NRR Benchmarks (2026): Net Revenue Retention by Segment, Stage & Valuation

NRR is a retention metric to your customer success team and a valuation metric to your investors. The number means something different to each — and the gap between those two readings is where most SaaS companies lose money before they realize it.

This is a benchmark reference: what's normal for net revenue retention (NRR) and gross revenue retention (GRR), how the numbers move by segment and stage, what the public leaders actually report, and how retention translates into valuation. It does not cover how to improve your NRR — for that, see the SaaS churn playbook and the guide to increasing subscription LTV. For the definition and formula, see the NRR glossary entry, and to run your own number, use the NRR calculator.

Every figure below traces to a named primary source with its reporting period attached. Aggregate benchmarks come from SaaS Capital's 2026 private-company survey (1,000+ companies), ChartMogul's SaaS Retention Report, Optifai's B2B SaaS Benchmarks, Bessemer's State of the Cloud and the BVP Nasdaq Emerging Cloud Index, m3ter's 2026 usage-based pricing data, and Scale Venture Partners' Scale Studio. Public-company figures come straight from each company's most recent 10-Q, 8-K, or earnings disclosure. Valuation multiples are from Software Equity Group's public-market analysis, date-stamped to the quarter. The mix of platform aggregations and public filings is deliberate — it's what makes the numbers checkable.

One caution before the tables: not every company measures retention the same way. NRR (also reported as dollar-based net retention or net dollar retention) captures revenue change from the existing base including expansion, contraction, and churn. MongoDB calls its version "net ARR expansion rate" — functionally the same. Gross revenue retention (GRR) strips out expansion, so it exposes the leaky bucket NRR can hide. And a few companies report a renewal rate — a contract or logo metric that is not NRR at all. We've labeled the metric in every row so the comparisons stay honest.

The single most reliable way to benchmark NRR is by average contract value. More than industry or funding stage, companies with similar ACVs share the structural traits that govern retention — how they sell, how they support customers, and crucially how much room a customer has to expand. A self-serve tool at $40/month has a low expansion ceiling; a six-figure enterprise platform has seats, modules, and use cases to grow into for years.

The ~20-point spread between the enterprise and SMB bands isn't a quality gap — it's structural. Higher-ACV relationships involve longer sales cycles, dedicated implementation, and account management, all of which raise switching costs, while bigger budgets give those customers more surface to expand across. The practical takeaway: judge your NRR against your ACV band, not against the headline median. A 100% NRR is solid for an SMB-focused product and a warning sign for an enterprise one.

NRR tends to climb as a company scales, and the reasons are mechanical rather than motivational. A larger base means no single account swings the number; expansion motions (upsell, cross-sell, usage growth) mature and get resourced; and the customer mix diversifies away from the fragile early adopters that dominate a pre-PMF book. The result is a fairly consistent walk upward by stage.

The $10–50M band is where investor scrutiny on NRR sharpens most — it's the stage where boards expect clean cohort data and start treating sub-110% NRR (for enterprise-selling companies) as a thesis risk rather than a growing pain.

No competitor benchmark page publishes this: current-quarter NRR from named public companies, each cited to its most recent filing. These are point-in-time figures — read them as "as of each company's most recent filing," not as a live ticker. The mechanism column is the interesting part: it explains why the order looks the way it does.

Reading the board top to bottom:

• Snowflake (126%, consumption). Customers pay for compute and storage as they use it, so expansion happens automatically as workloads grow — no upsell motion required to lift revenue. That's the structural reason consumption names sit at the top.
• Datadog (low-120s, consumption). More hosts monitored and more telemetry ingested means more spend, and customers steadily adopt additional products across the observability suite. Datadog usefully discloses GRR too (mid-to-high 90s), so you can see the expansion is sitting on a genuinely sticky base.
• MongoDB (121%, consumption). Atlas spend scales with the customer's own application data and traffic; as their app grows, so does their bill — expansion tracks customer success directly.
• Atlassian (120%+, seat + metered). The instructive case: nominally seat-based subscription pricing, yet it clears 120% because it has layered metered AI credits (Rovo) and bundled collections on top of seats. The metered surface is what lifts it above a pure seat model.
• Cloudflare (118%, hybrid). Blends subscription and per-seat pricing with usage-based consumption across its network and security products — a middle path that still produces strong expansion.
• HubSpot (103%, seat + cross-sell). Expansion comes from adding seats and cross-selling hubs, with no metered consumption layer underneath. That's the ceiling on a seat-led model, and it's exactly why HubSpot sits ~17 points below the consumption names despite a strong product and brand.

A note on companies people expect to see here: ServiceNow reports a 97% renewal rate (Q1 2026), which is a contract/logo-retention metric — not NRR — so it isn't directly comparable and we've kept it out of the ranking. And the names that were synonymous with sky-high NRR a few years ago have come back to earth: Snowflake reported 158% as recently as FY2023 and is now 126%; CrowdStrike (115% dollar-based net retention, GRR 97%, Q4 FY2026) and Twilio (114% dollar-based net expansion, Q1 FY2026) have both dropped below 120%. That trajectory is the real story — 120%+ is genuinely top-tier today, not the baseline it briefly seemed to be.

GRR is the metric almost no benchmark page covers as a first-class topic, and it's the one that prevents NRR from misleading you.

Why GRR is capped at 100%

GRR counts only the bad things — churn and contraction — and never adds expansion back. The formula is NRR's, minus the expansion term: GRR = (Starting MRR − Contraction − Churn) ÷ Starting MRR. Because you can't retain more than 100% of what you started with when expansion is excluded, GRR maxes out at 100% by construction. That makes it the cleaner read on stickiness: NRR can run to 120%, 130%, 140% on the strength of upsells, but GRR tells you, with no flattery, how much of last year's revenue base would still be there if no customer ever expanded.

