What's a good NRR for Series B SaaS in 2026?

The board meeting where a good number turned into a bad one

Picture a Series B company at $28M ARR, mid-market horizontal software, seat-priced, growing 55% year over year. The CEO opens the retention slide with 118% NRR and the room relaxes. Two quarters later the same slide reads 104%, nothing about the product changed, and nobody in the room can explain the drop. This is the single most common retention story at Series B, and it almost never means what the board first assumes.

What actually happened is visible only when the blended number comes apart. That 118% was carried by three accounts — a logistics customer that rolled the product out to two new business units, a fintech that upgraded two tiers after an audit requirement, and a retailer that tripled seats going into a peak season. Together those three accounts contributed roughly 60% of all expansion ARR in the period while representing about 9% of starting ARR. The other 91% of the base expanded at a blended rate closer to 103%. When the three whales had a flat year — not a bad year, a flat one — the headline collapsed by 14 points with zero change in the health of the rest of the business.

The second thing hiding under that number was segment mix. The company sold to mid-market as its stated ICP but had accumulated a long tail of sub-$15K SMB accounts through a self-serve motion nobody owned. That tail was retaining at roughly 91% gross and expanding almost not at all, and it had grown from 8% of ARR to 19% of ARR over two years. The blend absorbed it invisibly until the enterprise expansion stopped covering it.

The third thing was definitional. The finance team had been computing NRR monthly and compounding it to an annual figure, which smooths out the lumpiness of annual renewal cycles and reliably reads two to four points higher than a true trailing-twelve-month cohort measurement. When the Series C diligence team rebuilt the number from the raw billing export on annual cohorts, they got 109% for the period the deck had reported as 118%. Nothing was fraudulent. The methodology had simply never been written down, and each quarter's analyst made a slightly different judgment call.

The practical lesson for any Series B operator: the question "what's a good NRR" is unanswerable until you have specified segment, pricing model, cohort definition, and concentration. A 118% that is really 109%, carried by three accounts, and masking a 91% SMB tail is a worse business than a stable, broadly distributed 108%. Investors know this, which is why they rebuild the number rather than reading it off your slide. The fastest way to lose credibility in a Series C process is to present a figure the data room contradicts.

How net revenue retention actually gets built and broken

Net revenue retention measures what happened to the dollars you already had. Take every account that was active twelve months ago, lock that cohort, and compute: NRR = (starting ARR + expansion ARR − contraction ARR − churned ARR) ÷ starting ARR. New logos acquired during the period are excluded entirely — if they leak into the numerator you are no longer measuring retention, you are measuring total ARR growth with extra arithmetic.

A worked example makes the mechanics concrete. Suppose that cohort started with $20M of ARR. Over twelve months, existing accounts added $4.0M through seat growth, tier upgrades, and a second module. Separately, $1.0M was lost to downgrades and seat reductions from accounts that stayed, and $1.5M was lost to accounts that left entirely. NRR = ($20M + $4.0M − $1.0M − $1.5M) ÷ $20M = 107.5%. Gross revenue retention, which strips expansion out completely, = ($20M − $1.0M − $1.5M) ÷ $20M = 87.5%. That pairing tells the real story: an expansion engine producing 20 points of growth on top of a base leaking 12.5 points a year.

The variants matter as much as the formula. Dollar-based retention weights by revenue, so losing ten $5K accounts barely moves it while losing one $400K account is devastating; logo retention counts accounts and inverts those weights. Management NRR that includes signed-but-not-yet-billing expansion reads higher than a strictly billed view. Monthly cohorts compounded annually read smoother than true annual cohorts. Currency effects move ARR for international customers with no behavior change at all — a euro-denominated account that did nothing can appear as expansion or contraction purely on FX.

