Business Technology & SaaS · Reference

SaaS metrics, defined precisely

No accredited standards body defines these terms, so two companies can publish the same metric name and mean materially different things by it. A working reference: what each measure measures, the exact formula, and where each one goes wrong.

Reviewed August 2026 · The Insight Journal Editorial Team

In short

SaaS metrics are the recurring revenue, retention and unit economics measures used to describe a subscription software business: MRR, ARR, net revenue retention, churn, customer acquisition cost and lifetime value. No accredited standards body publishes definitions for them, so the same metric name can be calculated differently at different companies, and often is.

SaaS metrics on paper: printed line and bar charts fanned across a desk beside a pen and calculator in soft daylight.

The state of the answer

Four numbers that frame the problem

590

US monthly searches for the term itself, falling 45 percent year on year

DataForSEO, August 2026

+324%

Year-on-year rise in searches for "what is NRR in SaaS": the definitions are what people now want

DataForSEO, August 2026

8 of 9

Page-one results published by a party that sells software, capital or courses into the category

Live SERP analysis by The Insight Journal, 18 August 2026

3

Disclosures the SEC expects to accompany any performance metric a company publishes

SEC Release 33-10751, 2020

The root cause

Why the same metric name means different things

In short

There is no ISO standard, no accounting standard and no regulator that defines net revenue retention, ARR or customer acquisition cost. Each company picks a method, and unless it publishes that method, its number cannot be reproduced or compared. The nearest published discipline is a piece of SEC guidance that applies only to listed companies.

Accounting has authorities, and a buyer's subscription has a defined treatment, which our guide to business software by function covers on the purchasing side. The operating measures sitting on top of the accounts have no such authority.

One body has tried. The SaaS Metrics Standards Board publishes definitions for ARR, retention and CAC, and its own About page describes it as a coalition of financial platform vendors, metrics platform vendors, accounting firms, audit firms and SaaS investors.

It is a private membership organisation rather than an accredited standards body, and its last release of standards is dated June 2023. Useful, and not the same thing as a standard.

The formulas

Eleven measures, and where each one slips

Formulas first, then the sections below take the five that cause the most trouble in detail. There is no benchmark column, and that is deliberate: no benchmark figure could be retrieved from a free, dated, primary source, so none is published here.

Eleven SaaS metrics with the formula for each and the most common calculation error
Metric	Formula	The usual trap
MRR Monthly recurring revenue	Sum of the monthly value of active subscriptions	Setup fees and services work get folded in
ARR Annual recurring revenue	MRR x 12, or the sum of annualised contract values	Two different methods, both called ARR
ARPU Average revenue per user	Revenue for a period / customers at period end	Numerator and denominator taken from different dates
NRR Net revenue retention	Cohort revenue now / cohort revenue then	The cohort rule and window are rarely stated
GRR Gross revenue retention	(Cohort revenue then, less contraction and churn) / cohort revenue then	Capped at 100 percent, so it is quietly dropped
Logo churn Customer churn rate	Customers lost in period / customers at period start	Silent on what those accounts were worth
Revenue churn Gross revenue churn rate	Recurring revenue lost / recurring revenue at period start	Downgrades are sometimes excluded
CAC Customer acquisition cost	Acquisition spend in period / new customers in period	What counts as acquisition spend is disputed
CAC payback Months to recover CAC	CAC / monthly gross profit per customer	Computed on revenue rather than gross profit
LTV Customer lifetime value	Monthly gross profit per customer / monthly churn rate	Rests on an assumed lifetime, not a measured one
Rule of 40 Growth plus profitability	Growth rate percent + profit margin percent	Hides which of the two inputs is carrying it

Recurring revenue

MRR and ARR are run rates, not revenue

In short

MRR is subscription revenue normalised to a month, and ARR is the same figure annualised. Neither is revenue in the accounting sense: both are point-in-time run rates built from active subscriptions, and neither reconciles to the income statement. Treating ARR as revenue is the most common error in this vocabulary.

Two methods share one name

The run-rate method takes the last month of a period and multiplies by twelve. The contract-value method sums the annualised value of every active contract. They answer slightly different questions and produce different numbers in any business with seasonality or usage-based billing.

Backblaze uses the first, stating in its Form 10-K that ARR is the monthly revenue from its storage and backup arrangements in the last month of a period, multiplied by twelve. A legitimate method, clearly disclosed, and not the only one.

ARR does not have a standardized meaning and is therefore unlikely to be comparable to similarly titled measures presented by other companies.

