Annual recurring revenue: ARR, run rate, and the difference that matters
Annual recurring revenue is the value of a company's contracted, recurring subscription revenue expressed over one year, excluding one-time fees and services. It is the headline number in almost every SaaS board deck and funding announcement.
It is also two different metrics wearing the same three letters. One is built from signed annual contracts. The other is last month's revenue multiplied by twelve. They can differ by a wide margin for the same business, and the difference is usually the most interesting thing about the number.
Table of contents
- Annual recurring revenue vs annualised run rate
- The ARR formula and what it excludes
- Multi-year, ramped and discounted contracts
- How ARR gets inflated
- A worked example on a mixed book
- ARR and MRR should reconcile exactly
- Not all ARR is worth the same
- When ARR is the wrong metric
- What ARR cannot tell you
- Frequently asked questions
Annual recurring revenue vs annualised run rate
True annual recurring revenue comes from contracts. Annualised run rate comes from extrapolation. Most articles on the topic collapse the two into one formula, which is how a company with three months of history ends up quoting an ARR figure.
Annual recurring revenue applies to businesses selling annual or multi-year contracts. You know the contracted value because a customer signed it.
Annualised run rate applies to businesses billing monthly. There is no annual contract, so you take current MRR and multiply by twelve to express the current rate as an annual figure.
Annualised run rate = MRR x 12
Annual recurring revenue = total contract value / number of years
The gap between them is a forecast. Run rate assumes the current month repeats eleven more times, which for a monthly-billing business with meaningful churn it will not, which is the same assumption that breaks most customer lifetime value calculations. A company at $100,000 MRR with 3% monthly churn and no new sales does not earn $1.2m over the next year. It earns closer to $1.01m, because the base shrinks every month.
For a business on annual contracts, ARR carries real information: customers have committed. For a monthly-billing business, ARR is a presentation choice, and calling it "annual recurring revenue" rather than "run rate" overstates how much of it is actually locked in.
Neither is dishonest. Reporting one while your audience assumes the other is.
The ARR formula and what it excludes
The clean version, for a business on annual contracts:
ARR = sum of contracted annual subscription value across active customers
For a multi-year contract, divide total contract value by the number of years. ChartMogul's worked example: a $6,000 four-year contract contributes $1,500 of ARR.
Exclusions match the MRR rules and matter more here, because the numbers are twelve times larger and so are the errors:
- One-time fees. Setup, implementation, migration, onboarding.
- Professional services. Custom development, training, fixed-scope consulting. Recurring-looking services retainers with an end date are still not recurring revenue.
- Usage overages that are not contracted. A customer who happened to spend $30,000 above their commitment last year has not contracted to do it again.
- Non-renewing contracts already served notice. Still generating revenue, no longer recurring.
- Taxes and processing fees.
Discounts come off the top. A customer with $120,000 list price on a permanent 20% discount contributes $96,000, not $120,000.
Multi-year, ramped and discounted contracts
This is where the standard articles stop and where enterprise contracts actually live. Ramps, tiered discounts and mid-term modifications go unaddressed in essentially every widely-ranking explanation of the metric.
Ramped contracts. Enterprise deals frequently step up. A three-year contract priced at $100,000 in year one, $150,000 in year two and $200,000 in year three has a total contract value of $450,000.
Dividing TCV by years gives $150,000, and that figure is wrong for almost every purpose you would use ARR for. It overstates the revenue you are currently earning by 50% and understates where you will be in two years. What you are contracted to receive right now is $100,000.
The defensible convention is current contracted ARR: the annualised value of the contract year you are in. Year one shows $100,000, and the step-up appears as expansion when it takes effect. That keeps ARR aligned with what is actually recurring today and makes the growth visible when it happens rather than smearing it backwards.
If you report the TCV average instead, say so. Investors will ask, and finding out during diligence that reported ARR was a three-year average is worse than a lower number honestly labelled.
Discounts with an end date. A customer on $120,000 list with a 20% first-year discount contributes $96,000 in year one and $120,000 after it expires. Booking $120,000 from the start overstates current ARR by 25% and makes the eventual increase invisible, since the system already assumed it.
Mid-term modifications. A customer who adds fifty seats in month seven of a twelve-month contract has changed their contracted annual value from that point. Recalculate ARR from the amended contract rather than waiting for renewal, otherwise expansion lands in a lump at renewal months after it was earned.
