ARR Formula for Annual Contracts, Usage and Retainers

MRR times 12 is the standard ARR formula, and it is right for exactly one kind of customer. That customer pays the same amount every month and has no other fee attached.

That covers fewer contracts than you would guess. Once you sell a three-year deal, a usage plan with a minimum, or a service retainer, “times 12” produces a number that is either too high or impossible to defend, and the board deck and the pipeline report start to disagree.

The sources I trust on this agree on the core and fight about the edges. Tabs gives the multi-year rule as total contract value divided by the term in years, and its example is a $36,000 three-year contract that contributes $12,000 to ARR, not $36,000. Ordway says there is a real split on usage fees. I’ll take a position on each edge below, with the arithmetic shown, so you can copy the rule that fits your billing.

Start with the base formula and what stays out

For subscriptions billed monthly, ARR is MRR times 12. For annual and multi-year contracts, ARR is the recurring contract value divided by the term in years. Sum both across every customer whose contract is active today.

Three things stay out. One-time fees, meaning implementation, onboarding, hardware and professional services, are not recurring. Trials and expired contracts are not active. Credits and refunds against subscription charges reduce the run rate and should come out of it. The Tabs guide says the same about one-time fees, and it also says ARR only counts active contracts.

For anyone chasing a “100k ARR” or “1m ARR” goal, divide by 12 to get the monthly figure, which gives about 8,333 dollars and 83,333 dollars of MRR. Do the reverse when someone quotes MRR. This is also the whole of the ARR vs MRR distinction, which is why it trips people only when the two come from different systems.

Annual and multi-year contracts need the term, not the invoice

Invoices lie about timing.

A contract paid annually in advance shows a large invoice in one month and nothing for eleven, and a multi-year deal can be billed in a single payment. If ARR is built from invoices, it jumps and collapses with the billing calendar, so build it from the contract term instead, which is what the Tabs formula does.

These four contracts are made up to keep the numbers clean.

The three-year contract worth 36,000 is 12,000 of ARR, which matches the Tabs example. The one-year contract worth 24,000 paid up front is 24,000. The two-year contract worth 18,000 that also carried a 3,000 onboarding fee is 9,000, because the fee comes out and the rest is spread over two years. The one-year contract worth 30,000 that was signed last week but starts in January adds nothing today.

That adds up to 45,000 of ARR. If you add the signed contract you get 75,000, which is a different metric called contracted ARR, or CARR. It is useful for pipeline visibility and the Tabs guide defines it as including signed deals that haven’t started. Report it, label it, and never let it replace ARR in a slide that says ARR.

Ramped contracts are where I disagree with the simple formula. Say a customer signs 20,000 in year one, 40,000 in year two and 60,000 in year three. Total value is 120,000 and dividing by three gives 40,000, yet that customer pays you 20,000 a year today. I would report 20,000 as ARR and show the step-ups in a contracted line. Averaging a ramp makes ARR look better than the bank account for two years.

Usage billing: put the floor in ARR and show the rest separately

This is the edge where sources split. Ordway’s position is that the test for including usage fees is whether they are predictable, and it describes annualizing by taking the GAAP-recognized usage revenue from the most recent quarter and multiplying it by four. The Tabs guide offers two options instead: count only the contractually committed minimum, or use a consistent trailing average for the variable portion and report it apart from core ARR.

I would take the second route and keep it strict. The committed minimum goes into ARR. Usage above the minimum goes on its own line, usually called usage run rate, and it is never added in.

The clearest public case is New Relic’s fiscal 2021 10-K. After it moved to consumption pricing, the company said ARR is based on commitments and does not capture actual usage, and that it would lean on other metrics. When a public software company concludes that its own ARR definition stopped describing usage revenue, a startup should not stretch the same definition to cover it.

Here is the arithmetic. Take a customer who commits to 2,000 a month with a usage plan on top, so committed ARR is 24,000. Overage billed over the last three months was 1,500, 2,100 and 900, which is 4,500 for the quarter. Ordway’s quarter-times-four method turns that into 18,000. I would show it as usage run rate beside the 24,000, so the reader sees both and can discount the second however they like.

Ordway also warns against treating spikes as a signal, and that is the reason I won’t fold the variable piece in. One big month would raise ARR, and one quiet month would lower it, and neither says anything about the contract.

Retainers count only when the commitment is real

Service retainers are the hardest case because the billing looks recurring and the contract often isn’t. A retainer of 6,000 a month with 30 days’ notice is a commitment of one month, whatever the invoice history looks like.

My rule has three parts. A retainer counts in ARR when it has billed at the same amount for at least three consecutive months and the contract has no usage cap that resets the price. Project fees and hour bundles that expire are excluded. When the amount moves, use the average of the last three months times 12.

With numbers, a steady 6,000 retainer is 72,000 of ARR. A client who paid 4,000, 6,000 and 5,000 over the last three months averages 5,000, which is 60,000. A one-off project for the same client stays out of both.

I’d also write the rule down before the first retainer shows up in the pipeline. The temptation to count a good quarter of project work as recurring is highest right before a fundraise, which is exactly when someone will check.

Fields that make the CRM calculate ARR for you

The calculation above is only useful if the CRM computes it once and everything else reads from it. A board deck built from a spreadsheet and a pipeline report built from deal amounts will drift apart within a month. The fix is to store the inputs on the deal or the subscription record and compute ARR in one property.

The inputs I would add:

1. Revenue type, with three values: subscription, usage, retainer.
2. Contract start date and end date, so “active today” is a formula and not an opinion.
3. Term in months.
4. Recurring contract value, which excludes one-time fees.
5. One-time fees, kept in their own field so they are visible and never leak into the first one.
6. Committed minimum, for usage contracts.
7. Billing frequency.

Two computed properties come from those. ARR is the annualized recurring value when today falls between start and end. CARR is the same value for contracts signed but not started. Everything downstream reads those two properties. Nothing else.

This is the same discipline I argued for in RevOps for startups: build a system, not a team, where one source of truth matters more than the tool. It also belongs inside the larger revenue architecture blueprint as part of the data model, before the pipeline stages are designed.

Once the definition is written down it can also be checked automatically. A scheduled check that flags any active deal with a missing term, a recurring value that equals the full invoice, or a usage contract with no minimum catches the errors that otherwise surface in a diligence call.

What to do with the number once it is clean

A consistent ARR is the input to every other metric you quote, which is why it deserves this care. The five SaaS metrics that predict growth assume the revenue base is defined the same way each month, and the stage benchmarks in the $0 to $1M ARR playbook only mean something if your ARR follows the same rules as the numbers it is compared against.

If I were the founder, I would publish the definition in a single paragraph in the finance folder, with the four edge rules from this post, and put the date on it. Investors forgive a conservative definition. They do not forgive one that changed the quarter the growth rate flattened.

If you want help wiring this into your CRM, the CRM agents page shows what we build, and the free growth audit at app.momentumnexus.com covers where your current numbers disagree.