ARR used to mean one thing. Now it means whichever number makes the deck look best, and most investors have stopped asking which one they're getting.
One AI startup founder found this out the hard way. Last summer, he told a reporter his company was doing $7 million in ARR. By this spring, he'd admitted publicly that the real number was closer to $5.2 million.
He described the whole thing as an offhand answer to a cold call he wasn't expecting to matter. Except the call happened because his own PR team had pitched the story. A founder inflating his number to a reporter his own publicist booked. That pretty much sums it up.
Here's the taxonomy, since nobody's stated it plainly: ARR can mean Annual Recurring Revenue, Annualized Run Rate, or what I'd call Aspirational Revenue Roadmap. Founders know exactly which one they're pitching, they just don't say it out loud.
Contracted or committed ARR, CARR, gets rebranded as ARR, and a meaningful share of that contracted revenue never actually shows up as cash. One VC told TechCrunch he's seen companies where CARR runs 70% ahead of live ARR. Nobody in that deal flags it, because everybody's portfolio has a company doing the same thing. That's not an incentive problem at the founder level. It's an incentive problem at the fund level, and General Catalyst's Hemant Taneja said the quiet part out loud when he noted that going from 1 to 3 to 9 to 27 doesn't read as a good story anymore. It has to look more like 1 to 20 to 100. That pressure is what manufactures the fake hockey stick. Founders aren't lying because they're dishonest. They're lying because the shape of the curve is the pitch.
Phia is another example, and it's worse because the revenue was real. Bloomberg's reporting found that cookie stuffing accounted for roughly half of the merchandise value the shopping app was claiming, and once the feature got disabled, daily revenue reportedly dropped by more than half. The founders are alleged to have known for months before any of it became public.
Frothy markets and fraud aren't separate problems. The more inflated a market gets, the more lax investors become, and that laxity is exactly what lets fraud slip through the cracks. A founder at a well-known Bay Area accelerator will tell you, with a straight face, that they have no materials to share, no website, and don't care if that's a non-starter. You have 24 hours to decide, and it's uncapped. In frothy markets, young entrepreneurs get handed a lot of bad advice about how to sell investors, and it taints their sense of what the relationship between capital and founders is actually supposed to look like.
None of this requires a new set of ethics. It requires vocabulary. Revenue is cash collected or GAAP-recognized. ARR is the annualized value of live, paying, recurring contracts. Run rate is last month times twelve, and it only means something if the business is actually recurring and not spiky. CARR is signed but not yet live, net of expected churn. Everything else is pipeline. Every number in your deck should be labeled with one of those five words, and investors should never have to ask which one you mean.
Usage-based pricing breaks the whole framework, which is a problem most AI companies pretending to be SaaS companies haven't dealt with yet. If revenue moves with consumption, annualizing one good month is a story, a snapshot in time, not a number. Great entrepreneurs show six trailing months and let the investor do the multiplication themselves (and they show retention next to it).
Concentration is the tell most founders don't think to hide because they don't realize it's damning. $500k in ARR from three customers isn't traction. It's three relationships you can't afford to lose, and one churned logo is a 33% haircut you'll have to explain in the next update. A cohort table of monthly signups and their revenue over time takes ten minutes to build and can't be gamed the way a single headline number can.
Quality of revenue should carry a premium, and high growth paired with high churn should carry a heavy discount. Right now it runs backwards. The market rewards the hockey stick and shrugs off the churn, when the discount should go the other way.
LOIs deserve the same scrutiny. An LOI with a dollar figure, a timeline, and a named budget owner is worth something. One that says the customer is excited to explore is a marketing asset with a logo on it. Once a customer puts down a deposit, you're in a different conversation. Enthusiasm is free, money is not.
The actual risk here was never the embarrassment of a reporter catching the gap, but the next round. Tell your seed investor $2M when the live number is $1.2M, and your Series A pitch now has to explain why growth mysteriously slowed. Diligence is catching up. Bank statements, Stripe exports, and reference calls are standard now, and third-party verification is getting cheap enough that numbers you can't support in a data room shouldn't be in a deck.
The AI founder’s number was a lie to a reporter. Phia's alleged number was a lie to affiliate networks. The distance between aggressive framing and material misrepresentation is smaller than most founders think.
Every investor who read the TechCrunch piece is now discounting every ARR claim by default. The founder who shows the honest number, labeled correctly, with the churn and the concentration attached, is the one who stands out. The smaller, honest number sells much harder than a fake hockey stick.