Volume tiers. A contract priced per unit with tier breaks has no single annual value until you assume a volume. State the assumption. Committed minimum is the conservative choice and the one that survives audit.
How ARR gets inflated
ARR has no accounting standard behind it. Unlike revenue recognised under IFRS 15 and ASC 606, nobody audits an ARR figure, and every company defines it slightly differently. The common inflation tactics are worth knowing whether you are reporting the number or reading someone else's.
Best month times twelve. Take the strongest month of the year, annualise it, present it as ARR. ChartMogul singles out this practice specifically. It is most common in businesses with seasonality or a large one-time component.
Including one-time revenue. Setup fees and services rolled into the recurring line. A company with $2m of subscriptions and $800,000 of implementation work reporting $2.8m ARR is reporting something that will not recur without selling the same volume of services again next year.
Counting signed but not started. A contract signed in December that begins in March is bookings, not recurring revenue.
Counting pilots and LOIs. Paid pilots without renewal commitments, and letters of intent, are not contracted recurring revenue.
Ignoring notice given. A customer who has formally declined renewal is still paying, and their revenue is no longer recurring.
Gross rather than net of churn. Summing everything ever sold rather than what is currently active.
The test is simple. If every customer stopped buying anything new tomorrow and nothing else changed, would this figure still arrive over the next twelve months? If the honest answer involves an assumption about future behaviour, it is a forecast, and the label should say so.
A worked example on a mixed book
Most explanations calculate ARR for one tidy customer. Real books are mixed, and the mix is where the errors are. Here is a six-customer book with every awkward case in it.
| Customer | Contract | Correct ARR contribution |
|---|---|---|
| A | $2,000/month, monthly billing, no commitment | $24,000 (run rate, not contracted) |
| B | $60,000/year annual contract | $60,000 |
| C | $180,000 three-year flat contract | $60,000 |
| D | Three-year ramp: $100k, $150k, $200k, currently in year one | $100,000 |
| E | $120,000 list, 20% discount expiring after year one | $96,000 |
| F | $40,000/year plus $25,000 implementation fee | $40,000 |
Total ARR: $380,000.
The three most common wrong answers on the same book are worth naming. Counting customer D at the TCV average of $150,000 gives $430,000. Counting customer E at list gives $404,000. Rolling customer F's implementation fee into recurring revenue gives $405,000. Make all three mistakes and the book reports $479,000, which is 26% above the defensible figure.
None of those errors requires bad intent. Each is a plausible reading of an ambiguous definition, which is exactly why the definition needs writing down. Two people calculating ARR from the same billing system should not be able to produce numbers 26% apart.
One further note on customer A. Their $24,000 is annualised run rate sitting inside a figure otherwise built from contracts. Mixed books are normal, and the honest presentation splits contracted ARR from run-rate ARR so the reader knows how much of the total is committed. Most companies report one blended number. Being able to produce the split on request is what separates a metric you control from one you inherited.
ARR and MRR should reconcile exactly
ARR and MRR describe the same subscriptions at different resolutions, so they should reconcile to the twelve-times relationship without a plug.
ARR = MRR x 12
When they do not reconcile, the cause is nearly always one of three things. One of the two is being computed from invoices rather than subscriptions, so annual prepayments distort the month they land in. One includes one-time fees the other excludes. Or annual contracts are being counted at full value in MRR instead of normalised to a monthly figure.
Run the reconciliation monthly. It is the cheapest data quality check available on a subscription business, and a persistent gap between the two figures always means one of them is wrong.
The same movements that decompose MRR apply here: new, expansion, contraction and churn. Net new ARR is what the total movement was, and reading it without the decomposition has the same problem at annual scale that it has monthly. Net revenue retention is usually the more informative annual number, because it isolates what happened to the customers you already had.
Not all ARR is worth the same
Two companies reporting identical ARR can be worth very different multiples, because ARR measures size and says nothing about quality. Benchmark sets like Bessemer's Cloud Atlas consistently pair growth with efficiency and retention rather than reporting ARR growth alone. Four attributes separate them, and none of them appear in the headline figure.
Contract length and commitment. A million of ARR on three-year contracts with auto-renewal is more durable than a million on monthly plans that can cancel tonight. Weighted average contract length is the usual summary.
