LTV:CAC for subscription products: modelling payback before you scale spend
Every subscription business eventually has the same argument. Marketing wants more budget because the CAC "looks fine". Finance wants to cut budget because cash keeps leaving faster than it comes back. Both are reading real numbers; neither has the model that connects them. That model — LTV by cohort, CAC by channel, and the payback window between them — is the single document that should govern how much you spend and where, and most companies scale spend for years without building it.
The three numbers, defined honestly
CAC is fully loaded acquisition cost per paying customer — media spend plus creative production plus agency or team cost, divided by first payments, not trials. Segmented by channel and geography, because a blended CAC hides exactly the variation you need to act on.
LTV is the contribution margin a customer generates over their life, not their revenue. Payment fees, refunds, service costs come out first. And it must be cohort-based: the average lifetime of customers acquired three years ago tells you little about the ones you acquired last quarter — especially if channel mix shifted, because customers from different channels retain differently. TikTok cohorts and branded-search cohorts are not the same animal wearing different clothes.
Payback window is the number the other two exist to produce: how many months until a cohort's cumulative contribution margin covers its acquisition cost. This is the number finance actually feels, because it is the number your cash position is made of.
Building the model without a data team
A working version fits in a spreadsheet. Rows: monthly acquisition cohorts per channel. Columns: months since acquisition. Cells: cumulative contribution margin per customer, from your billing data. Against each cohort row, its CAC. The diagonal where cumulative margin crosses CAC is your payback frontier — and the visual alone usually settles the marketing-versus-finance argument in one meeting.
Two refinements earn their complexity early. First, retention curves flatten: model the tail conservatively (we cap projected lifetimes rather than extrapolating early retention forever — survivorship flattery is how LTV models lie). Second, discount nothing before month twelve matters less than getting refunds and involuntary churn in from day one; failed payments quietly eat a meaningful slice of naive LTV.
From model to bids
The model becomes operational when it sets targets. Decide the payback window your cash position can carry — subscription businesses commonly operate somewhere between three and twelve months, and the right answer is a finance decision, not a marketing preference. That window plus cohort margins yields a maximum CAC per channel and geography; those become your tCPA and cost-cap targets. Platforms then optimise inside boundaries your economics chose, instead of the other way round.
This also reframes the LTV:CAC ratio itself. The folk-wisdom "3:1" is a decent sanity check and a poor target: a 5:1 ratio with a two-year payback can still bankrupt a company that funds growth from cash flow, while a disciplined 2.5:1 at four-month payback can compound safely. Ratio measures efficiency; payback measures survivability. Scale on payback.
One warning from the trenches: the model is only as good as the conversion data underneath it. If your platform conversions are duplicated or your trial-versus-paid events are muddled, every cell inherits the error — which is why we treat measurement hygiene as a prerequisite of economics work, not a separate project.
Build the model before the growth push, revisit it monthly, and let it say no sometimes. The cheapest customer acquisition mistake is the one the spreadsheet talks you out of.