The 2026 SaaS Benchmarks Report: ARR, NRR, CAC, and 14 Metrics That Actually Matter

Every SaaS board meeting in 2026 starts the same way. Someone pulls up a slide with ARR, net revenue retention, and CAC payback. The numbers are compared against "industry benchmarks" that come from a 2023 report, a 2021 Bessemer State of the Cloud deck, or a KeyBanc survey that sampled 150 companies — half of which no longer exist.

The board nods. The metrics look reasonable. The company is "performing inline with benchmarks." Six months later, the company misses its Series B, cuts 30% of the team, or quietly starts exploring an acqui-hire.

The problem is not that SaaS leaders are tracking the wrong metrics. It is that they are benchmarking against the wrong numbers, in the wrong context, at the wrong stage. A 120% NRR is elite at seed. It is table stakes at $50M ARR. A 14-month CAC payback is fine if your gross margins are 82%. It is a death sentence if they are 64%.

This report covers 14 metrics that actually predict whether a SaaS company survives the next 18 months. Every benchmark is stage-specific. Every number reflects 2026 market conditions — where AI is compressing seats, capital is more expensive than 2021 but cheaper than 2023, and the bar for "good" has shifted in ways most operators have not internalized.

The 14 Metrics, Ranked by How Much They Actually Matter

Before we get into individual breakdowns, here is the hierarchy. Not every metric matters equally, and the ranking shifts by stage. But if you are a seed-to-Series B company and you can only obsess over three metrics, make them these:

1. Burn multiple — how efficiently you convert cash into ARR growth
2. Net revenue retention (NRR/NDR) — whether your existing customers are expanding or dying
3. Gross margin — whether your unit economics can ever work

Everything else is downstream of those three. CAC payback matters, but it is a function of gross margin and sales efficiency. LTV:CAC matters, but it is a function of NRR and churn. Rule of 40 matters, but only at scale.

Let me be blunt about what does not matter as much as people think: logo churn rate in isolation, magic number below $10M ARR, and ACV without context on sales motion.

ARR Growth Rate: The Number Everyone Knows and Nobody Contextualizes

ARR growth rate is the most commonly cited SaaS metric and the most commonly misused. A company growing 100% year-over-year at $2M ARR is in a fundamentally different position than a company growing 100% at $50M ARR. The first is expected. The second is exceptional. Most benchmarks do not separate these.

Here are the 2026 medians, by stage:

The story these numbers tell: growth expectations have compressed by 10–15 percentage points at every stage compared to 2021 benchmarks. A Series A company growing 100% was median in 2021. It is still median in 2026, but the top-quartile threshold has dropped because fewer companies are hitting the 200%+ rates that were common when free money fueled land grabs.

The AI-native exception is real but narrower than people think. Cursor hit $1B ARR growing at 400%+. But Cursor is not a benchmark — it is an outlier that built a generational product category. The median AI-native SaaS company at Series A is growing at 120%, which is better than non-AI peers but not the order-of-magnitude difference the hype suggests.

The number that matters more than growth rate: growth rate relative to burn. A company growing 100% while burning $3M/month is in a worse position than a company growing 60% while burning $400K/month. Which brings us to burn multiple.

Burn Multiple: The Metric That Predicts Survival

Burn multiple is net burn divided by net new ARR. If you burned $6M last year and added $3M in net new ARR, your burn multiple is 2.0x. It tells you how many dollars you are spending to generate each dollar of new ARR.

David Sacks popularized this metric, and for good reason: it is the single best predictor of whether a company will reach its next funding milestone or run out of cash.

The 2026 shift: burn multiples have tightened significantly. In 2021, a Series A company could raise with a 4.0x burn multiple because investors were underwriting growth at any cost. In 2026, anything above 2.5x at Series A triggers hard questions about unit economics, go-to-market efficiency, and whether the company has product-market fit or is buying revenue with venture dollars.

