The H1 2026 SaaS Retention and NRR Report

SaaS retention in H1 2026 has split by ACV tier: enterprise platforms hold 95–97% gross dollar retention and clear 110–120% NRR, while SMB SaaS is bleeding to AI-copilot feature creep and post-ZIRP fatigue. The Series A bar is 108% NRR minimum, 120%+ to outprice peers. This report has the benchmarks, the churn breakdown, and the retention signature that closes the round.

• The NRR benchmark 2026 table
• Gross retention benchmark by ACV tier
• SaaS churn 2026: where it is rising
• Expansion revenue 2026: the Series A bar
• User retention is not dollar retention, and the deck must say so
• The minimum viable retention signature for Series A
• How to calculate net revenue retention without lying to yourself
• Why this matters for your raise

The headline of H1 2026 SaaS retention is that the benchmark stopped being a single number. Mid-market is holding, enterprise platforms are widening their lead, and SMB SaaS is the cohort where churn is structurally re-rating upward because AI copilots are collapsing pricing power inside small-team workflows. Every retention number in this report is anchored to an ACV tier and a fundraising milestone, because a 92% gross retention rate is great for a $4K ACV product and disqualifying for a $90K one. If your deck has one NRR number on one slide, you are pitching the wrong granularity for 2026.

Capital backdrop matters here. Carta's Q4 2025 read of the market is that "a new type of normal has emerged" as the system "adjusts to the omnipresence of AI", and that startups raised "nearly $120 billion" in 2025 against a much harder underwriting standard (Carta , State of Private Markets: 2025 in Review). PitchBook-NVCA's Q1 2026 read is sharper: at the application layer, the underwriting standard is "customer references and evidence of real value creation, clear return on investment and strong feedback" (PitchBook-NVCA Venture Monitor , Q1 2026). Translation: retention is the proof, and the bar moved.

The NRR benchmark 2026 table

The cleanest way to read the H1 2026 net revenue retention bar is by ACV tier and PMF stage, not by a single median. The table below is the defensible synthesis of the four published rubrics that VCs actually price off in 2026.

Two things to read off the table. First, the 108% NRR line is the threshold between "I have PMF" and "I think I have PMF." If you are pitching a Series A and your NRR is 102%, you are arguing Level 2, not Level 3, and the conversation about price gets capped by that. Second, the 141% NRR / 18-month streak of net negative churn profile that First Round labels "Extreme PMF" is what closes priced rounds at premium multiples in 2026; it is not aspirational, it is the comp set you are getting compared against (First Round Review , Levels of PMF).

The hidden trap in the table is the ICONIQ band. A founder reading "110–120%" and reporting their own NRR at 112% feels safe, but ICONIQ's band is for broader enterprise software that has crossed materiality. At seed and Series A scale, 112% NRR on a tiny base is noise. What VCs want to see is the shape of the curve, not the spot value, which is why the next section breaks it by ACV.

Gross retention benchmark by ACV tier

Gross retention is the truth serum of SaaS retention, because expansion cannot hide it. a16z is explicit: mission-critical platform status "shows up in two key metrics: gross dollar retention (GDR) and average contract values (ACVs)", and customers of true platforms "rarely churn unless they go out of business" (a16z , Anatomy of an Enterprise Platform Company).

The defensible gross retention benchmark for H1 2026 by ACV tier:

• Enterprise (>$100K ACV): a16z's bar is "above 95% and often above 97%" gross dollar retention. If your $120K-ACV product is at 91% GDR, you do not have a platform, you have a tool, and the multiple compresses accordingly.
• Mid-market ($25K to $100K ACV): the practical floor for "Series A defensible" is 85–90% GDR. Below that you are likely losing accounts to consolidation or AI-native displacers and your land motion will outpace your save motion forever.
• SMB (<$25K ACV): 70–80% GDR is the realistic band for sub-$25K-ACV SaaS in H1 2026. SMB churn is structural, not a defect, and the way to clear a Series A on SMB is to make NRR compensate via seat growth, usage expansion, or a multi-product wedge.

The opinion to internalize: do not benchmark your GDR against a broad SaaS median, because the only honest median is the one inside your ACV tier. Compare yourself against the right cohort or the comparison is decorative.

a16z's framing of why GDR matters is the right one to put in your deck. The phrase is "customers rarely churn unless they go out of business" for true enterprise platforms; the reverse claim, that your customers leave when they have a budget cycle or hire someone who likes a different tool, is the definition of non-platform status. Mission-critical and gross dollar retention are the same fact stated twice.