The defensible standard for a Series B board deck is a trailing-twelve-month, dollar-based, cohort-locked NRR, reported alongside GRR, with new-logo revenue explicitly excluded and FX either normalized or disclosed separately. Write the methodology down in a one-page memo, get finance and the CRO to sign it, and never change it mid-fundraise. Investors will rebuild your number from the data room regardless; a definition that flatters you erodes trust far faster than an honest lower figure would have.

The other structural point buried in the formula: contraction and churn are subtracted from a base that also generates expansion, so the two forces compete inside a single number. This is why NRR alone can never diagnose anything. A company moving from 112% to 104% might have lost eight points of expansion, gained eight points of churn, or done both by four points each — three entirely different problems with three entirely different fixes, indistinguishable from the headline.

The 2026 benchmark bands, by segment and by pricing model

The honest 2026 bands are tighter and lower than the numbers most boards still quote, because the folklore in circulation was set during the 2021 peak and has not been updated.

For a Series B company between roughly $10M and $50M ARR: 105–115% is solid and fundable, 115–125% is strong and attracts competitive term sheets, 125%+ is genuinely elite and rare. The 100–105% band is yellow — not disqualifying, but you will face hard questions about whether an expansion motion exists at all or whether you simply have not started losing customers yet. Below 100% is a red flag: your base is shrinking, and every point of growth must be purchased through the most expensive channel available.

Segment moves those bands more than stage does. SMB-focused: 95–105% is realistic. Small businesses fail, get acquired, cut tools in a downturn, and switch for cheaper alternatives; logo churn of 2–4% monthly compounds to 25–40% annually. Expansion room is also limited — there is not much upside in growing a five-seat customer. An SMB-heavy Series B at 105% is doing genuinely well, 100% is normal, below 95% signals real trouble. Mid-market: 105–115%. Switching costs are higher, buyers are more deliberate, and there is meaningful room to add seats, departments, and modules; annual logo churn typically runs 8–15%. Enterprise: 115–130%. Annual logo churn of 3–8% is typical because procurement cost, integration depth, and political capital make switching painful, and the expansion runway across business units, geographies, and use cases is enormous. An enterprise-motion Series B below 115% has an expansion problem, not a market problem.

The spread comes down to two structural variables. Churn floor is how often customers leave for reasons you cannot control; expansion ceiling is how much headroom exists inside an account. SMB has a high floor and a low ceiling, so both forces push NRR down. Enterprise has a low floor and a high ceiling, so both push it up. Comparing your SMB company's 103% to a competitor's enterprise 126% tells you nothing whatsoever.

Pricing model is the second axis, and it is roughly as powerful. Seat-based pricing typically caps around 105–115%. Revenue is tied to your customer's headcount, a variable you do not control and which has frequently moved down since 2022. Expansion requires the customer to hire or roll out to new teams; contraction happens mechanically when they cut staff, with no product failure required. Usage-based pricing with real product-market fit can reach 115–140%. When you charge for data processed, API calls, compute, or transactions, a customer who succeeds consumes more and your revenue rises with zero sales motion. The canonical consumption-priced public companies — Snowflake, Datadog and their peers — have historically disclosed NRR well above 130%, and even after substantial compression as their bases matured they sit in territory a seat-priced company structurally cannot reach. Hybrid — a committed platform or seat base plus consumption on top — realistically supports 110–125% with materially less volatility than pure usage, and it is the model best shaped for AI features, which are naturally consumption-priced.

Public disclosures are the best available calibration because they are audited, comparable, and refreshed quarterly. Use them as anchors rather than targets: consumption-priced enterprise infrastructure sits at the top of the range, horizontal seat-priced mid-market software clusters near 110%, and SMB-heavy seat-and-tier businesses sit around or just above 100% — which for that segment is a structural reality, not a weakness.