AI is affecting this metric in two directions. AI-native companies often have lower burn multiples because their products grow through word of mouth and product-led adoption — Notion's AI features, for example, drive expansion with zero incremental sales cost. But AI infrastructure costs can inflate burn if a company is running expensive model inference without usage-based pricing to offset it. The median AI SaaS company spends 18–24% of revenue on model inference costs, which is a new line item that did not exist two years ago.

> If your burn multiple is above 2.0x at Series B, you do not have a growth problem. You have an efficiency problem. Fix your go-to-market before you fix your growth rate.

Net Revenue Retention: The Only Metric That Compounds

NRR (sometimes called NDR — net dollar retention) measures how much revenue your existing customer cohort generates over time, including expansion, contraction, and churn. An NRR of 120% means that a cohort of customers who paid you $1M a year ago is now paying you $1.2M — without any new customers.

This is the single most powerful metric in SaaS because it compounds. A company with 130% NRR doubles its existing revenue base every 2.5 years without acquiring a single new customer. A company with 95% NRR loses half its revenue base every 14 years even if it keeps every logo.

The 2026 reality: NRR has declined 5–8 points across the board since 2022. Budget scrutiny is real. Procurement teams are killing expansions. Companies like Datadog and Snowflake, which posted 130%+ NRR during the cloud migration boom, are now reporting 115–120% because customers are optimizing usage rather than expanding indefinitely.

Why NRR Peaks at Series B

This is something most benchmarking reports miss. NRR typically peaks around Series B because that is when a company has enough product surface area for meaningful expansion but has not yet penetrated its TAM deeply enough for saturation to set in. At scale, NRR naturally compresses because your largest customers are already on your highest-tier plans — there is less room to expand.

HubSpot is the canonical example. At $30M ARR, their NRR was above 125% because customers were adopting additional hubs (Marketing, Sales, Service). At $2B+ ARR, NRR has settled around 108% because the expansion curve flattens at the enterprise tier.

AI's Impact on NRR

AI is creating a paradox for NRR. Products that add AI capabilities see higher expansion — Figma's AI features drove meaningful plan upgrades through 2025. But AI is simultaneously enabling competitors to offer similar capabilities at lower price points, increasing contraction pressure. The net effect depends on whether your AI features are proprietary or commoditized.

Cursor's NRR is estimated to be above 150%, which is extraordinary. But Cursor's expansion comes from developers converting from free to paid and from individual to team plans — not from AI features alone. The AI capability is the wedge, but the expansion revenue comes from organizational adoption. That distinction matters.

Gross Margin: The Metric AI Is Quietly Destroying

SaaS gross margins are supposed to be 75–85%. This is one of the fundamental structural advantages of software: you build it once, and the marginal cost of serving an additional customer is near zero. That premise is under attack.

AI-native SaaS companies are running significantly lower gross margins because model inference is not free. Every time a user sends a prompt, the company pays for compute. This makes AI SaaS structurally more similar to marketplace or fintech businesses than traditional software.

The good news: AI-native gross margins are improving as model costs decline. The cost of running GPT-4-class inference has dropped roughly 90% since early 2024. Companies that were paying $0.06 per output token are now paying fractions of a cent. Anthropic's Claude pricing has followed a similar curve. This is allowing AI-native companies to claw back margin.

The bad news: usage is scaling faster than cost reductions. A company whose per-query cost dropped 80% but whose query volume increased 500% is still spending more on inference than it was a year ago.

The benchmark that matters: gross margin after AI inference costs. If you strip out inference, your gross margins look like traditional SaaS. If you include inference, they do not. Boards need to see both numbers. Most are only seeing one.

CAC Payback Period: Overrated but Not Irrelevant

CAC payback measures how many months it takes to recover the fully loaded cost of acquiring a customer. It is one of the most cited SaaS metrics and one of the most misleading at early stages.

Here is why: CAC payback is a function of three underlying metrics — customer acquisition cost, average revenue per account, and gross margin. If any of those numbers is unstable (and at seed and Series A, all three are), your CAC payback calculation is noise, not signal.

