Amey-Thakur SaaS metrics
SaaS finance is a small set of numbers that everyone miscomputes slightly differently. Define each once, compute from the billing source of truth, and read them in cohorts: aggregates smooth over exactly the problems you need to see.
Method
- Build MRR from components. New + expansion + reactivation - contraction - churn = net new MRR, each component tracked separately (the mix is the diagnosis: flat net-new from booming new + bleeding churn is a different company than steady low both): normalized to monthly (annual contracts / 12), excluding one-time fees and unpaid trials (see warehouse-modeling's definition-once rule; this is its revenue instance).
- Compute churn in both currencies, by cohort. Logo churn (customers lost / customers at start) and revenue churn (MRR lost / MRR at start) diverge when customer sizes vary: small-customer churn with large-customer retention can be healthy; the reverse is a fire. Monthly cohort curves (see churn-analysis for the diagnosis toolkit) beat blended rates: blended churn improves mechanically as you grow even when every cohort is worsening.
- Make NRR the headline retention number. Net revenue retention (a cohort's MRR now / its MRR a year ago, expansions included, new customers excluded): above 100% means the installed base grows itself (usage-based pricing and expansion tiers drive it: see saas-pricing's value metric). Sustained NRR is the single strongest signal of product-market durability; report it beside gross retention so expansion is not hiding churn.
- Compute CAC fully loaded, by channel. Sales + marketing cost (salaries included) / customers acquired, per channel and segment: blended CAC hides that paid-channel CAC is 5x organic's; payback period (CAC / monthly gross margin per customer) is the cash-reality version: 12-18 months is conventional health for SMB SaaS, longer tolerable with strong NRR (see cloud-cost-optimization's unit-economics ethic: same discipline, different cost object).
- Treat LTV as an assumption-laden estimate. Gross margin per customer x expected lifetime (1/churn is the naive version and explodes at low churn: cap horizons, use cohort-observed survival instead: see scikit-survival adjacency for the honest math); LTV:CAC of ~3 is folklore-standard but the inputs deserve more scrutiny than the ratio. Never let a projected LTV justify runaway spend that cash payback contradicts.
- Review the panel monthly, decisions attached. MRR components, NRR/GRR, cohort curves, CAC payback by channel, burn multiple (net burn / net new ARR): trended, against targets, with an owner and an action per red number (see product-metrics' review ethic; status-updates' no-surprises for the board version). Benchmarks (public SaaS medians) calibrate ambition, but your own cohort trends decide what to fix.
Boundaries
- Billing-system truth and analytics-event truth drift; revenue metrics compute from billing (see data-quality-checks reconciliation between them), and finance signs the definitions.
- These metrics assume recurring-revenue mechanics; marketplaces, usage-spiky infrastructure, and services businesses need adapted definitions, not forced fits.
- Metrics describe; they do not decide. A great dashboard over a product nobody loves is instrumentation on a sinking ship (see product-discovery).