The Only 9 SaaS Metrics That Matter in 2026 (Benchmarked Against 1,300+ Companies)

The Only 9 SaaS Metrics That Matter in 2026 (Benchmarked Against 1,300+ Companies)

Median gross revenue retention in the 342-company Aleph × Benchmarkit panel fell from 88% to 84% in calendar 2025 (Aleph × Benchmarkit, 2026 SaaS & AI Performance Benchmarks). Four points, in one year. Yet the saas metrics that matter in most board decks are still scored against thresholds David Sacks published in October 2021, back when capital was effectively free. Those bars aren't slightly stale. They're calibrated to a market that stopped existing. What follows is nine metrics, every figure traced to a named 2026 report with its sample size, and six metrics we think you should drop.

Key Takeaways

• Usage-based SaaS posts 108% median NRR versus 98% for seat-based, a 10-point gap (Benchmarkit 2026, n=230).
• Median gross revenue retention fell from 88% to 84% year over year (Benchmarkit 2026).
• Bootstrapped SaaS growth slowed from 20% to 15% median at $3–20M ARR (SaaS Capital 2026, n>1,000).
• Nine metrics matter in 2026. Six commonly tracked ones don't, and we name them.

What Do the 2026 SaaS Benchmarks Actually Say?

The 2026-edition SaaS metrics benchmarks, all reporting full-year 2025 actuals, describe a market that got harder on growth and slightly better on efficiency. Median net revenue retention sits at 102% and median gross revenue retention at 84%, both down year over year, while median CAC payback improved from 18 to 16 months (Aleph × Benchmarkit, CY-2025; NRR n=230, CAC payback n=198, GRR n not disclosed).

Here's the whole thing in one table: the saas metrics cheat sheet people keep searching for. Screenshot it, cite it, argue with it. Every figure is CY-2025 actuals from a 2026-published benchmark, dollars USD as published. Cuts a source doesn't publish read n/a, never an estimate. One labelling note: SaaS Capital calls its figures "2026 medians" where Aleph and Benchmarkit label the same reporting period CY-2025. We use CY-2025 throughout for consistency; the underlying data year is the same.

Read the median as "half the market is below this," not as a passing grade. The nine group into three buckets:

1. Retention: NRR, GRR, logo retention
2. Efficiency: CAC payback, burn multiple, ARR per employee
3. Growth quality: growth rate, gross margin, Rule of 40

Half the industry is being measured against a bar that was written when capital was free.

How We Built This Benchmark (Methodology and Sample Sizes)

We did not run a survey. What this is: a normalized meta-benchmark assembled from named, published 2026 reports. Every figure belongs to one of them and is attributed by name. The one section drawing on our own work, on how these numbers break in implementation, carries no statistics at all, because we haven't measured any.

The arithmetic behind "1,300+": 342 companies in the Benchmarkit panel plus more than 1,000 in the SaaS Capital panel is more than 1,342. We count the SaaS Capital panel once, since two of its reports appear above off the same respondent base.

On definitions we normalized to the SaaS Metrics Standards Board: NRR measures expansion, contraction and churn on a fixed cohort and excludes new logos entirely. That's exactly what the SQL further down enforces.

Deliberately left out: the "$129,724 revenue per employee" figure and "only 11–30% of SaaS companies clear the Rule of 40." Both trace to an aggregator's own analysis, not a primary report. Not one page in the top ten of this search result explains where its numbers came from.

One disclosure, because this post would be hypocritical without it. Our two sources disagree with each other on ARR per employee. Benchmarkit's summary page headlines $175K, up 17%; Aleph's per-metric page reports $193K, up 29%, off the same panel. We use $193K because that page discloses its sample (n=96) and publishes the full distribution, while the summary headline discloses neither. Reasonable people could take the other number. What nobody should do is quote either one without saying which page it came from.