The two numbers that jump out: PLG companies at seed have radically faster CAC payback because their acquisition cost is near zero — users sign up, try the product, and convert without touching a sales rep. Notion, Figma, and Canva all had sub-6-month CAC payback at seed because their growth was organic. Sales-led companies at seed can have CAC payback above 24 months and still be healthy if their ACV is high enough and their NRR is strong.

The 2026 opinion: CAC payback is overrated below $10M ARR and underrated above $50M ARR. Below $10M, your CAC is dominated by founder-led sales and one-off experiments that do not represent steady-state unit economics. Above $50M, CAC payback is a genuine indicator of sales efficiency because the motion is repeatable.

LTV:CAC — The Vanity Metric That Should Be Retired

I will die on this hill: LTV:CAC is the most abused metric in SaaS.

The standard guidance is that LTV:CAC should be above 3.0x. The problem is that LTV (lifetime value) is a theoretical number based on assumptions about future retention, future ARPU, and future gross margins — none of which are knowable at early stages. A company with 12 months of data calculating a "lifetime" value is doing creative fiction, not analysis.

Worse, LTV:CAC is trivially gameable. Extend your assumed customer lifetime from 5 years to 7 years and your LTV:CAC improves by 40% without changing anything real about your business. Lower your blended CAC by including organic signups in the denominator and the ratio looks even better.

What to use instead: gross-margin-adjusted CAC payback plus NRR. If your CAC payback is under 18 months and your NRR is above 110%, you do not need LTV:CAC to tell you the unit economics work. If either number is bad, LTV:CAC will not save you — it will just help you construct a convincing board slide while the business deteriorates.

For boards that insist on seeing it: median LTV:CAC in 2026 is 3.2x at Series B and 4.5x at growth stage. Top quartile is 5.0x+. But I would trade any LTV:CAC calculation for an actual cohort retention curve with 24 months of data.

Gross Revenue Churn and Logo Churn: Context Is Everything

Gross revenue churn is the percentage of ARR lost from existing customers before any expansion. Logo churn is the percentage of customers lost. These are related but not interchangeable, and the distinction matters.

A company can have 3% logo churn and 8% gross revenue churn if the customers leaving are disproportionately large. Conversely, a company can have 12% logo churn and 4% gross revenue churn if the customers leaving are all on the $29/month plan while the enterprise accounts stay.

Healthy ranges in 2026: - Gross revenue churn: 8–12% annually (median), under 6% (top quartile) - Logo churn: 10–18% annually for SMB-focused, 5–10% for mid-market, under 5% for enterprise

The AI Churn Cliff

There is a new phenomenon in 2026 that is not captured in legacy benchmarking: AI-driven churn spikes. Companies that sold workflow tools — project management, document editing, basic analytics — are seeing churn increase as AI-native alternatives emerge. A $500/month Asana contract gets replaced by a team using Notion AI and Linear. A $2,000/month BI tool gets replaced by a natural-language analytics layer from a startup that did not exist 18 months ago.

This churn is not gradual. It arrives in clusters — usually when a single champion inside the customer org discovers an AI alternative and pulls the team over within a single renewal cycle. The result is that monthly churn can spike from 0.8% to 2.5% in a quarter with no warning in the leading indicators.

If you are running a SaaS company in a workflow category, your churn risk model needs to account for this. The traditional predictor of churn — declining usage — still applies. But you also need to track whether your customers are experimenting with AI alternatives in adjacent categories.

Rule of 40: Still Relevant, Still Misunderstood

The Rule of 40 states that a SaaS company's revenue growth rate plus its profit margin (typically EBITDA or FCF margin) should exceed 40%. It balances growth against profitability and is the closest thing SaaS has to a single summary statistic.

In 2026, the Rule of 40 is still the primary framework institutional investors use to evaluate SaaS companies at scale. But it has two significant flaws:

Flaw 1: It treats growth and profitability as interchangeable. A company growing 50% with -10% margins (Rule of 40 = 40) and a company growing 10% with 30% margins (Rule of 40 = 40) are valued identically by this framework. In practice, the first company is worth 3–4x more because growth is harder to manufacture than profitability.