Backblaze, Inc., Form 10-K for the year ended 31 December 2024, filed 11 March 2025. The company is telling its own investors not to compare its ARR with anyone else's.

Retention

Net revenue retention, defined four incompatible ways

In short

Net revenue retention measures what a cohort of existing customers is worth now against what it was worth then, after expansion, contraction and cancellation, excluding customers won during the period. Above 100 percent means the cohort grew. The measurement window, the cohort rule and the revenue base all vary by publisher, so the percentage alone carries almost no information.

Four published definitions of the same metric name, captured on 18 August 2026. Two are filings made to a regulator, two are vendors selling into the category, and no two would produce the same number from the same data.

Snowflake Inc.

FY2026 Form 10-K, filed 20 March 2026

SEC registrant reporting its own figure

Window: Trailing two years
Cohort: Customers on capacity contracts active at any point in the first month of year one
Base: Product revenue

A customer that leaves stays in the calculation at zero. Adjusted for acquisitions and spin-offs.

Backblaze, Inc.

FY2024 Form 10-K, filed 11 March 2025

SEC registrant reporting its own figure

Window: Trailing four-quarter average
Cohort: Accounts active at the end of the same quarter one year earlier
Base: Recurring revenue

A rolling average of four year-on-year quarterly comparisons, not one measurement.

Maxio

SaaS metrics guide, captured 18 August 2026

Sells subscription billing and metrics software

Window: "A given period", unspecified
Cohort: Not stated
Base: MRR

(Starting MRR + expansion, less contraction and churn) / starting MRR. A snapshot, not a cohort.

Stripe

SaaS metrics guide, captured 18 August 2026

Sells payments and Stripe Billing

Window: "A given period of time", unspecified
Cohort: Not stated
Base: Base recurring revenue

Retained revenue over base recurring revenue. Not reproducible as written.

Illustrative only

Take a cohort worth $50,000 in monthly recurring revenue at the start of a period. Over the period it gains $8,000 in upgrades, loses $2,000 to downgrades and $4,000 to cancellation.

NRR = ($50,000 + $8,000 − $2,000 − $4,000) / $50,000 = 104%
GRR = ($50,000 − $2,000 − $4,000) / $50,000 = 88%

Same cohort, same period. One number says the base is growing, the other says twelve percent of it left. Figures invented for illustration and checked arithmetically; they are not a benchmark.

Churn

Logo churn and revenue churn are not the same question

In short

Logo churn counts accounts lost as a share of accounts held. Revenue churn counts recurring revenue lost as a share of recurring revenue held. When customers vary in size, the two can diverge enormously, so a company reporting a single unqualified "churn rate" has not yet told you anything.

Backblaze reports the two separately in its Form 10-K, describing its gross customer retention rate as one that "reflects only customer losses and does not reflect the expansion or contraction of revenue" from existing customers. That separation is the practice worth copying.

Identical 5 percent logo churn in both rows. Read on revenue, one is a rounding error and the other is a fifth of the book. Figures constructed for illustration and verified arithmetically.

Unit economics

CAC, payback and the lifetime value assumption

In short

Customer acquisition cost divides acquisition spend by new customers won, but what counts as acquisition spend is genuinely disputed between published sources. Lifetime value then divides gross profit per customer by a churn rate, which makes it a forecast rather than a measurement. CAC payback period is the more checkable of the two.

The dispute inside CAC

Maxio defines CAC as total sales and marketing costs divided by the number of new customers acquired. Stripe scopes the numerator to the cost of sales and marketing directed at potential new customers, which excludes spend aimed at retaining and expanding existing ones.

Both wordings are defensible and they are not equivalent. Salaries are the second fault line: a paid-media-only figure and a fully loaded one can differ by a multiple in the same quarter.

Illustrative only

A quarter with $60,000 of paid media, $90,000 of sales salaries, $30,000 of marketing aimed at existing customers, and 60 new customers won.

All sales and marketing: $180,000 / 60 = $3,000
New-customer spend only: $150,000 / 60 = $2,500

Why lifetime value is the softest number here

LTV divides monthly gross profit per customer by the monthly churn rate, which is arithmetic for "assume this churn rate holds forever". It is a projection, and small changes to the assumed churn move it a long way.

Illustratively, at $250 monthly revenue per customer and an 80 percent gross margin, gross profit is $200 a month. LTV comes out at $10,000 on 2 percent monthly churn and $6,667 on 3 percent, so against a $2,500 CAC the ratio falls from 4.0 to 2.7 on a one-point change in an assumption.

