monetary value, often shortened to RFM.

How long building the tiers takes, step by step, is in How long does it take to build a customer segmentation model?

A next-best-offer model puts several propensity models together. For each customer and each offer, it multiplies the chance of taking the offer by what the offer is worth if taken, less what it costs to make. The offer with the highest expected value wins, within the contact rules you set, such as no more than one offer a month.

Each propensity model needs enough past takers, as a churn model needs enough past leavers; how many is on the churn prediction page.

Who it is for. Subscription and repeat-purchase businesses with no data team, deciding how much to spend to win or keep a customer, and what to offer whom.

The questions it answers. What a customer is worth, by plan, channel and tenure. Which customers make up most of the value. Who is likely to upgrade or buy more. Which offer to make each customer next.

What you need to have. Billing history per customer, with plan and price changes; start and end dates; a margin per plan or product, even a rough one; and, for offers, a record of past offers and who took them.

The method, in plain words. Retention curves by sign-up month, plan and channel; a value per customer over an agreed horizon; tiers that are checked for stability; then propensity models for each offer, tested on months they never saw, and the expected-value rule for choosing offers.

What you receive. A decision memo of up to two pages on what customers are worth and where the value sits. A value and a tier for every customer, as a file or a recurring report. For offers, a ranked offer list per customer. A method note listing the checks that ran.

How it is checked. Agents build the models and draft the memo. Every deliverable gets number and claim checks run by scripts, not AI, and a review by an agent. On what you act on, such as a model going live or a decision memo, that review is a panel of three reviewer agents and a judge, and then the analytics director signs it; routine refreshes do not need a sign-off. The checks are described on how it works.

Your data. Where it goes, which model providers process it and what happens when the work ends: see data handling.

Price. Fixed per deliverable. A customer-value model or an offer model is a sprint, from €12k, fixed in a one-page proposal before work starts. Repeat work runs on an async retainer, from €8k a month.

Who builds it. FinDaS AI’s analytics director built propensity, churn and next-best-offer models during eight years in retail-banking and consumer-finance CRM analytics. Customer value is one of the money questions the same team works on.

A subscription software company sells three plans, all at an 80% margin.

A Scale customer pays ten times what a Starter customer pays and is worth over thirty times as much, because they also stay more than three times as long. Worked on revenue instead of margin, Starter would read €580 (€29 ÷ 5%), overstating it by a quarter.

Now the next offer for a Starter customer. The company has two to choose from:

Each move is worth the gap between the two plans’ values if taken. Fewer customers take the move to Scale, and it still wins. It rests on customers who move up then behaving like that plan’s own customers, an assumption the work tests on past movers before anyone acts on it. How a lasting cut in churn changes revenue, and how sure that is, is shown in our revenue forecast demo, on the same company.