Neider says companies without cash profits brag about page views, downloads, or token maxing, a metric that ignores efficiency.
Neider says older companies are evasive about AI impact, while AI-first startups have upside-down unit economics reliant on other LLMs.
He says examine EBITDA minus legitimate add backs and capex, citing a company that was $58m cash EBITDA and arguably $25m after capitalized expenses.
How investors spot numbers that don't add up
Lead Edge Capital's Brian Neider wrote The Hierarchy of Metrics Bullshit after his firm kept finding trickery in company numbers. The premise is that a company with cash profits talks about cash profits, while one without them brags about page views or a best place to work award. The episode shows what that looks like in EBITDA, in incomplete diligence answers, and in AI-first companies with upside-down unit economics.
Takeaways
Companies brag about the metrics they lack. A business with cash profits talks about cash profits. One without them reaches for page views or a best place to work award. Token maxing is the current version, since counting tokens induces spend without any look at efficiency.
EBITDA gets dressed up to clear committees. One company showed up as $58m of cash EBITDA, $38m of EBITDA and arguably $25m after capitalised expenses. Neider strips legitimate add backs and capex, and calls recurring costs dressed as one-time, like recruiting fees or M&A legal work, unjustifiable.
Partial diligence answers are close to dishonesty. Teams handed ten or twenty diligence items give back eight, usually omitting the one that matters, and older companies are most evasive about AI. Neider's method is to say he does not understand and ask why until the story is rebuilt.
AI-first companies run upside-down unit economics. They share revenue and token counts but carry heavy cost structures and lean on other LLMs to run the business, setting those costs aside on a pro forma basis in the hope that open source or cheaper models arrive later.
Quality and entry price both set returns. Neider frames a good investment as company quality meeting valuation, like a house among $1m comparables listed at $10m, an A-plus company at an F-minus valuation. Lead Edge's bar is B-plus, and an F-minus asset at an A-plus price can work if you do the work.
AI risk splits software by category. Vertical software for one end market, such as AppFolio in real estate or Procore in construction, is insulated by nuance and know-how. Horizontal applications for functions like accounting or HR are more exposed, and developer tools are most at risk, while observability and cybersecurity benefit.
The best assets are sole sources of truth. Rentrak is the only box office data source, accepted as neutral ground by theatres and studios whose release decisions depend on it. Clearwater holds the same position on institutional ownership, where trust and distribution carry the value. Electric sells scarce medium voltage parts into data centres.
The $50m Happy Gilmore check
Tap a name to go there · tap a line to see why they are linked
On this episode 2
Show 1
Neider was told not to bring a term sheet to Duo Security.
He had a giant Happy Gilmore style check for $50m delivered to a friend's apartment.
He took an Uber double XL so it would fit in the trunk.
Lead Edge became the first call and largest investor in Duo's next round.