CAC payback avoids the forecast entirely. At $2,500 CAC and $200 of monthly gross profit it is 12.5 months, where dividing by revenue instead of gross profit would report 10. It answers a question you can actually check against a cash flow forecast.

The nearest thing to a rule

What a regulator expects a published metric to carry

In 2020 the Securities and Exchange Commission issued guidance on key performance indicators in management discussion and analysis. It binds listed companies only, and it is the closest thing to a rule this vocabulary has.

The three disclosures

The Commission states it would generally expect a metric to be accompanied by a clear definition of the metric and how it is calculated, a statement of why it provides useful information to investors, and a statement of how management uses it.

And when the method changes

A company changing how it calculates a metric should consider disclosing the difference from prior periods, the reasons for it, and the effect on both current and previously reported figures. Where the change is significant, it should consider recasting the earlier numbers.

Which metrics it covers

The guidance names average revenue per user and daily and monthly active users among its examples. Separately it notes that operating and statistical measures such as subscriber counts fall outside the non-GAAP regime, which places many of these measures in an even less governed space.

Almost no reader of this page files with the Commission, but the discipline transfers. Write down what the metric is, how it is calculated, why you track it, and what a change of method would oblige you to disclose. Full text in the SEC guidance published at 85 FR 10568.

Reading them together

Which of these travel between companies

Some survive a comparison with another company and some do not. Each is more useful next to a companion measure than on its own.

Whether each metric is comparable between companies, the reason, and the metric to read alongside it
Metric	Comparable?	Why	Read it alongside
ARR	No	Two annualisation methods share the name, and it reconciles to nothing	Read against GAAP revenue for the same period
NRR	Only if the method is published	Cohort rule, window and revenue base all vary by reporter	Read against GRR and logo retention
CAC	No	What counts as acquisition spend is unsettled; salaries move it most	Read against payback and gross margin
LTV	No	It contains an assumed lifetime, so it is a forecast wearing a metric name	Read against CAC payback, which uses no forecast

Whoever publishes a number should be able to say how it was built, and the method should not change quietly between one board pack and the next. For the delivery model itself, start with the standards-based definition of SaaS. For where the model is heading, see what the evidence says about AI displacing SaaS.

Questions

Common questions about SaaS metrics

What is the 3 3 2 2 2 rule of SaaS?

It is a growth heuristic that circulates widely: triple revenue for two years, then double it for three. It is a rule of thumb about pace, not a metric with a formula, and we could not trace it to an authoritative published source. Treat it as folklore about ambition rather than a benchmark.

What is a good net revenue retention rate?

We are not going to publish a benchmark we could not verify. The figures that circulate come from surveys run by venture funds and vendors whose sample and method are usually undisclosed, and no free, dated, primary source was retrievable for any of them. Compare your own NRR with your own GRR instead, tracked on an unchanged method.

Is ARR the same as revenue?

No, and treating it as revenue is the most common mistake here. ARR is a run rate assembled from active subscriptions at a point in time, while revenue is recognised over the period the service is delivered. Backblaze states in its Form 10-K that ARR "does not have a standardized meaning and is therefore unlikely to be comparable to similarly titled measures presented by other companies".

Should churn be measured on customers or on revenue?

Both, because they answer different questions and can point in opposite directions. Logo churn counts accounts lost and speaks to product fit; revenue churn counts recurring revenue lost and tells you what it cost. Backblaze reports gross customer retention separately from net revenue retention for exactly this reason.

What counts in customer acquisition cost?

This is genuinely unsettled. Maxio defines CAC as total sales and marketing costs divided by new customers, while Stripe scopes it to spend directed at potential new customers, which excludes spend aimed at existing ones. Salaries are the other fault line: a paid-media-only CAC and a fully loaded CAC can differ by a multiple.

Why does my churn rate differ between two systems?

Usually because the two disagree about when a customer left, not about the arithmetic. A billing system records the churn when the subscription lapses; a CRM may record it when the cancellation was logged, and an annual contract cancelled in month two does not lapse until month twelve. Fix the event definition first.

Method and disclosure

Definitions on this page were captured live on 18 August 2026. Two come from annual reports filed with the Securities and Exchange Commission and retrieved from EDGAR; two come from vendor guides, each labelled with what its publisher sells. Where sources disagree, both readings are shown rather than one being chosen silently.

No benchmark appears anywhere here, because none could be traced to a free, dated, primary source. We sell no software and take no placement fee on this page; our policy on placement and sourcing sets out the rest.