What Is ARR? Annual Recurring Revenue, Defined — and How Founders Stretch It
Annual recurring revenue is the annualized value of the subscriptions you have under contract — not services, not one-time fees, not your best month times twelve. The strict math, and how to read a founder-quoted ARR.

On this page
Annual recurring revenue (ARR) is the annualized value of the recurring subscription revenue a company has under contract right now. Take every active subscription, normalize what it pays to a yearly figure, and add it all up — that sum is ARR. One-time fees, implementation charges, and consulting work are excluded by definition, no matter how reliably they show up.
or, for contract-billed businesses:
ARR = sum of the annualized value of every active subscription
That is the strict version. In practice, ARR is the most-quoted and least-standardized number in SaaS. It is not a GAAP metric, no auditor signs off on it, and across 1,500+ founder interviews on the Latka show, we have heard "ARR" used to mean everything from clean subscription revenue to last month's total billings times twelve. This page covers the strict math, the two calculation conventions, and how to read a founder-quoted ARR without getting fooled — including two episodes from our own archive where the headline number and the tape disagree.
What counts in ARR — and what doesn't
The test is contractual recurrence: is the customer committed to paying this amount on a repeating schedule? If yes, annualize it and count it. If it's a one-off — however predictable — it's revenue, but it isn't ARR.
| Counts toward ARR | Excluded from ARR |
|---|---|
| Subscription fees (monthly, annual, or multi-year, normalized to one year) | One-time setup and implementation fees |
| Recurring seat, platform, and license fees | Professional services and consulting projects |
| Recurring add-ons already live (extra seats, support tiers billed as subscriptions) | Usage overages the customer hasn't contractually committed to |
| Contractually committed minimums in usage-based deals | Signed-but-not-yet-live bookings — that's contract value, not ARR |
| Perpetual license sales |
A multi-year deal counts at its per-year value: a $240,000 two-year contract contributes $120,000 to ARR, not $240,000.
ARR vs. revenue vs. run rate
These three get used interchangeably, and they measure different things.
Revenue is the GAAP number: what the company actually recognized over a trailing period, under accounting rules, including services, one-time fees, everything. It looks backward and it's auditable.
ARR is a forward-looking snapshot: what the current subscription base would produce over the next twelve months if nothing changed. It excludes non-recurring revenue and no accounting standard governs it — which is precisely why it gets stretched.
Run rate is the loosest of the three: take a recent period — a month, a quarter, sometimes a single big week — multiply it out to a year, and include whatever revenue happened to land in it. A December spike, a large services engagement, a one-time license deal: all of it annualizes. When a company says "annualized run rate," it is telling you the number includes non-recurring revenue; when it says "ARR," it is claiming the number doesn't. Hold it to that.
The two ARR conventions
Companies that bill monthly typically compute ARR as MRR × 12: build the month's recurring revenue from all its moving parts, then annualize. Companies that sell annual or multi-year contracts typically sum contract values instead — each active contract contributes its per-year price, and average contract value × customer count should land on the same total. Both are legitimate. What matters is knowing which one you're looking at, because MRR × 12 moves every month while a contract-sum ARR only moves when deals are signed, renewed, or churned.
A worked example that closes
Say you start the month at $100,000 in MRR — $1.2M in ARR. Over the month:
− MRR lost to cancellations: $5,000
− MRR lost to downgrades: $1,000
+ expansion MRR from existing customers: $8,000
+ MRR from new customers: $10,000
= Ending MRR: $112,000 → ARR: $1,344,000
Check it: 100,000 − 5,000 − 1,000 + 8,000 + 10,000 = 112,000, and 112,000 × 12 = 1,344,000. That's 12% month-over-month MRR growth.
The component definitions, briefly. Churned MRR is revenue from customers who canceled outright. Downgrade MRR is the delta when a customer stays but pays less: a customer moving from a $1,000/month plan to $250/month costs you $750 of MRR. Expansion MRR is new recurring revenue from customers you already had — upsells and add-ons only, never new logos. New MRR is recurring revenue from customers who signed this month.