Retention: NRR, GRR, and Logo Churn

Net revenue retention (NRR)

Net revenue retention, also called net dollar retention, measures what a fixed customer cohort pays you twelve months later, counting expansion, contraction and churn but excluding new logos. The CY-2025 median is 102%, with the 25th percentile at 92%, down from 95% in 2024 (Aleph × Benchmarkit, n=230).

Formula: NRR = (starting MRR + expansion − contraction − churn) ÷ starting MRR. A cohort starting at $100K that expanded $12K, contracted $4K and churned $6K lands at 102%.

Now the finding worth the whole post. Split that panel by pricing model and usage-based companies post a median NRR of 108% against 98% for seat-based, a 10-point gap (Aleph × Benchmarkit).

Aleph × Benchmarkit, 2026 SaaS & AI Performance Benchmarks, n=230 reporting NRR. Below the 100% line, the existing customer base shrinks without new logos.

Sit with the 98%. The median seat-based SaaS company is below breakeven on its own base: stop acquiring tomorrow and it shrinks. On the metric everyone treats as a customer-love score, how you charge beats what you built. If your NRR is stuck in the high 90s, ask whether anything in your pricing grows when your customer does.

Gross revenue retention (GRR)

Gross revenue retention is NRR with the expansion credit removed: contraction and churn only, floored at each customer's starting revenue. The median fell from 88% to 84% and the 75th percentile from 95% to 91% (Aleph × Benchmarkit, CY-2025).

Formula: GRR = (starting MRR − contraction − churn) ÷ starting MRR. It can never exceed 100%.

NRR hides churn behind expansion; GRR can't. A company at 105% NRR and 84% GRR isn't retaining customers, it's selling hard into the ones who stay.

Logo retention and customer churn

Logo retention counts customers, not dollars, so it diverges from GRR whenever churn concentrates in small accounts or in large ones. We could not trace a logo-churn median to a primary 2026-edition report, so we're not printing one.

What we will say: "5–7% annual churn is healthy" is 2019-era folk wisdom that a top-ten competitor still publishes with no citation, and it's meaningless without an ACV band.

How Efficient Is Your SaaS, Really? CAC Payback, Burn Multiple, ARR per Employee

CAC payback period

CAC payback period is the months of gross-margin-adjusted revenue needed to repay the cost of acquiring a customer. The CY-2025 median is 16 months, improved from 18 in 2024; the top quartile repays in 6 months or less and the bottom quartile takes 24 or more (Aleph × Benchmarkit, n=198).

Formula: CAC payback = CAC ÷ (new MRR × gross margin).

Aleph's argument that CAC payback should replace LTV:CAC as your primary efficiency metric is right, for a plain arithmetic reason: payback uses two numbers you know, LTV:CAC multiplies three estimates. The same panel reports a median magic number of 1.37, or $1.37 of new ARR per $1 of sales and marketing. The split underneath is sharper: a new customer costs $1.63 per dollar of new ARR against $0.80 for expansion (Aleph × Benchmarkit, NRR benchmarks). Expansion is half price.

Burn multiple

Burn multiple, David Sacks' own invention, is net burn divided by net new ARR: how many dollars you consume to manufacture one dollar of recurring revenue. Under 1 is amazing and under 2 is still quite good, per his 2021 framing.

Here's an honest gap: none of the 2026-edition reports we could trace publishes a burn-multiple distribution, so we're not inventing one. If someone quotes you a current median burn multiple, ask which report, which data year, what n. Capital efficiency is the most-asserted, least-sourced corner of this category.

ARR per employee

ARR per employee is annual recurring revenue divided by full-time headcount, and it's the metric that moved most in 2025: median $193K, up 29% from $150K, with the top quartile near $279K and the bottom at $126K (Aleph × Benchmarkit, n=96). Usage-based companies lead at $291K.

A 29% jump in one year isn't a productivity miracle, it's partly a hiring freeze. Tooling moves this one: the tools small teams use to stay small and automating the work instead of hiring for it both land in this ratio. Companies growing above 50% post $235K per employee against $136K for the 31–50% cohort.