SaaS churn 2026: where it is rising

SaaS churn 2026 is not uniformly up. It is up in two specific cohorts, and flat-to-down in the others. Naming the cohort that is bleeding is the whole point of this section, because none of the top SERP retention pieces will do it for you.

Where churn is rising in H1 2026:

• SMB SaaS in single-workflow categories. AI copilots inside Notion, ChatGPT, Linear, and the platform incumbents have collapsed the "one-trick tool" wedge. If your product was a thin layer of workflow logic over a database, the median SMB buyer is canceling and using a copilot inside their existing stack. This is the dominant churn vector below $25K ACV.
• Mid-market in EU geographies with hard macro exposure. Logos in DACH, Nordics, and UK mid-market are seeing 1–3 point gross retention compression as procurement cycles lengthen and renewals get re-priced down. The fix is not retention tactics, it is multi-year contract structure and pre-renewal QBRs that lock value in writing.
• Prosumer subscriptions at the $10–30/mo price point. Atomico's read of the 2024 European fundraising environment, that only "20% of founders find it easier to raise capital from investors" (Atomico , State of European Tech 2024), is the same macro that is squeezing prosumer subscription budgets. Discretionary SaaS spend gets cut first.

Where churn is flat or improving:

• Enterprise platforms with mission-critical status. GDR is widening, not tightening, because a16z's underlying claim holds: real platforms have customers that "rarely churn unless they go out of business" and AI is making those platforms more, not less, defensible.
• Vertical SaaS with embedded payments or data dependencies. Retention is climbing because switching costs are real.

The contrarian read on H1 2026 SaaS churn: the headline "churn is up" is true for SMB and prosumer, and false for enterprise. If you are pitching a Series A on a vertical or mid-market platform with embedded data dependencies and your gross retention is climbing, you have the most pitchable retention story in the market right now, because the macro narrative makes your numbers look anomalously good.

✅ Good: "GDR climbed from 88% in 2024 to 93% in H1 2026 as customers moved their billing system onto our rails. Embedded payments are the moat." Says what changed, why, and why it compounds.

❌ Bad: "Retention is strong and improving." Says nothing the VC cannot disprove with a 10-second look at your usage data.

Expansion revenue 2026: the Series A bar

Expansion revenue 2026 is the single most diagnostic Series A metric for application-layer SaaS, because it is the cleanest signal that your customers are getting compounding value. a16z's framing makes the mechanism explicit: increasing returns to scale "show up in multi-module adoption and net revenue retention (NRR)", which is why the >120% NRR bar for sub-$1B platforms is set where it is (a16z , Anatomy of an Enterprise Platform Company).

The expansion revenue bar to clear in H1 2026, by motion:

• Seat-led expansion: healthy is 20–30% of new ARR coming from existing accounts via seat growth. If you are below 15%, your product is not landing across teams.
• Usage-based expansion: healthy is 25–40% of new ARR from existing accounts via consumption growth. Usage-based SaaS that does not show this is mispriced.
• Multi-product expansion: at Series A, 10–20% of new ARR from cross-sell on a second SKU is the credibility line. Below 10% you are a one-product company, which is fine at seed and a problem at Series B.

The opinion to hold: expansion is not a marketing channel, it is a product fact. Lenny's read on the structural difference between B2B and consumer subscriptions is the right anchor here. B2B SaaS businesses, the argument goes, "increase Net Revenue Retention (NRR) by growing the value of retained accounts to offset churn" through upsell, while consumer subscriptions usually have a single SKU and limited expansion lever (Lenny's Newsletter , The Subscription Value Loop). The implication is that B2B SaaS founders who do not show expansion are leaving the structural advantage on the table.

For consumer or prosumer SaaS, the bar is different and harsher. Lenny's data point is the one to internalize: out of hundreds of thousands of consumer subscription apps, "fewer than 50 have ever reached $1B+ valuations, and fewer than 10 are publicly traded companies with $10B+ market caps" (Lenny's Newsletter). Consumer retention has to be exceptional to clear the same Series A bar that average B2B retention will clear, because the expansion lever is structurally weaker.

