Every company you've funded came with a financial model and a cap table somebody on your team pulled apart line by line. Fewer came with anyone testing whether the CAC number, the channel mix, or the growth curve behind the ask was real, or just a confident story.
That gap isn't a theory. A widely cited academic survey of practicing VCs found that 9% use no quantitative deal-evaluation metric at all, and 20% don't even forecast cash flows before investing. The team gets 95% of the attention. The growth story gets whatever's left.
of VC firms don't forecast cash flows before they invest. (Gompers, Kaplan & Mukharlyamov, Harvard Business School / NBER)
Commercial due diligence for PE buyouts is a mature, heavily outsourced industry: roughly 36,000 engagements run per year, and mid-market PE firms spend an average of $340,000 per deal on it, mostly through external specialist firms. That's not the world a seed-to-growth VC round lives in.
Builder.ai (the clearest global anchor, collapsed 2025): Builder.ai raised more than $450 million at a peak $1.5 billion valuation, backed by Microsoft, the Qatar Investment Authority, SoftBank's DeepCore, and the IFC, on the claim that its "Natasha" AI engine could build apps automatically, "as easy as ordering pizza." In reality, roughly 700 engineers in India were writing the code by hand behind the interface. The company claimed $220 million in 2024 revenue; a post-collapse audit put the real figure closer to $50 million, an overstatement of roughly 300%. Bloomberg's reporting also surfaced an alleged "round-tripping" arrangement with Indian ad-tech firm VerSe Innovation: roughly $60 million in mutual payments between 2021 and 2024 that both sides booked as revenue (VerSe denies wrongdoing, though it has since restated its own FY23 revenue down by about 24%). None of this surfaced in time to stop the raises. It surfaced when a new CEO opened the books in early 2025, after which a lender seized $37 million from the company's accounts and the business filed for insolvency across three countries.
Financial modeling and cap-table review are now standard practice on almost every round, regardless of size. Builder.ai had audited financials and marquee institutional investors through multiple rounds, and still ran for years on a product claim, an AI that writes code, that nobody outside the company seems to have tested against the actual engineering team. Commercial diligence isn't standard anywhere in this market. Not because funds don't want it, but because even the specialists who build tools and run practices specifically to do this well say the discipline itself is still immature: GTM datasets need specialist interpretation most deal teams don't have, reports get built under days of time pressure, and, in the words of one commercial-diligence advisory to PE firms, "a vendor who confirms the seller's story and hands you a deck of pipeline charts has increased your risk while making you feel better, which is the worst possible outcome." Worth noting too: even the 73% diligence-failure research above doesn't track "commercial/GTM forecast risk" as its own category, which is closer to the gap itself than a stat about it.
Microsoft, a sovereign wealth fund, and a development-finance institution all backed Builder.ai across multiple rounds. None of them appear to have asked the one question that would have unraveled it: who is actually writing this code? That's not a knock on any one fund. It's the actual state of the discipline.
...and the $500K deal where nobody needed to commit fraud at all.
GoMechanic (still the strongest India anchor): GoMechanic raised more than $55 million from institutional investors on inflated sales figures and underreported costs. KPMG and PwC both audited the company in prior years and reported no discrepancies. SIDBI, a government-backed lender and LP to several of the funds involved, is now questioning those funds' own diligence process directly.
Medial (Shark Tank India, smallest end of the scale): Medial, a startup networking app, got a deal on Shark Tank India from OYO founder Ritesh Agarwal. Its investor projected 65–70% subscriber renewal rates and profitability within 12 months, based on the startup's own numbers. The app reached 5 lakh users with claimed "10X" growth. It shut down in 2026, having never found a working monetization channel. No fraud here, nobody fabricated anything. The forecast was just never stress-tested, and it didn't hold.
Bain & Company, India Private Equity Report 2026: more deals, each needing diligence that scales down with the check size rather than disappearing below some threshold.
GoMechanic shows this gap can hide behind two clean financial audits at $55 million. Medial shows it doesn't even need fraud to matter at $500,000, an honestly optimistic forecast that nobody pressure-tested was enough on its own.
KPMG and PwC checked GoMechanic's books. Nobody checked whether its growth numbers meant what they said. Medial's numbers weren't even fabricated, just never tested. Same gap, two very different ways into it.
Read both signals together and the shape is the same everywhere a deal closes. The financial numbers get checked by someone trained to check them. The growth numbers next to those financials get checked by whoever happens to be in the room. What changes between a mature market and India is how the gap gets found, not whether it exists.
| Dimension | Global | India |
|---|---|---|
| Who typically catches the gap | A down round, a failed next raise, or an LP's own question | A collapse, a whistleblower, or a regulator like SIDBI asking the VC directly |
| Typical failure mode | Forecast never stress-tested (Medial-style), surfaces as "ran out of cash" | Fabricated sales and underreported cost, missed by real financial audits (GoMechanic) |
| What's changing | GTM diligence software and specialist advisories emerging, by their own account still immature | Smaller, more numerous deals raising the stakes on getting this right |
| Scale it shows up at | $500M+ (Builder.ai) down to small seed checks | $55M+ (GoMechanic) down to $500K (Medial) |
| Check | What it actually verifies |
|---|---|
| Rebuild CAC from the general ledger, not the deck | Whether personnel, tools, and agency cost are already included, or quietly excluded |
| Ask for attribution by model, not by channel | Whether the reported channel mix survives a shift away from last-click |
| Test channel concentration at the account level | Whether revenue-driving traffic sits on one channel the company doesn't control |
| Independently verify the top 20 logos in the customer list | Whether they're active, paying, and at the stated contract value |
| Request cohort retention by signup month, not blended | Whether blended retention is hiding a newer cohort decaying faster than average |
| Check who reviewed the growth numbers, not just the financials | Whether anyone with commercial expertise reviewed the story, or only accountants reviewed the ledger |
Builder.ai, GoMechanic, and Medial look like unrelated stories, a $450M global AI company, a $55M Indian auto-services startup, and a $500K Shark Tank deal. Pulled apart, they're the same failure at three different scales.
None of the funds in any of these three deals were careless by the standard most rounds get held to today. That's the actual problem this newsletter is about.
Think about the last deal you actually funded. Can you name, right now, who outside that company tried to break its growth forecast before the money moved? If you can't answer in five seconds, you already have your answer, and so will every other fund that's about to find out the hard way.
Sources
Before You Go
If this matches what you're actually seeing in your own deal process, or if it doesn't, I'd like to know. This isn't a funnel, there's no call to book here, just a straight line to my inbox.