Growth Quality: Growth Rate, Gross Margin, and the Rule of 40

Growth rate

Median year-over-year growth for bootstrapped B2B SaaS at $3–20M ARR fell from 20% to 15%, while NRR held at 103% and GRR at 91% (SaaS Capital, 2026 Benchmarking Metrics for Bootstrapped SaaS Companies, n>1,000). The 90th percentile fell too, from 51% to 42.3%.

SaaS Capital, 2026 Benchmarking Metrics for Bootstrapped SaaS Companies, n>1,000 private B2B SaaS. Retention held flat over the same period, which locates the slowdown in acquisition rather than churn.

Growth fell. Retention didn't. That means the slowdown is an acquisition problem, and most boards are pointing at the wrong thing.

Gross margin

Software gross margin medians sit at 80%, with the top quartile at 86% or better, the bottom quartile at 50%, and blended total-revenue margin at 76% (Aleph × Benchmarkit, n=228–232).

Everyone expects AI inference costs to be eating SaaS margins. The report found no compression: software gross margin held at 79–81% across all four years of benchmark data. Its own advice is to instrument AI costs now, which starts with knowing what LLM inference actually costs per token inside your product.

Rule of 40

The Rule of 40 says growth rate plus profit margin should clear 40. The median rose from 15% to 25% in CY-2025, and the top quartile reaches 43% (Aleph × Benchmarkit). Formula: growth % + FCF or EBITDA margin % ≥ 40.

The pricing thread ties off here: hybrid subscription-plus-usage models lead the cohort at that 43% mark. Note what the rising median means, though. Companies got closer to 40 by cutting, not growing, which is the least satisfying way to pass a test.

Do These Benchmarks Change by ARR Band?

Yes, substantially, and this is where single-number guides fail hardest. NRR ranges from 94% at sub-$5M ARR to 103% above $100M, and ARR per employee peaks in the middle bands rather than at the top.

Sources by row: NRR, ARR per employee and gross margin from Aleph × Benchmarkit 2026 (CY-2025); GRR and growth rate from SaaS Capital 2026 (n>1,000). An n/a means no primary source published that cut. Read your own column: at $8M ARR growing 15% with 103% NRR, you're precisely average.

Why Your MRR Dashboard Disagrees With Your P&L

The formulas above are the easy part. Computing them correctly against real billing data is not, and every guide in this SERP skips it.

Start with published proof that the definitions themselves aren't settled. The SaaS Metrics Standards Board, the body that exists to standardize exactly this, records the argument as still open: "there is debate on the inclusion of usage-based pricing revenue greater than the committment [sic] level (overages) in the NRR calculation." It does settle how to compute it: "the cohort method is the most accurate way to measure NRR and is the preferred approach in most situations." Aleph names the worst single error outright: "the most common error is letting new-logo revenue leak into the expansion term," which in their words inflates NRR and hides a retention problem (Aleph × Benchmarkit).

So comparing your NRR to a benchmark median assumes both were computed the same way, and the standards body says they probably weren't. Here are the five failure modes we design against when we instrument this at Techsy, stated as engineering practice rather than survey findings:

1. New-logo revenue leaking into the expansion term. If your query filters on a date range rather than freezing a customer cohort, every new customer inflates your retention.
2. Contraction and churn collapsed into one bucket. GRR needs them separated, and merged you can't recover the split without a rerun.
3. ARR snapshotted at different moments by billing and by the dashboard, so a mid-cycle upgrade booked at full contract value in one system and prorated in the other never reconciles.
4. Monthly and annual contracts normalized differently. One system divides the annual value by 12, the other recognizes it on invoice date. Both are defensible; running both is not.
5. A cohort window that silently shifts, usually because dunning states count as active. Stripe's past_due is not active, and a naive status != 'canceled' filter treats it as though it were.

