SaaS Metrics That Actually Predict Anything

Every SaaS company measures ARR, churn, CAC, and LTV. Very few can tell you which of those, at their stage, is load-bearing. The distinction matters because a metric you cannot act on is a report, not a control.

Net revenue retention is the one

If you keep a single number, keep NRR: revenue from the cohort you had a year ago, measured today, including expansion, contraction, and churn.

NRR = (starting ARR + expansion − contraction − churn)
      ÷ starting ARR

It works because it compounds. Above 100% the existing base grows without new logos, which means growth continues even if acquisition stalls — the single most valuable property a subscription business can have. Below 100%, sales is running to stand still, and every efficiency gain gets eaten by the leak.

Gross retention tells you whether the product is needed. Net retention tells you whether the business is a compounding one.

CAC payback beats LTV/CAC

LTV/CAC is a ratio built on an estimate of customer lifetime — a number that, for a young company, is largely invented. CAC payback asks a question you can answer with data you already have: how many months of gross profit does it take to recover the cost of acquiring a customer?

CAC payback = CAC ÷ (new MRR × gross margin)

It maps directly to cash. A twelve-month payback with eighteen months of runway is a different company from a twelve-month payback with three years of runway, and the ratio version hides that.

The magic number, honestly

Sales efficiency — net new ARR divided by prior-period sales and marketing spend — is useful precisely because it is crude. It answers "if we spend another pound here, what comes back?" Anything meaningfully above 0.75 usually justifies more spend; well below, and the correct move is fixing conversion or retention before adding headcount.

What to stop reporting

Logo churn without revenue weighting. Losing thirty small accounts and one large one are not the same event.
Blended CAC. Mixing paid and organic acquisition produces a number that describes no channel and cannot be optimised.
ARR without a definition. Committed, billed, and annualised-last-month are three different things, and the gap is where surprises live.
Pipeline coverage as a health metric. It measures optimism as much as demand unless stage definitions are enforced.

The FP&A version

A forecast is a set of assumptions with arithmetic attached. The useful discipline is to state each assumption as a driver — sales headcount, ramp time, quota attainment, win rate, average contract value — and to review last quarter's assumptions against outcomes before setting next quarter's. A model that is never scored against reality becomes a wish with a spreadsheet around it.

General business commentary, not financial advice.