Two health signals fall out of the same ledger. Gross churn here is $6,000 lost against $100,000 — 6% monthly, which is high. But the existing base also expanded by $8,000, so net revenue movement from existing customers is +$2,000: net revenue retention above 100% even before new sales. Those retention dynamics flow directly into customer lifetime value — the mechanics are in our LTV:CAC ratio guide.
How founders stretch ARR — and how to read it
After enough interviews, the inflation patterns repeat. The common ones: annualizing the best month rather than a normal one; folding services and one-off projects into the recurring number; quoting signed contract value as if it were live ARR; counting uncommitted usage revenue; and saying "run rate" quietly enough that listeners hear "ARR."
This isn't hypothetical, and we'll audit our own archive to show it. When Rosterfy CEO Shannan Gove came on the show in September 2021, the episode went out as "Rosterfy Hits $3.2m ARR." On the tape, the actual split: roughly $3.2M in total revenue that year, of which about $2M was pure recurring and about $1.2M was consulting contracts. Gove himself was the honest one — he told Latka a 10–15x multiple should apply to the recurring revenue only. The headline "ARR" was about 60% ARR.
Same pattern a rung up. The December 2019 Bombora episode ran as "Bombora CEO Hits $30m ARR." On the tape, CEO Erik Matlick says the company was "well over $30 million" in annual revenue — and that about 65% of it was subscription revenue. Strictly counted, ARR was closer to $20M. Neither founder was hiding anything; the split is right there in the interview. But the headline number and the defined metric were different things, and that's the norm, not the exception. Our own company pages label their charts "reported revenue / ARR" for exactly this reason: founders report the blended figure, and where the tape gives a split, the page notes it.
The fastest sanity check is whether the arithmetic closes: customers × average contract value should approximately equal the claimed ARR. Sendbird's 2024 figures on GetLatka — $40M revenue, 422 customers, $94.8K average ACV — close almost exactly (422 × $94.8K ≈ $40M). So do Bombora's: 300 customers × $186.7K ≈ $56M for 2024. When a founder's customer count, pricing, and claimed ARR can't be multiplied into each other, one of the three is wrong.
Five questions that separate a real ARR from a stretched one:
- As of when? ARR is a snapshot; an undated ARR is a soft number.
- MRR × 12 or contract sum? Annualizing one strong month is the oldest trick in the book.
- Are services, implementation, or one-off projects inside the number?
- Is usage revenue contractually committed, or just what customers happened to spend?
- Does customers × average contract value land near the claim?
Dated examples from real companies
| Company | Quoted figure | What the tape / page shows |
|---|---|---|
| Sendbird | "$20m in ARR" — Feb 2019 interview with CEO John Kim | Consistent with the $21.8M revenue GetLatka records for March 2020; $40M reported for 2024 |
| Bombora | "$30m ARR" — Dec 2019 episode title | CEO Erik Matlick on tape: $30M+ total revenue, ~65% subscription — strict ARR ≈ $20M |
| Rosterfy | "$3.2m ARR" — Sep 2021 episode title | CEO Shannan Gove on tape: ~$2M pure recurring + ~$1.2M consulting |
Want to pressure-test more claims yourself? The GetLatka SaaS database lists reported revenue, customer counts, and average contract values side by side for thousands of companies — the three columns you need to check whether an ARR closes.
This piece draws on:
- GetLatka interview with Sendbird CEO John Kim, recorded February 6, 2019, and the Sendbird company page (getlatka.com/companies/sendbird), last updated November 2025
- GetLatka interview with Bombora CEO Erik Matlick, published December 1, 2019, and the Bombora company page (getlatka.com/companies/bombora), last updated August 2026
- GetLatka interview with Rosterfy CEO Shannan Gove, recorded September 16, 2021, and the Rosterfy company page (getlatka.com/companies/rosterfy)