Why the compression happened matters for diagnosis, because it tells you whether your softness is market or self-inflicted. Four forces reinforced each other. The end of cheap capital meant customers who bought ahead of need in 2021 now audit every line item. Procurement professionalized — spend-management platforms and finance gatekeepers now sit between you and the budget owner, negotiating out auto-expansion clauses and cutting nice-to-have modules at renewal. Layoffs across tech mechanically reduced seat counts for horizontal seat-priced vendors, the single biggest driver of sub-100% NRR in recent cohorts. And AI-driven efficiency at the customer is the newest structural force: a support team that needed 50 agent seats needs fewer once AI handles tier-one volume, which is direct contraction for a seat-priced vendor even when the product is working perfectly.

Choosing between the levers, and what each one costs you

Every point of NRR comes from either adding expansion or removing contraction, and at Series B those are two separate programs competing for the same limited CS and product capacity. Choosing badly wastes a year.

On the expansion side there are five distinct levers. Seat expansion is driven by adoption depth — the more teams actively using the product, the more seats get added organically. It is the most common lever for seat-priced software and the most exposed to your customer's macro conditions. Tier upgrades are high-margin and predictable if packaging is designed so growing customers naturally hit a ceiling; bad packaging leaves customers comfortable on the cheap tier forever. Cross-sell of a second module has the highest ceiling for multi-product companies, and the discipline at Series B is making the second product a natural extension sold to the same buyer rather than a random adjacency requiring a new sales cycle. Usage growth requires no sales motion at all but is the least controllable — it tracks the customer's own trajectory. Price increases — a 5–7% annual uplift on a sticky base — are a legitimate and chronically underused lever at Series B, because founders fear churn that rarely materializes among healthy accounts when the uplift comes with notice and value framing.

On the contraction side the drivers are equally distinct. Seat reduction from customer headcount cuts is largely macro-driven but somewhat predictable if you watch hiring signals. Downgrades are value-perception failures that surface at renewal and are usually preventable through value demonstration months earlier. Full churn clusters into uncontrollable causes (business failed or acquired), semi-controllable (champion left and the replacement never valued the product), and fully controllable (never adopted, or displaced by a competitor). And silent contraction is the most dangerous because billing shows nothing: the customer quietly stops using features or lets seat utilization drift below what they pay for, and the entire adjustment lands at once on renewal day.

The trade-off between the two programs is real. Expansion work is more attractive — it produces visible bookings, it is cheaper per dollar than new-logo revenue, and it makes the headline move fast. Contraction work is unglamorous, and its wins are invisible because they show up as things that did not happen. But pouring expansion resources onto a gross-retention problem is a treadmill: you are growing a base that leaks underneath you, and the moment expansion plateaus the churn becomes fully visible with nothing covering it. The rule is to hold the GRR floor first — 85% SMB, 90% mid-market, 92%+ enterprise — and only then spend on expansion.

There is also a genuine alternative to both, and it is the highest-ceiling move available: repricing. If your value metric is structurally disconnected from value delivered — a flat platform fee, or seats that do not track outcomes — no expansion team can manufacture what the pricing model forbids. A customer getting twice the value should pay roughly twice as much, and when that holds, expansion is automatic. When it does not, expansion requires a sales motion every single time. Repricing is slow, risky, and disruptive to the existing base, which is why most Series B companies avoid it. It is also why some of them stay stuck at 106% no matter how many CSMs they hire.

Ownership is the last trade-off and the most fumbled. CSM-owned expansion works when CSMs are genuinely commercial and comped accordingly, and corrupts the trusted-advisor relationship when they are not. AM-owned expansion separates adoption from the commercial conversation cleanly but introduces handoff friction. AE-owned expansion preserves continuity but loses to next quarter's new-logo quota nearly every time. The pragmatic Series B answer is usually CSMs measured on GRR and health, with a commercially comped role carrying the expansion number. The fatal pattern is diffuse ownership — when everyone owns NRR, nobody does, and the number drifts.

The mistakes that cost Series B companies their Series C

Presenting NRR without GRR. Showing net retention alone signals to any experienced investor that you are hiding the gross number. A 112% NRR with 95% GRR is a genuinely healthy company; a 112% NRR with 80% GRR is a business in trouble that does not know it yet. Report both, side by side, every board cycle, with the decomposition into expansion, contraction, and churn underneath.

