Last Updated on July 7, 2026 by Taya Ziv
Last month, Cluely CEO Roy Lee told TechCrunch his company’s annual recurring revenue had doubled in a week to $7 million.
Then, on X, he said he “told her some BS.” The real number was $5.2 million.
An Andreessen Horowitz-backed CEO. Lying about revenue. On the record. And then admitting it like it was a parking ticket.
But what actually bothered me about this story wasn’t the lie. It was the reaction. Or the lack of one. No investor pulled their money. No board member resigned. A16z stayed quiet. The startup world collectively shrugged and went back to scrolling.
And that tells you everything about where we are right now.
ARR Used to Mean Something
For two decades, Annual Recurring Revenue was the gold standard of startup health. Clean. Predictable. If a SaaS company told you they had $10M ARR, you knew what that meant: $10 million in contracted, repeating subscription revenue coming in every year.
Bloomberg just called ARR “Silicon Valley’s least trusted metric.” And they’re not wrong.
The problem started when AI companies tried to use a metric built for subscription software to measure something that doesn’t work like subscription software. AI products don’t have neat annual contracts. They have usage-based pricing, token-based billing, pilot programs that may or may not convert, and “partnerships” that are really just one-time enterprise deals counted twelve times.
So what counts as “recurring”? A three-month pilot? A usage-based customer who might churn next week? A free tier user who upgraded for one month?
There are no SEC definitions for ARR. No audit requirements. No industry standards. The number can mean whatever the founder needs it to mean during a fundraise.
And that’s not a bug in the system. That IS the system.
The Problem Is Bigger Than One Lying CEO
Roy Lee is the headline. But he’s not the story.
The story is that the entire AI startup ecosystem is running on a metric that has no agreed-upon definition, no oversight, and no consequences for getting creative. Lee just happened to be dumb enough to say it out loud.
Think about what’s happening in AI funding right now. Q1 2026 saw $297 billion in global startup funding. OpenAI alone raised $122 billion, pushing its valuation to $852 billion. Anthropic, xAI, and a dozen others raised billions more.
All of these valuations are built on ARR multiples. But if ARR itself is unreliable, then the valuations are unreliable. And if the valuations are unreliable, then the entire funding stack is basically a collective agreement to not ask too many questions.
Which, if you’ve been around long enough, sounds familiar. We’ve done this before. Different decade, different metric, same human tendency to inflate numbers when money is flowing fast.
Why This Matters If You’re Raising (Or Planning To)
If you’re a pre-seed or seed-stage founder, you might think this doesn’t apply to you. You’re not faking $7 million in ARR. You probably don’t have ARR at all.
But it applies to you in two ways that are worth understanding.
First, the investors you’re pitching are getting burned by inflated metrics at the growth stage. That skepticism trickles down. When a seed investor asks you for “evidence of traction,” they’re partly reacting to the fact that they got burned by a Series B company whose ARR turned out to be creative accounting. Your job just got harder because of someone else’s lie.
Second, it changes what you should actually measure and present. The smartest founders I work with have stopped leading with ARR entirely. Instead, they lead with activation metrics: how many users completed the core action in the first 48 hours. Retention cohorts: what percentage came back after week 1, week 4, week 8. And the most powerful one, willingness to pay. Not “they signed up for a free trial.” But “they entered their credit card, and the charge went through, and they didn’t cancel.”
Boring? Absolutely. But boring metrics are the ones that actually survive due diligence.
Revenue Quality Over Revenue Quantity
I keep coming back to one concept that I think separates the founders who raise successfully from the ones who get ghosted after the second meeting: revenue quality.
It’s the difference between a company with $5M ARR from 3 enterprise customers (one of whom is the CEO’s college roommate) and a company with $2M ARR from 200 self-serve customers with 95% monthly retention.
The second company is worth more. Every serious investor knows this. But ARR as a headline number doesn’t capture it.
What does capture it:
Net Revenue Retention (NRR). Are existing customers spending more or less over time? An NRR above 120% means your product is so good that customers expand without you selling them anything new. An NRR below 90% means you’re filling a bucket with a hole in it.
Payback Period. How many months until a customer’s payments cover what it cost to acquire them? If the answer is 18+ months and you’re pre-Series A, your unit economics have a problem you’re not facing.
Revenue Concentration Risk. What percentage of your revenue comes from your top 3 customers? If it’s above 40%, one phone call can cut your “ARR” in half.
None of these are as exciting as a big ARR number on a pitch deck slide. But they’re a lot harder to fake. And in a world where investors are increasingly aware that even $122 billion companies can build products nobody actually wants to use, “hard to fake” is becoming a competitive advantage.
What I’d Do If I Were Raising in 2026
Stop chasing ARR as a vanity number. I know that sounds counterintuitive when every VC blog post and Twitter thread still references it. But the smart money is already looking past the headline.
Build a revenue quality dashboard before your pitch deck. Put NRR, payback period, and revenue concentration on one page. If those numbers are strong, the ARR number matters less. If those numbers are weak, a big ARR number won’t save you anyway.
Be the founder who says the real number. In an ecosystem where everyone’s inflating, honesty is a genuine edge. I’ve seen founders close rounds specifically because the investor said “you’re the first person who gave me a number I actually believe.” When trust is scarce, the person who offers it first gets rewarded.
Track activation, not sign-ups. The most dangerous metric in early-stage isn’t low revenue. It’s high sign-ups with low activation. It looks good on paper and means nothing in practice. It’s the startup equivalent of getting a standing ovation for a movie trailer and then watching the theater empty out when the actual film starts.
Show the cohort, not the snapshot. A single month’s revenue number is a photograph. A 6-month cohort chart is a story. Investors fund stories they believe in, not photographs they can’t verify.
Maybe I’m wrong about ARR losing its crown completely. It’ll probably stick around in some form because investors love simple numbers and founders love showing big ones. But the version of ARR that meant “we have a healthy, growing, repeatable business” is being drowned out by the version that means “we did some math that makes our fundraise look better.” And in a market where companies are being valued at $10-25 million per engineer regardless of what their revenue actually looks like, maybe the metric was always more fiction than fact.
The founders who’ll raise successfully in 2026 aren’t the ones with the biggest ARR number on their deck. They’re the ones whose numbers mean something when someone actually checks.
