90% fail hides where startups really die

An investor leans back and says, "You know 90% of startups fail, right?"

You nod. Because what else are you supposed to do? But here's the thing: that number doesn't actually help you. You can't budget with it, can't plan hiring with it, can't decide how much to raise with it. It's just ambient dread floating around the room.

The Core Problem: "90% Fail" Is a Vibes Number

The real risk isn't a smooth 90% chance of death spread evenly across all startups like some kind of existential coin flip. Risk spikes at specific stages—pre-seed to seed, seed to Series A, A to B, B to profitability. And it looks totally different in SaaS versus biotech versus hardware versus fintech.

If you don't model that curve for your stage and your vertical, you're flying blind.

Failure Is Lumpy: Think Survival Curve, Not One Big Coin Flip

Imagine 100 companies that raise a proper seed round in a given year. What actually happens next? Based on recent US and Europe data, the breakdown looks something like this:

Seed to Series A: roughly 30–40% make it. Series A to Series B: about 40–60% of those make it. Series B to C / growth: around 60–70% of those make it. Beyond that, some go public or get acquired, and some just stall out completely.

But here's where it gets interesting. Add the step before that:

Idea / pre-seed to real seed round: easily 70–90% never get to institutional seed at all.

So the actual curve looks something like this: 100 "serious" attempts at a startup, maybe 10–30 raise pre-seed, then 10–20 raise seed, then 3–8 raise Series A, and finally 1–3 get to a meaningful exit.

Compress all of that into "90% fail," and you lose the information you actually need.

At pre-seed, your biggest risk is: "Can we even build something people use and convince anyone to fund the next 18 months?" At seed, your biggest risk is: "Can we show enough traction to look like the top 30–40% that raise an A?" At Series A, it shifts to: "Can we prove repeatable sales and sane unit economics before we run out of cash?" At Series B+, it's: "Can we scale without everything breaking, and can we keep justifying growth-stage valuations?"

So don't ask, "Will we be in the 10% that survive?" Ask, "What percentage of companies from this stage make it to the next stage, and what did they prove along the way?"

That's tractable.

Your Vertical Has Its Own Mortality Rate

Even at the same funding stage, risk isn't uniform. Roughly how different verticals behave:

SaaS (especially B2B): Lower upfront capex, faster feedback, earlier revenue. Seed to Series A usually wants to see things like $20–80k MRR, growing 8–15% month-over-month and some proof of repeatable sales—not just founder heroics. Survival curve: lots of small deaths early; if you get to real revenue plus net retention, odds improve fast.

Marketplaces: Hard mode early, easier later. Need both sides of the market plus liquidity, which can take 2–3 years. Investors often want a clear wedge (specific niche where liquidity is real) and strong engagement or take-rate, even if GMV is still modest. Many marketplace startups die before seed or between seed and A, when they can't get liquidity fast enough.

Fintech: Regulated, higher trust bar, often higher CAC. Big risk pockets: licensing, compliance, fraud, and banking partners. Survival heavily depends on getting regulatory and banking relationships right and not blowing up unit economics with fraud or chargebacks. Failure might come late—after raising several rounds, when a risk model breaks.

Hardware / Deep Tech: Huge upfront R&D and manufacturing cost. Long timelines. Pre-seed and seed are often about technical feasibility, not revenue. Many of these "failures" are actually: "tech didn't work / too slow / too expensive" after $5–20M invested.

Biotech: Almost no revenue for years. All about milestones: target validation, preclinical data, Phase I/II, etc. Survival curve is tied almost entirely to trial results and regulatory gates, not "got to $1M ARR."

All of these could show up in a generic dataset as "startups." But a pre-seed SaaS founder and a Series A biotech founder do not share the same risk profile in any useful sense.

So when you hear "90% fail," translate it to: "Some unknown mix of SaaS, biotech, crypto, hardware, and random apps did a bunch of different things and mostly died at different times for different reasons."

Not super helpful.

Plan Around Milestone Risk, Not "Startup Risk"

The useful unit of analysis isn't "company." It's milestone.

Companies tend to die right after they miss one of these:

Product validation: People actually use it and come back.

Go-to-market repeatability: You can acquire and close customers in a way that isn't purely founder hustle.

Unit economics: Gross margin, payback period, and churn aren't obviously broken.

Regulatory / technical proof: Approvals, trials, safety, uptime, whatever your world requires.

Follow-on financing: You hit enough of the above to convince the next stage of capital.

So your planning question becomes: "What is the next proof point we must hit to be fundable or self-sustaining, and how much time and money do we need to get there with a margin of error?"

That's your runway design.

A practical way to do this:

1. Name the next 1–2 financing or revenue milestones. Seed: maybe it's "$30k MRR with 8–10% MoM growth" or "10 design partners with clear ROI stories." A: maybe it's "$1–2M ARR, with >70% gross margin and early payback visibility."

2. Work backwards to the leading indicators. How many users, pilots, sales conversations per month does that imply?

3. Add a 3–6 month buffer. Because things always slip.

4. Raise (or cut burn) to cover that window.

And instead of asking VCs, "How risky is my startup?" ask:

"In your last fund, what percentage of seed checks in SaaS/fintech/etc. raised a Series A?"

"What did the winners actually look like at the next round?"

"How long did it typically take them to get there?"

That gives you a survival curve you can act on.

So What?

Ignore the 90%.

Build your own survival model:

• By stage (where do companies like mine usually die next?)
• By vertical (what milestones actually matter here?)
• By milestone (what proof do we need in the next 12–24 months, and do we have enough runway to get it with slack?)

You can't control whether "startups fail 90% of the time." You can control whether you give yourself a real shot at the next step on your curve.