Reading the blend instead of the cuts. A single company-wide figure is an average that can look entirely healthy while a segment bleeds. A 112% blended number might be a 128% enterprise segment masking a 92% SMB segment that is destroying value. Always segment before you present: by customer size, by product line, by acquisition channel, by geography. The blend is exactly where problems hide, and a board that only sees the blend cannot govern.

Ignoring concentration. If your blended NRR is 122% but your NRR excluding the top ten accounts is 103%, you do not have a 122% company. Present the concentration cut every time — NRR ex-top-5 and ex-top-10, and the share of total expansion ARR that came from the top decile. Concentration-carried NRR is fragile in a way distributed NRR is not, and it swings on accounts whose renewals come with the most pricing leverage against you.

Skipping cohort analysis. A rising blended number with deteriorating recent cohorts is a company about to hit a wall. Track NRR by signup cohort, comparable at the same age: if your Q1 2024 cohort retained better at month 18 than your Q1 2025 cohort does at month 18, your ICP targeting, onboarding, or qualification has degraded, and the blend — dominated by large mature cohorts — will hide that for another year. Series C diligence teams ask for the cohort table directly.

Diagnosing before segmenting. When NRR falls, the disciplined sequence is: segment it, cohort it within the troubled segment, decompose into GRR versus expansion, then identify the specific mechanism — logo churn or seat contraction or downgrades; seats or tiers or cross-sell or usage. Teams that "work on NRR" from a blended average and an intuition burn quarters. Teams that recover fast diagnosed precisely first.

Never reconciling CRM to billing. NRR is computed from the intersection of your CRM and your billing system, and those systems disagree constantly — deals marked closed-won that are not yet billing, billing changes never reflected in the CRM, currency handled differently in each. Before a Series C you should be able to tie NRR back to billed revenue exactly, with a documented reconciliation. A beautiful dashboard on unreconciled data is a liability: the moment diligence finds the gap, every number you have ever presented is in question.

Riding the NRR treadmill. Discounting expansion to manufacture the number lifts NRR while quietly rotting unit economics and training the base to expect discounts. Tilting all energy toward the existing base because expansion is cheaper lets the new-logo engine atrophy, capping total growth at whatever the current base can absorb. Steering product toward whatever the installed base will pay for next produces real revenue and real strategic drift. NRR is a means; the end is durable, efficient growth, which needs expansion and new logos and healthy margins together.

Forecasting from the spreadsheet instead of the leading indicators. NRR is a trailing metric with real leading indicators: the distribution of health scores (not just the average), usage depth and breadth trends, support sentiment and escalation clustering, and champion turnover. A credible forecast combines the contracted renewal and expansion pipeline with a leading-indicator overlay. Sophisticated investors are testing exactly this when they ask how you forecast retention.

Finally, understand what the number is worth. NRR is, after growth rate itself, the metric that most shapes how a SaaS business is underwritten, because it proxies product-market fit, capital efficiency, and predictability simultaneously — a 120% NRR company can grow 20% annually selling nothing to anyone new. That optionality is why the rough heuristic holds that each ten points of NRR below the segment benchmark costs roughly half a turn to a full turn of forward-revenue multiple, with the effect strongest around the 100% line where the entire growth story changes character. It is also why NRR work outranks nearly any other operating improvement available to a RevOps leader at this stage: ten points does not just add expansion revenue this year, it re-rates the durability of every future year.

Related questions

What NRR do investors require to lead a Series C?

At or above your segment benchmark: roughly 105%+ for SMB-heavy, 110%+ for mid-market, 115%+ for enterprise — paired with a healthy GRR floor and an expansion motion that is systematized rather than founder-dependent. A stable 110% beats a 115% that fell from 128%.

Is GRR or NRR the more important metric?