The fix for all five is that cohort method: pick a month, freeze the exact customer list active in it, then ask what those same customers pay you a year later. New logos are excluded by construction, not by a filter. Here it is on PostgreSQL 16 against a monthly MRR snapshot table:

-- Cohort NRR and GRR with expansion, contraction and churn separated.
-- The cohort is frozen at the anchor month, so new logos cannot enter it.
WITH anchor AS (
  SELECT customer_id, SUM(mrr_cents) AS start_mrr
  FROM mrr_monthly
  WHERE month = DATE '2025-06-01'
    AND status = 'active'          -- past_due and unpaid are NOT active
  GROUP BY customer_id
),
later AS (
  SELECT customer_id, SUM(mrr_cents) AS end_mrr
  FROM mrr_monthly
  WHERE month = DATE '2026-06-01'
    AND status = 'active'
  GROUP BY customer_id
),
movement AS (
  SELECT
    a.start_mrr,
    COALESCE(l.end_mrr, 0)                            AS end_mrr,
    GREATEST(COALESCE(l.end_mrr, 0) - a.start_mrr, 0) AS expansion,
    CASE WHEN COALESCE(l.end_mrr, 0) = 0
         THEN a.start_mrr ELSE 0 END                  AS churned,
    CASE WHEN COALESCE(l.end_mrr, 0) BETWEEN 1 AND a.start_mrr
         THEN a.start_mrr - l.end_mrr ELSE 0 END      AS contraction
  FROM anchor a
  LEFT JOIN later l USING (customer_id)   -- LEFT JOIN keeps churned accounts
)
SELECT
  SUM(start_mrr)   AS start_mrr,
  SUM(expansion)   AS expansion,
  SUM(contraction) AS contraction,
  SUM(churned)     AS churned,
  ROUND(100.0 * SUM(end_mrr) / SUM(start_mrr), 1) AS nrr_pct,
  ROUND(100.0 * (SUM(start_mrr) - SUM(contraction) - SUM(churned))
                / SUM(start_mrr), 1)              AS grr_pct
FROM movement;

Two things a naive MRR sum can't give you. The LEFT JOIN keeps customers who went to zero, so churn is counted instead of silently vanishing from both sides of the ratio. And the movement components come out as separate columns, so NRR and GRR derive from the same figures rather than from two queries that drift apart.

Then write the boring artifact: a one-page metric-definition doc naming, per metric, the system of record, the exact filter, the recognition-date rule and the currency rule. Drift usually starts when a CRM field and a billing field are both called "MRR" and nobody wrote down which wins, which is what a custom HubSpot API integration has to resolve, on top of the AI stack we build SaaS products on.

Which SaaS Metrics Should You Stop Tracking?

Six metrics take up dashboard space and return nothing. Cut them. Competing guides list between 15 and 38 metrics with no priority order, and 38 items isn't a dashboard, it's a filing cabinet.

1. NPS as a growth metric. Fine as a CX signal. But NRR measures in dollars what NPS only implies in sentiment, and not one 2026 benchmark report above publishes an NPS figure.
2. Raw signups. Counts intent, not value. A launch spike and a bot-farm spike look identical in this chart.
3. Total registered users. A cumulative number that can only go up isn't a metric, it's a trophy. It can't tell you you're shrinking.
4. Page views. Marketing input, not business output. It belongs in a marketing report, not next to NRR.
5. MAU/DAU without paid qualification. Free-tier activity mixed with paid produces a number that moves for reasons you can't act on. Split it or drop it.
6. LTV as a standalone number. The strongest call here. LTV compounds three estimates (ARPA, gross margin, churn) into one confident-looking figure, and small churn errors move it enormously. Use CAC payback: two numbers you actually observe.

What Changed Since the 2021 SaaS Metrics Playbook?

David Sacks and Ethan Ruby's October 2021 essay is the origin document for this exact phrase and deserves its credit: the six-category framework is the clearest organizing model in the category, and it popularized the burn multiple. David Skok's "SaaS Metrics 2.0" sits alongside it as the other canonical reference.