What healthy GRR looks like by segment

GRR follows the same segment logic as NRR but at lower absolute levels, since it can't borrow from expansion. Directionally, enterprise SaaS tends to hold GRR around 90%+ annually and mid-market in the high-80s — though those segment-level cuts are estimates pending confirmation (see the table note below) — while SMB and self-serve products run lower, often 80–88%, since smaller customers churn more readily with less contractual lock-in. The publicly sourced anchors carry the pattern cleanly: SaaS Capital's 2026 data puts the private-company GRR median at 91%, and among the public leaders Datadog discloses GRR in the mid-to-high 90s, CrowdStrike at 97%, and HubSpot in the high-80s — the higher the ACV and the stickier the product, the closer GRR sits to its 100% ceiling.

What the NRR–GRR gap signals about the business model

The gap between the two numbers is the expansion contribution, and its size tells you what kind of growth engine you're looking at. A wide gap on a low GRR base (say 110% NRR / 82% GRR) is an expansion-led model: growth depends on continuously out-expanding heavy churn, which works until expansion slows and the leak shows. A narrow gap on a high GRR base (105% NRR / 93% GRR) is a retention-led model: less dramatic, more durable, and far more valuable in a downturn when expansion is the first thing customers cut. Neither is automatically wrong — but they carry very different risk, and NRR alone hides the difference.

A worked example

Take two companies that look similar on NRR:

• Company A — 110% NRR, 85% GRR. The 25-point gap means expansion is masking 15% gross churn. Strip out the upsells and this is a leaky bucket; growth depends on continuously out-expanding the losses.
• Company B — 105% NRR, 92% GRR. A 13-point expansion contribution on a base that barely leaks. This is a genuinely sticky business.

Company B is the healthier company despite the lower headline NRR. This is why investors look at both numbers, and why the gap between them — not NRR alone — tells you whether you're looking at an expansion-led or a retention-led growth model. (Scale Venture Partners' Scale Studio framework treats gross and net churn as separate vital signs for exactly this reason.)

It's tempting to conclude "usage-based pricing produces higher NRR." The aggregate data points that way — m3ter's 2026 analysis puts usage-based models around 115–130% NRR versus flat-rate's 95–105%. But the leaderboard complicates the clean version of that story, and the more accurate framing is about mechanism, not pricing label.

Atlassian is the case that breaks the simple pricing-model story. It's seat-based subscription pricing, yet it clears 120% — outrunning consumption-blended Cloudflare — because it has layered a metered surface onto seats (AI credits and bundled collections that customers consume and expand into). HubSpot, also seat-based but without that metered layer, sits at 103%. The gap between two "subscription" companies is the whole argument: what compounds NRR is metered expansion headroom, not the pricing label on the contract.

The practical implication isn't "rip out your pricing model." It's "give existing customers a metered way to spend more" — consumption tiers, usage-linked modules, AI credits. For the playbook on actually building those expansion motions, see the guide to increasing subscription LTV.

NRR is one of the cleanest predictors of enterprise value, because investors pay for revenue that compounds on its own. The clearest read comes from a single index at a single point in time — here, the Software Equity Group SaaS Index, public EV/TTM revenue, Q4 2024:

In SEG's own words, the >120% cohort traded at nearly triple the multiple of companies below 100% NRR. An earlier SEG cut (2Q24) showed the same shape at lower absolute levels — the >120% group at 9.3x versus a 5.7x median, a 63% premium. The important point is what happened next: absolute multiples compressed market-wide, with the SEG index median falling to roughly 3.6x by Q1 2026 amid an AI-driven re-rating, yet the premium commanded by top-retention companies persisted across both cuts. Retention didn't stop being rewarded; the whole market simply re-priced downward together.

Two cautions on reading multiples. First, indices differ by composition — the SaaS Capital Index sat near 6.4x in early 2026 and the BVP Emerging Cloud Index higher still, because they weight toward different companies than SEG. We anchor to one index here precisely so the comparison holds together; don't mix them. Second, these are public EV/revenue comps. Private lower-middle-market SaaS M&A is a different population that prices differently — those deals more typically close at 3–6x ARR, with 7–9x reserved for genuinely elite assets (strong Rule-of-40 plus 120%+ NRR). Keep the two worlds separate.

For how NRR fits alongside other valuation drivers in a churn-reduction context, see the SaaS churn rate guide.

What a few points of NRR compound to

NRR isn't linear — it compounds. Here's what each NRR level does to your existing customer base over five years, assuming you acquire zero new customers:

A company at 120% NRR roughly 2.5x's its existing-customer revenue over five years without signing a single new logo. A company at 90% loses 41% of it. That spread — entirely a function of retention and expansion — is why the metric carries the valuation weight it does, and why two otherwise-similar companies at 105% and 120% NRR diverge so sharply over time.

Use your ACV band to locate a realistic median target and a stretch goal:

NRR = (Starting MRR + Expansion − Contraction − Churn) ÷ Starting MRR. Start a month at $100K MRR, add $5K from upgrades, lose $2K to downgrades and $3K to cancellations, and you land at 100%. It's usually measured monthly and reported annualized. The NRR calculator handles the math; the glossary entry walks the formula in full.

NRR has exactly two levers: reduce the losses (churn and contraction) and grow the gains (expansion). On the loss side, the highest-impact moves are well-targeted cancel flows with save offers like discounts, pauses, and plan switches, plus one-click reactivation to win back customers who do leave, with real-time analytics to measure the impact. On the gain side, it's expansion headroom — the metered surfaces discussed above. This page is the benchmark reference; for the actual execution playbooks, see the SaaS churn guide and the LTV guide.