Net Revenue Retention — A Field Guide

Net Revenue Retention is the metric that separates a SaaS business that grows because it keeps land-and-expanding within its existing base from one that's a leaky bucket refilled entirely by new sales. It's also one of the easiest metrics to compute inconsistently across two companies claiming the same number — the formula looks simple until you have to decide exactly which dollars go in which bucket.

The formula

NRR = (Starting MRR + Expansion − Contraction − Churn) / Starting MRR, measured over a cohort of customers that existed at the start of the period, with no new customers added to the numerator. It's a retention metric, not a growth metric — new logo revenue never enters the calculation at all.

WITH cohort AS (
  -- customers active exactly 12 months ago
  SELECT customer_id, mrr AS starting_mrr
  FROM mrr_snapshots
  WHERE snapshot_month = DATE_TRUNC('month', CURRENT_DATE) - INTERVAL '12 months'
),
current_state AS (
  SELECT c.customer_id, c.starting_mrr,
         COALESCE(s.mrr, 0) AS current_mrr
  FROM cohort c
  LEFT JOIN mrr_snapshots s
    ON s.customer_id = c.customer_id
   AND s.snapshot_month = DATE_TRUNC('month', CURRENT_DATE)
)
SELECT
  SUM(starting_mrr)                         AS starting_mrr,
  SUM(current_mrr)                          AS ending_mrr,
  ROUND(SUM(current_mrr) / NULLIF(SUM(starting_mrr), 0) * 100, 1) AS nrr_pct
FROM current_state;

Note the LEFT JOIN: a customer from the cohort who churned entirely contributes current_mrr = 0, not a missing row. Drop churned customers from the denominator instead of counting them at zero, and you'll compute a number that looks better than reality — this is the single most common NRR calculation bug I've seen in homegrown dashboards.

Why anything above 100% is the headline

NRR over 100% means your existing customer base would grow revenue even if you closed zero new logos next year — expansion (upsells, seat growth, upgrades) is outpacing contraction and churn within the same cohort. Public SaaS benchmarks put "good" NRR around 100-110% and "best-in-class" north of 120%, though the number that matters is the trend for your own business more than any external benchmark, since NRR is heavily influenced by pricing model (per-seat pricing naturally expands as customers grow headcount; flat-rate pricing doesn't).

Gross retention vs. net retention

Gross Revenue Retention (GRR) is the same formula but caps expansion at zero — it only measures how much revenue you'd have kept with no upsells at all: (Starting MRR − Contraction − Churn) / Starting MRR, and it can never exceed 100%. GRR isolates your churn and contraction problem from your expansion motion; NRR blends both together. A company with 85% GRR and 115% NRR has a real churn problem masked by a strong expansion motion — worth knowing before you conclude "retention is great" from NRR alone.

Cohort-based calculation pitfalls

Beyond the churned-customer-as-zero mistake above, the other recurring error is mixing cohort vintages: computing NRR against "customers active at any point in the trailing 12 months" instead of a fixed point-in-time cohort inflates the number, because it excludes customers who churned early in the window from the denominator entirely. NRR should always be anchored to a specific starting snapshot date, with the same set of customer IDs tracked forward — no substitutions, no backfilling the cohort with customers who joined partway through.

Wrapping up

NRR is one formula, but the inputs — what counts as expansion, how churned accounts are handled in the denominator, which cohort snapshot you anchor to — are where the number actually gets made or inflated. Compute GRR alongside it so you know whether a shift in NRR came from expansion or from churn, and anchor every calculation to a fixed cohort rather than a rolling window.