GRR is the truer health metric because it cannot be inflated by a few expanding accounts; NRR measures how well you monetize the base you keep. Report both. When NRR is strong and GRR is weak, fix retention first — expansion on a leaking base is a treadmill.

Why did NRR benchmarks drop from the 2021 numbers?

Four reinforcing forces: the end of cheap capital ending speculative buying, professionalized procurement gating every uplift, headcount cuts mechanically reducing seats in seat-priced products, and AI efficiency letting customers do the same work with fewer licensed users. Median NRR fell roughly 10–20 points depending on segment.

Does usage-based pricing really produce better NRR?

On average yes — consumption models convert customer success directly into revenue with no sales motion, reaching 115–140% with strong fit. But the same mechanism cuts downward without a renewal event to defend, and quarterly swings of 15–20 points on macro conditions are common. Higher mean, higher variance.

How often should a Series B company review NRR?

Monthly at the leadership level with segment and cohort cuts, quarterly as a board deep-dive with GRR and decomposition, and always-on for the leading indicators — health scores, usage, support sentiment — that the CS and account teams actually act on. NRR itself is trailing-twelve-month and does not need daily refresh.

FAQ

Is 110% NRR good for a Series B SaaS company in 2026?

For mid-market it is right at the healthy midpoint and comfortably fundable. For SMB-focused it is strong, above the realistic 95–105% band. For an enterprise-motion company it is below benchmark and signals an expansion problem, since enterprise accounts should support 115–130%. The same figure is good, great, or concerning depending entirely on segment.

What is the minimum NRR to avoid a red flag at Series B?

100% is the hard line — below it your existing base is shrinking and every dollar of growth must come from the most expensive channel available. Between 100% and 105% you are in a yellow zone that invites hard questions about whether a real expansion motion exists. For enterprise-motion companies, sub-100% is close to disqualifying.

How is NRR different from gross revenue retention?

NRR includes expansion revenue and can exceed 100%; GRR strips expansion out entirely and is capped at 100%. GRR = (starting ARR − contraction − churn) ÷ starting ARR. Hold GRR floors of 85% for SMB, 90% for mid-market, and 92%+ for enterprise — below those, whatever NRR you report is built on sand.

Can a good NRR hide a bad business?

Yes, in two ways. Concentration: a handful of whale accounts expanding can carry a blended 122% while the rest of the base sits at 103%. And a leaky bucket: aggressive expansion from surviving customers can mask 18% annual base erosion. Both are exactly why diligence teams rebuild NRR cohorted, segmented, and ex-top-accounts.

Does raising prices count as legitimate NRR expansion?

Yes. A 5–7% annual uplift on a sticky base is a real and chronically underused lever at Series B, and executed with advance notice and value framing it rarely drives churn among healthy accounts. The caution is discounting in the other direction — buying expansion bookings with price concessions manufactures the number while rotting unit economics.

How much does ten points of NRR affect valuation?

The working heuristic is roughly 0.5–1.0x of forward revenue multiple per ten points relative to the segment benchmark, with the effect strongest around the 100% line where the growth story changes character entirely. That is why NRR improvement outranks most other operating work — it re-rates the durability of all future growth, not just this year's ARR.

Sources

• https://www.bvp.com/atlas/state-of-the-cloud
• https://openviewpartners.com/blog/
• https://www.saastr.com/
• https://a16z.com/
• https://investors.snowflake.com/
• https://investors.datadoghq.com/
• https://ir.hubspot.com/
• https://www.klipfolio.com/resources/kpi-examples/saas/net-revenue-retention
• https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights
• https://www.sec.gov/edgar/searchedgar/companysearch

Related on PULSE

• What's a healthy gross revenue retention rate by segment?
• How do you build a customer expansion motion at Series B?
• Should CSMs or account managers own expansion revenue?
• How does usage-based pricing change SaaS unit economics?
• What CAC payback period do investors expect in 2026?
• How do you run a cohort retention analysis in a RevOps stack?