Concentration. If the largest customer is 30% of ARR, the business has a single point of failure that a growth rate will not reveal. Reporting the share held by the top five and top ten accounts alongside the total is standard in diligence and useful long before then.
Retention. ARR growth funded entirely by new sales, against a base that leaks, is more expensive to sustain than the same growth on a base that expands on its own. This is what net revenue retention measures, and it is why NRR usually gets asked about in the same breath as ARR.
Gross margin. Recurring revenue requiring heavy human delivery to sustain is not the same asset as software revenue, even when both recur on a signed contract.
The practical version: report ARR with its composition attached. Total, split by contract length, with concentration and retention beside it. A single number invites the reader to assume the most flattering composition, and sophisticated readers will assume the least flattering one instead.
When ARR is the wrong metric
ARR assumes a contracted annual commitment exists. Three situations break that assumption.
Monthly-billing businesses. No annual commitment exists, so the figure is run rate. It is fine to report, and it should be called run rate.
Usage-based pricing. With no commitment and consumption that varies, there is no contracted annual value. Reporting committed ARR alongside trailing usage revenue as separate lines is more honest than blending them into one number.
High-churn consumer subscriptions. At consumer churn rates, twelve months is long enough that the annualised figure describes a customer base that will have substantially turned over. Monthly metrics and cohort retention curves carry more information.
There is also a scale problem. Below roughly a hundred customers, ARR is dominated by a handful of contracts, and a single renewal decision moves the headline number more than a year of product work. The figure is still correct. It is just mostly reporting the behaviour of your largest customer.
What ARR cannot tell you
ARR is the most heavily lagging metric in the standard SaaS set. It is an annual figure, usually reviewed monthly, describing commitments made over the preceding twelve months.
It cannot tell you composition. Two companies at $5m ARR, one with 500 customers at $10,000 and one with 5 customers at $1m, are not comparable businesses. The second has a concentration risk that the headline hides completely.
It cannot tell you when a change started. Annual contracts churn on renewal dates, so a retention problem that began in March surfaces as a decline in November, when the affected cohort's renewals come due. By then the next cohort is already forming the same opinion.
It cannot tell you why. A flat ARR line can mean a healthy business at equilibrium, or aggressive new sales exactly cancelling aggressive churn. Those demand opposite responses and look identical at annual resolution.
It cannot tell you which accounts. The renewals at risk today are visible in usage decline, support volume, contraction and champion departures, none of which appear in an annual aggregate until the renewal is already lost.
Frequently asked questions
What is the difference between ARR and MRR?
Resolution and, sometimes, source. MRR is the normalised monthly run rate. ARR is the same subscriptions expressed annually. For a business on annual contracts, ARR comes from signed contract values. For a monthly-billing business, ARR is MRR multiplied by twelve, which makes it a run rate rather than a contracted figure. They should always reconcile at twelve times.
Does ARR include one-time fees?
No. Setup, onboarding, implementation and professional services fees are excluded because they do not recur. Including them is one of the most common ways ARR gets overstated.
How do you calculate ARR for a multi-year contract?
Divide total contract value by the number of years for a flat contract. For a ramped contract, report the annualised value of the contract year you are currently in and let the step-ups appear as expansion, rather than averaging future increases into today's figure.
Is ARR the same as annual revenue?
No. ARR is a forward-looking run rate of contracted recurring revenue. Annual revenue is backward-looking recognised revenue under IFRS 15 or ASC 606, including one-time and services revenue. A company can report $5m ARR and a very different recognised revenue figure for the same year, and both can be correct.
What is a good ARR growth rate?
It depends almost entirely on stage. Percentage growth rates fall as the base grows, so a rate that is unremarkable at $1m is exceptional at $50m. Compare against your own trailing performance and your segment rather than against headline figures from benchmark reports, which skew heavily toward venture-funded companies at their fastest-growing moment.
Ready to see what is actually behind your ARR?
An annual recurring revenue figure is a single number summarising thousands of decisions made over twelve months. It tells you the total. It does not tell you which contracts expanded, which customers quietly stopped using the product six months before their renewal, or what changed in the weeks before a major account gave notice.
GainSignal decomposes recurring revenue into new, expansion, contraction and churn, connects each movement to the accounts behind it, and links those movements back to what changed first. Renewal risk becomes visible while there is still time to act on it, rather than at the renewal date.