The problem isn't the framework. It's that the thresholds were calibrated to 2021 capital markets, and nothing on either page tells a 2026 reader that. A founder benchmarking against them concludes they're failing when they're at median:

• NRR: the 2021-era "healthy" bar was 110%+. The CY-2025 top quartile is 108% and sliding, so 110% is now above the top quartile, not a baseline.
• CAC payback: 12 months was the old target. The CY-2025 median is 16 months, and only the top quartile clears 6.
• Growth: the ZIRP-era default for a sub-$20M bootstrapper was 30%+. The CY-2025 median is 15%.

The 2021 playbook isn't wrong. It's calibrated to a capital market that no longer exists.

Frequently Asked Questions

What is the most important SaaS metric?

Net revenue retention, and it isn't close. NRR is the only single number capturing churn, contraction and expansion together, and it predicts whether the business compounds without new sales. The CY-2025 median is 102% (Aleph × Benchmarkit, 2026 edition, n=230). Pair it with GRR so expansion can't hide churn.

What is a good NRR for SaaS in 2026?

Above 102% puts you past the median, and top-quartile territory has been sliding downward rather than rising (Aleph × Benchmarkit, n=230, CY-2025). Pricing model matters enormously: usage-based companies post a 108% median NRR against 98% for seat-based. Judge yourself against your own pricing model, not the aggregate.

What is the Rule of 40 in SaaS?

Growth rate plus profit margin should total at least 40. A company growing 30% with a 10% free-cash-flow margin clears it exactly. The CY-2025 median is 25%, up from 15%, and the top quartile reaches 43% (Aleph × Benchmarkit). Most of that gain came from cost cuts, not growth.

What is a good CAC payback period?

The CY-2025 median is 16 months, improved from 18 the prior year. Top-quartile companies repay acquisition cost in 6 months or less; the bottom quartile takes 24 or more (Aleph × Benchmarkit, n=198). Companies growing above 50% average 10 months.

How many SaaS metrics should you track?

Nine, grouped into retention, efficiency and growth quality. More is actively worse: a 38-metric dashboard has no priority order, so nobody looks at it and nobody owns any single number. Pick nine, give each an owner and a written definition, review on a fixed cadence.

What SaaS metrics do VCs actually look at?

Growth rate, net revenue retention, CAC payback and burn multiple, roughly in that order, with the Rule of 40 as the valuation screen. Investors weight capital efficiency more heavily than in 2021, which is why burn multiple and CAC payback have displaced LTV:CAC in most diligence conversations.

Which SaaS metrics are vanity metrics?

Raw signups, total registered users, page views, unqualified MAU/DAU, NPS as a growth predictor, and LTV quoted standalone. Each either only moves up, mixes free with paid activity, or compounds several estimates into one confident-looking figure you can't act on.

What's the difference between NRR and GRR?

NRR includes expansion revenue and can exceed 100%; GRR excludes it and never can. A business at 105% NRR and 84% GRR is losing customers and covering it with upsells to survivors. The CY-2025 medians are 102% and 84% (Aleph × Benchmarkit).

Why doesn't my MRR dashboard match my accounting system?

Usually one of five mechanical causes: new-logo revenue leaking into the expansion term, contraction and churn merged into one bucket, ARR snapshotted at different moments by billing and by the dashboard, monthly and annual contracts normalized differently, or dunning states counted as active. Write one definition doc and pick a system of record.

The Short Version

Three things. Median NRR is 102% and median GRR is 84%, both down, so retention is quietly getting harder for everyone (Aleph × Benchmarkit, n=230, CY-2025). Usage-based pricing beats seat-based by 10 points on NRR, making contract structure a retention decision, not just a pricing one. And bootstrapped growth fell from 20% to 15% while retention held flat, so the slowdown lives in acquisition (SaaS Capital, n>1,000).

Then there's the other half of the pre-scale list: what else to get right before you scale. If your numbers and your P&L disagree, that's a plumbing problem with a known fix, so get in touch if you'd rather not debug it yourself.