Skip to main content

ARR & Net Revenue Retention

Updated Jun 24, 2026 at 8:22pm

Research Draft High 1,124 words

Annual Recurring Revenue (ARR) and Net Revenue Retention (NRR) are the two metrics that most define how subscription-software (SaaS) businesses are described, valued, and underwritten. ARR is a level — the annualized run-rate of recurring subscription revenue under contract — while NRR is a rate of change that isolates how the revenue from an existing cohort of customers evolves over a year once you net upsells and price increases against churn and downgrades, deliberately excluding any new logos. The core tension is that both are powerful precisely because they compress a messy income statement into one momentum number, and both are non-GAAP — unaudited, undefined by any standard-setter, and therefore as much a narrative tool as a measurement.

How they're calculated

ARR is the sum of the annualized value of all active recurring contracts at a point in time: monthly subscriptions × 12, annual contracts at their yearly value, multi-year contracts at their per-year value. It deliberately excludes one-time fees, professional services, and usage overages that are not contractually recurring. Variants proliferate: MRR (monthly), CARR (committed/contracted ARR, which includes signed-but-not-yet-live deals), and "ARR" definitions that quietly fold in non-recurring revenue. None is GAAP — ARR is a run-rate, not the revenue a company will actually recognize under ASC 606.

Net Revenue Retention measures a fixed cohort over (usually) twelve months:

NRR = (Starting ARR + Expansion − Contraction − Churn) ÷ Starting ARR

New-customer revenue is excluded — that is the whole point. Above 100% means the existing base is growing on its own; below 100% means it is leaking. NRR is frequently labelled NDR (Net Dollar Retention) or DBNRR (Dollar-Based Net Revenue Retention); these are the same metric.

Gross Revenue Retention (GRR) is the conservative sibling: it nets only contraction and churn against starting ARR and gives no credit for expansion, so it is mathematically capped at 100%. The gap between NRR and GRR tells you how much of "retention" is really expansion masking churn. Per CFI and ChurnZero, the two should always be read together.

How they're used in practice

ARR growth is the headline momentum metric: the rate of change in ARR, plus the Rule of 40 (revenue-growth % + a profitability-margin % ≥ 40 — most commonly EBITDA margin, sometimes FCF or operating margin depending on the analyst), is what most software valuation conversations start from. NRR is the quality check on that growth. A business with 120%+ NRR has a compounding engine — it would grow even if it stopped acquiring customers — which is why analysts treat high NRR as evidence of pricing power, product stickiness, and a low-cost growth flywheel. Investors decompose ARR growth into new logo vs. expansion; a company leaning heavily on expansion (high NRR) is structurally more efficient than one buying growth through sales-and-marketing on new logos. NRR is also a leading indicator: deterioration shows up in the installed base before it shows up in total revenue, making it an early warning for a growth slowdown.

Adoption, debate & evidence

Both metrics are near-universal in SaaS investor decks, S-1s, and earnings releases, and most public software companies disclose some form of NRR/NDR. Benchmarks (treat as directional, definitions vary by source):

  • Private-SaaS median NRR is commonly cited around 101–102% in 2024–25, reported as slipping from roughly ~105% at the 2021 peak (SaaS Capital's 2025 retention work reports a 102% median in the $25K–$50K ACV band, and independent aggregators corroborate the ~105%→~101% decline). Median GRR is repeatedly cited near ~91% by SaaS Capital across multiple survey years.
  • NRR rises sharply with deal size: roughly ~118% enterprise (ACV >$100K), ~108% mid-market, ~97% SMB in commonly cited segment data — small customers churn far more.
  • Public SaaS median NDR held near ~110% through several 2023–24 quarters per Growth Unhinged; public figures run above private medians partly from survivorship bias and reporting incentives, a caveat the source flags explicitly.

The serious debate is over ARR's lack of definition. Because ARR is unregulated, companies define it to their advantage — bundling maintenance tails of legacy license businesses, including non-recurring items, or counting CARR as ARR. CFO Dive reports SaaS metrics are frequently misreported even at CFO-staffed firms, and the SEC keeps non-GAAP measures among its top comment areas. NRR is equally manipulable: the chosen cohort window, whether downgrades net within the period, and whether partially-churned-then-re-expanded accounts are counted all swing the number. Two companies' NRR figures are rarely strictly comparable.

Strengths & limitations

These metrics work best as a pair and a trend: ARR growth for momentum, NRR for the quality of that growth, GRR to expose churn hidden under expansion, all watched over several quarters. They genuinely capture something GAAP revenue smears out — the health and stickiness of the installed base.

They fail when read in isolation or taken at face value. The #1 misuse is treating a single high NRR as proof of a healthy business: a handful of large-account upsells (or usage-driven expansion at a few customers) can lift NRR above 100% while the company is quietly losing most of its small customers — the GRR gap reveals this. Other traps: a single mega-renewal distorting the cohort; usage-based models where "expansion" is just a customer's good quarter (and contracts in a downturn); ARR that includes a shrinking license-maintenance base dressed as recurring SaaS; and the fact that ARR ignores cash timing, gross margin, and CAC entirely — a company can grow ARR fast while burning cash unsustainably. Because none of it is audited, always reconcile ARR against GAAP revenue and deferred-revenue/billings trends before trusting it.

Sources

Disputes flagged: ARR and NRR are unregulated non-GAAP metrics with no standard definition; cross-company comparisons and precise benchmark figures vary materially by source and methodology (segment, cohort window, public vs private, survivorship bias). All benchmark numbers above are directional, not authoritative.