Flaw 2: It only applies above $30M ARR. Below that, the Rule of 40 is meaningless because margins swing wildly with single hires and one-time costs. At $5M ARR, adding one enterprise sales rep swings your margins by 15 points. The Rule of 40 was designed for evaluating scaled SaaS businesses, and it should only be used there.

2026 medians for companies above $50M ARR: median Rule of 40 score is 32 (below the threshold), top quartile is 55+, and the companies actually hitting 40+ are disproportionately product-led or AI-native. Datadog runs above 55. CrowdStrike is above 50. The median Series B company is nowhere close.

Magic Number: The Sales Efficiency Metric That Needs a Reboot

The magic number measures sales efficiency: net new ARR divided by sales and marketing spend in the prior period. A magic number of 1.0 means you are generating $1 of net new ARR for every $1 spent on sales and marketing.

The traditional interpretation: above 0.75 means you should invest more in sales. Below 0.5 means something is broken.

The 2026 problem: the magic number does not account for PLG revenue. If 40% of your new ARR comes from self-serve signups with zero sales involvement, your magic number is inflated relative to your actual sales efficiency. Conversely, if you are running a pure enterprise motion, your magic number will look low compared to blended benchmarks that include PLG companies.

Median magic number in 2026: 0.65 (Series A through Growth). Top quartile: 0.9+. But these numbers are increasingly meaningless without segmenting by go-to-market motion.

Expansion Revenue Percentage: The Underrated Growth Lever

Expansion revenue as a percentage of total new ARR is, in my view, the most underrated metric in SaaS. It measures how much of your growth comes from existing customers versus new logos.

At scale, the best SaaS companies generate 30–40% of their new ARR from expansion. Snowflake generates over 35% from consumption expansion. Datadog consistently generates 30%+ from customers adopting additional products. This is structurally superior to new logo acquisition because expansion revenue has near-zero CAC and significantly higher conversion rates.

The 2026 benchmark: - Below $10M ARR: expansion should be 15–20% of new ARR - $10M–$50M ARR: 25–35% - Above $50M ARR: 30–45%

If expansion is below 15% at any stage, your product does not have enough surface area or your customers are not getting enough value to want more. Either way, it is a problem.

Quick Ratio: The Metric That Catches Leaky Buckets

SaaS quick ratio is (new MRR + expansion MRR) / (churned MRR + contraction MRR). It tells you how much revenue you are adding for every dollar you lose. A quick ratio of 4.0 means you add $4 for every $1 that churns.

Healthy quick ratios in 2026: - Seed: 4.0+ (you should not be losing much revenue yet) - Series A: 3.5+ - Series B: 3.0+ - Growth and Scale: 2.5+

The quick ratio is useful because it surfaces a problem that aggregate ARR growth can hide: if you are growing 80% but your quick ratio is 1.8, you are growing by brute-forcing acquisition while losing customers at an alarming rate. That growth is not durable.

ACV: The Metric That Defines Your Go-to-Market

Average contract value determines everything about your business model — sales cycle length, team structure, support requirements, and viable growth rate.

The 2026 trend: ACVs are compressing in categories where AI alternatives exist. A company that sold a $50K/year analytics platform is competing against AI-native tools at $12K/year. The response from incumbents has been to add AI features to justify pricing, but this only works if the AI features are meaningfully differentiated. "We added an AI chatbot" is not differentiation — it is the 2026 equivalent of "we have a mobile app."

Meanwhile, ACVs are expanding in categories where AI increases the value delivered. Security platforms, developer tools, and data infrastructure companies are seeing ACV increases of 20–30% as AI capabilities make the products more valuable to larger teams.

What AI Is Actually Changing About SaaS Benchmarks

The narrative that "AI changes everything about SaaS" is partially true and mostly unhelpful. Here is what is specifically, measurably changing:

1. Gross margins are structurally lower for AI-native companies. This is real, significant, and permanent — though the gap is narrowing as inference costs decline. Boards need to reset their gross margin expectations from 78% to 65–70% for AI-native SaaS and adjust valuation frameworks accordingly.

