Skip to content

MRR — Monthly Recurring Revenue

Normalized monthly revenue from all active subscriptions. The number everything else is built on.

The idea

MRR converts subscriptions billed on different cycles into a comparable monthly figure — an annual plan contributes one twelfth per month, a weekly plan contributes 52/12. What makes it interesting is not the arithmetic but the boundary decisions: what counts as active, when a change is recognised, and how non-committed revenue is treated.

Those decisions are where tools disagree, and why metric transparency matters more here than anywhere else.

Why it matters

MRR is the base of the dependency chain. Churn is measured against MRR at period start, NRR and GRR are ratios over it, and LTV derives from ARPU which derives from MRR. An idiosyncratic MRR definition propagates into every downstream number.

ARR is simply MRR × 12. (ChartMogul calls the same quantity Annual Run Rate; Tidemill uses the more common Annual Recurring Revenue.)

How Tidemill computes it

Tidemill splits MRR into two components:

Subscription MRR — committed recurring revenue from licensed items, normalized to a monthly interval and computed at subscription-event time.

Usage MRR — a trailing 3-month average of finalized usage charges, matching ChartMogul and Baremetrics convention. Smoothing keeps bursty workloads from whipsawing the number; the cost is a ~1.5-month lag. Customers with under three months of history average over what exists, so a first-month customer with $40 of usage carries $40, not $13.33.

The smoothed usage component is deliberately not the same thing as usage revenue actuals, which report raw monthly charges unsmoothed. Both are backed by the same table, so they cannot drift apart.

Every subscription change produces exactly one movement — new, expansion, contraction, churn, or reactivation — and the MRR waterfall chains these month over month.

Authoritative formulas: definitions.md — MRR. Implementation: tidemill/metrics/mrr/, documented in metrics.

Contested ground

  • Annual plans — spread monthly (Tidemill and the category standard) or recognised at payment time? Spreading is near-universal in subscription analytics but disagrees with cash accounting, which is a common source of CFO-vs-dashboard arguments.
  • Usage revenue — whether it belongs in MRR at all. It is not committed, so counting a smoothed average as "recurring" is a convention, not a fact. Tidemill states the convention explicitly and exposes actuals separately.
  • Mid-period extrapolation — ChartMogul projects incomplete periods to a full-period rate; Tidemill shows actuals only.

Sources