PitchBook-NVCA's Q1 2026 framing is the one to quote in the deck if you are consumer or prosumer: "In consumer or 'prosumer' businesses, retention is the key metric and the best proxy for product-market fit" (PitchBook-NVCA Venture Monitor , Q1 2026). For consumer SaaS the entire valuation lens collapses onto the retention curve.

User retention is not dollar retention, and the deck must say so

The single most common SaaS retention deck mistake in 2026 is conflating user retention and dollar retention. Pendo's frequently cited 2025 benchmark, that "software products lose 70% of users after three months", is a user retention statistic; it is not what a VC underwrites a Series A on. Dollar retention is the metric on the slide that matters.

The distinction in one paragraph: user retention measures whether a human keeps using the product, dollar retention measures whether the account keeps paying for and expanding the product. For a $90K-ACV mid-market deal, you can lose six of ten daily active users at the buyer account and still have 110% NRR on that logo because the seats that remain are sticky and the org bought a second SKU. The deck slide that gets this right shows both curves, labeled, with the dollar curve as the headline.

The opinionated call here: lead the retention slide with NRR and GRR, then put the user-retention curve below as evidence of stickiness. The reverse ordering, user retention up top and dollar retention as a footnote, signals that the founder is benchmarking against blog posts and not against VC underwriting standards.

YC's hard rule on retention curves is the right anchor for the user-retention curve below the NRR line. David Lieb's framing is unambiguous: "the only thing that matters is whether the curve flattens"; if it does not flatten, "you haven't made something people want" (YC Startup Library , How To Improve Cohort Retention). A flat 25% curve beats a 60% curve that keeps sliding to 8%. The asymptote, not the day-one number, is the signal.

YC's adjacent rule on how to define an active user is the one that prevents you from showing a flatter curve than you actually have. Lieb's instruction: never define an active user by passive signals like "opening the app" or by billing alone; for paid products an active user must be "paying AND an active user" (YC Startup Library). Users typically stop using the product before they cancel, so a billing-only retention curve is a lagging indicator that will surprise you at renewal.

In a Series A retention review, a 25% flat user-retention curve paired with 115% NRR will outprice a 45% declining user-retention curve paired with 98% NRR, every time. The asymptote and the dollars are the only two numbers that matter.

The minimum viable retention signature for Series A

A Series A retention signature in H1 2026 is the combination of four numbers, presented together, that lets a VC underwrite the round in under five minutes. None of the top three SERP retention pieces combine these into a single rubric, which is why founders keep arguing one number at a time and losing.

The minimum viable retention signature, by ACV tier:

What this rubric says in plain English: you need all four numbers to be above the floor for your tier. A 130% NRR for an SMB SaaS with 35% logo churn and 60% GRR is a leaky bucket with a strong upsell motion sitting on top, not a Series A. Sophisticated underwriters in 2026 spot the leaky-bucket pattern in 30 seconds.

The contrarian call inside this rubric is logo churn. Most retention guides skip it because dollar metrics flatter the story. But logo churn is what tells the VC whether your TAM math works. If you need 5,000 enterprise logos to hit a $500M revenue plan and your logo churn is 12% annually at $100K ACV, the math breaks before you get there. Logo churn is the metric that says whether the business model exists at scale, regardless of how good the dollar retention looks today.

First Round's PMF playbook reframes how to think about the engine behind these numbers. The argument, which is the right one to absorb, is that "in SaaS, focusing on making your customer successful is a retention strategy, not a cost center" (First Round Review , Levels of PMF). The CS org is the gross-retention engine, not a margin drag. Underfund it and the signature in the table above is unreachable.

How to calculate net revenue retention without lying to yourself

The mechanical definition of net revenue retention is the percentage of recurring revenue retained from a starting cohort of customers, including expansion, over a defined period (almost always 12 months). The formula:

NRR = (Starting ARR + Expansion ARR - Downgrade ARR - Churn ARR) / Starting ARR

That formula is simple. The way founders mislead themselves with it is not.

The three most common NRR self-deception patterns in 2026:

• Including new-logo ARR in the numerator. New logos do not belong in NRR. NRR measures the cohort that existed at t=0. If you mix in new logos to inflate the number, you are calculating something else and a VC will catch it inside one diligence call.
• Defining the cohort by customers who were "active" rather than paying. A free-trial conversion or a paused subscription does not belong in the starting cohort. The starting ARR is the contracted, billing ARR on the cohort start date.
• Reporting NRR on a too-recent cohort. A six-month NRR can hide downgrade waves that happen at the 9-month or 12-month mark. The right NRR for a Series A deck is 12-month trailing on a cohort that has had time to renew.