2. NRR is bifurcating. AI-native products with strong usage loops (Cursor, Jasper, Midjourney's enterprise offering) are posting NRR above 140%. Traditional SaaS products facing AI substitution are seeing NRR compress below 105%. The median is stable but the distribution has widened dramatically.

3. CAC is dropping for AI-native products. Products that deliver immediate, visible value — generate a report, write code, create an image — have organic virality that traditional SaaS did not. This shows up as lower CAC, faster payback, and higher magic numbers. The median AI-native startup at Series A has a CAC payback of 8 months versus 16 months for non-AI peers.

4. Burn multiples are better because teams are smaller. An AI-native company with 15 engineers can ship product that previously required 60. Cursor reached $1B ARR with roughly 70 employees. This structural efficiency advantage means burn multiples are lower at every stage for AI-native companies — not because they grow faster (though some do), but because they spend less.

5. The Rule of 40 is easier to hit with AI-native cost structures. Smaller teams, lower CAC, and product-led growth mean AI-native companies can grow fast while maintaining positive margins. This makes the Rule of 40 less of a stretch and more of a baseline expectation for the best AI SaaS companies.

The Metrics That Are Overrated vs. Underrated in 2026

Overrated

• LTV:CAC — too easily gamed, too dependent on assumptions, and not useful at early stages
• Logo churn in isolation — meaningless without revenue-weighting and segmentation
• Magic number below $10M ARR — your sample size is too small and your motion is too inconsistent
• Monthly active users — engagement without monetization is a consumer metric, not a SaaS metric

Underrated

• Burn multiple — the single best predictor of fundraising outcomes and survival
• Expansion revenue as a percentage of new ARR — reveals whether your product has genuine depth or is a one-trick sale
• Gross margin after inference costs — the metric that separates AI SaaS companies that will scale from those that will burn through cash
• Payback period segmented by channel — knowing that your blended CAC payback is 14 months is less useful than knowing that organic is 4 months, paid is 22 months, and outbound is 19 months
• Quick ratio — catches the companies that are growing on paper but leaking revenue faster than they realize

How to Use These Benchmarks Without Lying to Yourself

The most dangerous thing a founder or operator can do with benchmarks is cherry-pick the stage, the metric, and the comparison set that makes their numbers look best. Every company is "top quartile" at something if you choose the right frame.

Here is a more honest framework:

Step 1: Identify your three weakest metrics from the tables above. Not your best — your worst. The metrics where you are in the bottom quartile for your stage.

Step 2: Determine whether those weak metrics are correlated. If your gross margin is low AND your burn multiple is high AND your NRR is below median, you do not have three problems. You have one problem: the business does not have product-market fit at a price point that works.

Step 3: Pressure-test your strong metrics. If your NRR is 140% but it is driven by one enterprise customer's expansion, it is not 140% NRR — it is one deal that happened to renew big. Segment by cohort, segment by customer size, and see whether the strength holds.

Step 4: Compare against the right stage and the right motion. A PLG company at $5M ARR should not benchmark against enterprise SaaS at $5M ARR. The metrics are structurally different because the business models are structurally different.

The goal of benchmarking is not to prove you are doing well. It is to identify where you are weak before your investors, your customers, or your runway forces the conversation.

The 18-Month Test

Every metric in this report answers the same underlying question: will this company be in a stronger position 18 months from now than it is today?

If your NRR is above 115%, your burn multiple is below 2.0x, and your gross margins are above 70%, the answer is almost certainly yes. The compounding math is in your favor. Your existing customers are growing, your cash is being converted efficiently into new ARR, and your unit economics support the growth.

If any of those three metrics is in the bottom quartile for your stage, the math is working against you. Not slowly — quickly. An NRR below 100% means your revenue base is decaying. A burn multiple above 3.0x means your cash runway is shorter than you think. Gross margins below 60% mean you need dramatically more revenue to reach profitability than your model assumes.

The companies that will matter in 2028 are the ones reading these benchmarks honestly today — not to validate what they have built, but to identify what they need to fix before the market fixes it for them.