Outgrace
OutgraceThe Leadership Magazine, home
Growth & MarketingHow-to guide

How to build a customer payback model finance will sign off

Separate acquisition cost, contribution margin and churn, measure each by cohort, and set spend limits from the result. None of the public sources checked defines these terms.

A close-up of a hand holding a sign stating Where customers matter against a bright yellow background.
Photo: RDNE Stock project / Pexels

Finance rarely rejects a growth budget because the marketing figures are wrong. It rejects them because three different things have been fused into one number: what a customer costs to win, what that customer keeps after the cost of serving them, and how long they stay. Separate those three and the argument becomes far shorter. This guide shows how to build a payback model from them, and how to turn the result into a spend limit finance will accept.

Before you start

Decide what the model is for. A payback model can answer three different questions: whether a channel is worth funding, how much a channel may spend, and whether the business can afford faster growth. Choose one for the first version. Models built to answer all three at once tend to answer none of them well, and a finance reviewer will spot the mismatch within minutes.

Agree the definitions before you open a spreadsheet. None of the four outlets we checked defines these terms, so the definitions below are our working convention, not a published standard. Use them as a starting point, check them against your own ledger, and confirm with your auditor if your reporting depends on how acquisition spend is classified.

  • Acquisition cost: the sales and marketing spend attributable to winning a group of new customers, divided by the number of customers won.
  • Contribution margin per customer: revenue from a customer less the costs that rise with serving that customer, such as product costs, payment fees and fulfilment. Fixed overheads stay out.
  • Churn: the share of a cohort that stops paying in a given period. Track it per cohort, not as one blended average.
  • Payback period: the number of periods a cohort needs for cumulative contribution to cover its acquisition cost.

Two choices will move the answer more than any others: whether salaries are counted in acquisition cost, and whether revenue from customers who are about to leave is counted in the months before they go. Agree both with finance in writing. A model that changes its own rules halfway through a review is dead on arrival, however sound its arithmetic.

  1. Step 1

    Define the cohort

    Group customers by the month or quarter they started paying, and by channel where channels differ in cost. The cohort is the unit the model reasons about. What usually goes wrong: teams group by the date a deal closed instead, which moves revenue into the wrong period.

  2. Step 2

    Load the acquisition spend

    Take the sales and marketing cost for each period and allocate it to the cohort it produced. Be explicit about lag, because spend in one month may win customers in the next. What usually goes wrong: all spend is charged to whichever customers happened to close that month, so a campaign that ran for a quarter looks either brilliant or ruinous.

  3. Step 3

    Calculate contribution per customer

    For each cohort, take average revenue per customer per period and subtract only the costs that rise with each extra customer. Keep fixed platform and staff costs in the overhead schedule. What usually goes wrong: gross margin from the management accounts is used without checking whether it contains costs that do not vary with volume.

  4. Step 4

    Model churn as a curve

    Measure how many customers from each cohort are still paying after one, two, three and more periods, then plot the survival curve. A single average churn rate hides the early drop-off that usually decides payback. What usually goes wrong: a blended rate flatters a business whose newest customers leave fastest.

  5. Step 5

    Calculate payback from the curve

    Multiply contribution per customer by the share still paying in each period, add the results cumulatively, and find the period in which the total passes acquisition cost. Run this on each cohort's actual history first, then on forecast cohorts. What usually goes wrong: the model assumes every customer stays, so payback is quoted as if churn did not exist.

  6. Step 6

    Set the spend limit

    Turn the payback target into a maximum acquisition cost per customer, then into a maximum monthly budget for each channel. Work backwards from the target period, the contribution it implies, the survival curve, and the acquisition cost those support. What usually goes wrong: the limit is set from the best-performing channel and applied to all of them.

  7. Step 7

    Write the stop rule

    Agree in advance what result would cut spend: for example, a cohort that fails to pay back within the agreed period, or a churn curve that falls outside the band the model assumed. Fix a review date. What usually goes wrong: nobody decides beforehand, so the budget is extended by default.

Mistakes that cost the most

The most expensive mistake is treating a payback period as a promise. It is a forecast built on the survival curve, and when churn rises the forecast moves with it. A randomised field experiment on innovation-driven entrepreneurs, published as NBER Working Paper 34755 by Esther Bailey, Daniel Fehder, Eric Floyd, Yael Hochberg and Daniel J. Lee, found that training may help founders recognise when a venture no longer merits further investment of time and resources. The same paper says effective training could cause some startups to shut down sooner while improving outcomes for others. The study concerns founders, not marketing budgets; the parallel for spend is ours. A stop rule that is never used protects nobody.

Other errors are more mundane. Teams count revenue from customers who have not yet been billed, leave out the cost of onboarding and support, or charge all spend to the newest channel. Each one makes payback look shorter than it is. Finance reviewers look for three things: whether the cohorts match the ledger, whether the churn assumption comes from actual history, and whether the spend limit follows from the payback target. A model that passes all three rarely needs a long argument. Answer those three questions in the first draft.

Should the sales team's salaries go into acquisition cost?

Usually yes, if the team's work is winning new customers, and the treatment must be applied consistently. Split out any time spent on renewals, because renewal work is not acquisition. Agree the split with finance and keep it fixed from one quarter to the next. A change in method will look like a change in performance even when the market has not moved.

What if we have too few customers for cohorts to be meaningful?

Use longer periods, such as quarters rather than months, and say so in the model. Treat any payback figure built on a small number of customers as provisional. Mark those cohorts rather than dropping them, because dropping them quietly is the easiest way to lose a finance director's trust.

Which inputs must come from outside this guide?

Your own ledger for spend, revenue and cost of service, and your own customer records for start and end dates. The four public sources we checked do not supply these terms. According to the U.S. Census Bureau's retail trade main page, its Quarterly E-Commerce Report carries no margin, acquisition-cost or churn figures. Benchmarks drawn from a company's own filings or a finance text are more useful than broad public averages.

Can a payback model justify any spend?

No. It tells you whether a given spend pays back within the period you agreed, and nothing more. It does not show whether the market is large enough or whether the product will keep improving. Use it to set limits and to test channels, not as the sole argument for growth.

How often should the spend limits be reviewed?

Review them on the date written into the stop rule, and again whenever the churn curve moves outside its agreed band. Quarterly suits most budget cycles, but the fixed date matters more than the interval. A review date that exists before spending begins is worth more than a perfect model that nobody revisits.

Pick one channel and one cohort from the last twelve months. Put the four definitions on a single sheet, load the spend and revenue, and check whether the payback you get matches what your team has been quoting. Send the result to your finance lead before the next budget round, with the stop rule written underneath it. If the numbers disagree with the quoted figures, that disagreement is the most useful thing the model will produce this year.

Newsletter

The Briefing

One email a week: the leaders we profiled, the ideas worth stealing, and nothing else.

Weekly, free, one click to leave. Read our privacy policy.