The opinion to hold: report NRR on a 12-month trailing cohort with no new logos in the numerator, or do not put it on the slide. Anything else is a footnote that diligence will turn into a credibility problem.

The companion metric to put next to NRR is net negative churn in dollars. First Round's "Extreme PMF" profile cites an "18-month streak of net negative churn and renewals higher than plan" as the Level 4 signature (First Round Review). Net negative churn means the dollars expanded from existing accounts exceeded the dollars lost to churn and downgrades, every month, in absolute terms. A founder who can show eighteen consecutive months of net negative churn is presenting the strongest expansion signal a Series A underwriter will see in 2026.

YC's growth context is the right pair for the NRR section in your deck. Tom Blomfield's framing of growth rates is the calibration: 15% MoM user growth is "good" (about 5x per year), 10% is "okay" (about 3x per year), 5% or lower is "unlikely for breakout success", and consumer retention must "flatten off somewhere" to compound (YC Startup Library , Consumer Startup Metrics). NRR without growth is a maintained business; growth without NRR is a leaky bucket. Both numbers belong on the slide.

Why this matters for your raise

Retention is the metric a VC underwrites a Series A on in H1 2026, full stop. Carta's read that the market has settled into a "new type of normal" as it "adjusts to the omnipresence of AI" (Carta , State of Private Markets: 2025 in Review) is the macro context for why this is true. AI hype reset the bar for proof, and proof at the application layer means retention and expansion economics, not top-of-funnel growth.

The fundraising bridge: the retention signature in this report is what your Series A deck needs to show. If your numbers clear the floors in the minimum-viable-retention-signature table, you have the data to pitch confidently and to negotiate price; if they do not, the work in the next two quarters is to fix the leakiest of the four numbers before you go out, because no amount of pitch craft will compensate for a 92% NRR in a 110%+ market. If you are at the stage where the deck and the investor list are the next step, the Causo platform helps founders match retention numbers to VCs whose published theses underwrite at the ACV tier you are in, so the conversations land in front of partners who already believe the math.

FAQ

What is a good NRR for a SaaS company in 2026? For sub-$1B-ARR enterprise platforms, a16z puts best-in-class NRR above 120%; broader enterprise software clusters at 110–120% per ICONIQ. At seed and Series A, 108% net revenue retention is the "Strong PMF" line from First Round's Levels framework, and anything under 100% will not clear a 2026 underwriting bar.

What is the average SaaS churn rate? There is no single average that means anything across ACV tiers. The defensible 2026 anchor is First Round's "Strong PMF" profile at 8% regretted annual churn, with sub-PMF SaaS sitting at 10–20%. Enterprise platforms with mission-critical status clear 95–97% gross dollar retention per a16z, meaning 3–5% annual logo or dollar churn.

What NRR do Series A investors want in 2026? The minimum credible NRR for a 2026 Series A SaaS pitch is 108% (First Round's Level 3 PMF), with 120%+ as the "Extreme PMF" bar. Below 100% NRR you are net-shrinking the installed base, which is disqualifying in a market where Carta says capital is being underwritten on proof, not narrative.

What is the difference between gross revenue retention and net revenue retention? Gross revenue retention (GRR) measures dollars kept from the starting cohort, capped at 100%, after churn and downgrades. Net revenue retention (NRR) adds expansion (upsell, cross-sell, seat growth) on top, so it can exceed 100%. GRR proves the product is sticky; NRR proves the account base compounds.

What is a good gross retention rate for a SaaS company? For enterprise platforms, a16z's bar is "above 95% and often above 97%" gross dollar retention. Mid-market sits 5–10 points below that. SMB SaaS under $25K ACV rarely clears 80% GRR and is not penalized for it at Series A as long as NRR and expansion mechanics compensate.

Related on the hub

• Traction metrics for VCs in 2026: what IC memos screen for — for when the playbook turns into a raise.
• Customer expansion at seed 2026: double ARR per customer — Related retention guide.
• H1 2026 B2B SaaS GTM Benchmark Report — Related gtm business model guide.
• The H1 2026 SaaS pricing report — Related pricing guide.