Which Startups Actually Die? A Death-Rate League Table by Sector

The most useful thing a founder can look at before building isn't a trends report or a market-size slide. It's a graveyard. Which kinds of companies died, how often, and — if you're willing to sit with it — why. So we built the graveyard.

We took more than 6,000 Y Combinator companies, kept the 2010–2019 batches (old enough that the story has finished playing out), and ranked them by the one outcome nobody puts on a pitch deck: how often, sector by sector, the company was simply dead.

What is the startup death rate by sector?

Here is the league table, worst to best, for accelerator-backed companies from those batches. "Dead" means no longer operating, drawn from public records:

• Social — ~58% dead
• Consumer — ~43%
• Consumer health & wellness — ~42%
• Retail — ~40%
• Home & personal — ~39%
• Content & media — ~35%
• Productivity — ~34%
• Engineering & design tools — ~33%
• Food & beverage — ~32%
• Real estate & construction — ~30%
• Industrials — ~30%
• B2B — ~27%
• Healthcare — ~27%
• Infrastructure — ~25%
• Marketing — ~23%
• Fintech — ~21%
• Healthcare IT — ~20%
• Education — ~19%
• Recruiting & talent — ~15%

Top to bottom, that's a four-fold spread. A social company died nearly four times as often as a recruiting company from the same vintage, out of the same programme, funded on the same terms. That is not noise. When a gradient is that wide and that orderly, something structural is producing it — and it's worth knowing what, because you can see it before you write a line of code.

Why do some sectors die so much more than others?

Two forces do most of the work. Neither is a secret. Both are visible from the outside.

The first is distance from a signed contract. Read the top of the table and it's the consumer-facing world — social, consumer, retail, home and personal. Read the bottom and it's the machinery a business pays for on a contract — recruiting, fintech, healthcare IT, infrastructure. A company that a business pays under contract has received confirmation, in the only language that counts, that the problem is real. A consumer who taps "download" on a free app has confirmed almost nothing. Enterprise revenue is a lagging proof of demand; consumer traction is a leading guess at it. Guesses are wrong more often, and being wrong about demand is how most companies die. (I wrote the B2B-versus-consumer version of this argument here; the death table is the same force, seen at higher resolution.)

The cleanest proof is inside a single domain. Take health. Healthcare IT — software sold to providers and payers — died at about 20%. Consumer health and wellness — the same broad problem sold to individuals — died at about 42%, more than twice as often. Same domain, same year, same accelerator. The thing that changed was who signs the cheque. Who signs the cheque predicts survival better than the domain does.

The second force is defensibility of the wedge. Distance from a contract isn't the whole story, because a few consumer-adjacent categories sit lower than the contract logic alone would put them, and a few tool categories sit higher. What separates them is how easily the wedge gets commoditised or absorbed. Social and content compete for attention against platforms that can copy a good feature in a sprint and reach a billion people by Tuesday — the wedge evaporates. Recruiting, fintech, and education, by contrast, accumulate switching costs, compliance moats, and workflow lock-in that make them annoying to rip out once installed. Annoying-to-remove is a survival trait. It's not glamorous, and it is why the least glamorous category on the list — recruiting software — is the one still standing.

So should I just build in a low-death sector?

No — and reading the table that way is its own mistake. The companies at the bottom of the death list weren't founded by people optimising a percentage. Build recruiting software you don't care about because a spreadsheet said it survives, and you'll discover the death rate you were dodging has a cousin: the slow, quiet kind, where the company never dies and never matters either. Different failure mode, same origin — a founder who chose the market instead of being chosen by the problem.

What the table actually buys you is an honest read on the terrain. A high death rate isn't a stop sign; it's a warning that the market hasn't confirmed demand for you yet and won't do it gently — so build like someone who knows that, with faster reads on whether it's working and no borrowing of the enterprise playbook that doesn't apply to you. A low death rate isn't a green light; it's a note that demand here has already been confirmed by the companies ahead of you — which means the survival is real and the crowding is too. The number describes the ground. It does not describe your company. That part is still yours to prove.

How much can this data actually claim?

The caveats are load-bearing, so here they are. Every company in this table was admitted to Y Combinator — roughly one applicant in a hundred. These are death rates for startups that already cleared an extreme filter, not for startups in general; the true mortality of the ideas that never got in is far higher and invisible here. "Dead" is a public-record status, not an audited one. We used older batches on purpose, because the recent ones haven't resolved — which makes this an honest map of 2010–2019 and a silent one about whether the same gradient holds in an AI-reshaped 2026. The forces underneath it — confirmed demand, defensible wedge — are older than any batch, so my guess is the shape survives even as the sector labels churn. But it's a guess, and I'd rather tell you it's a guess than dress it as a law.

Frequently asked questions

Which startup sector has the highest failure rate?

Among accelerator-backed companies in Y Combinator's 2010–2019 batches, social startups had the highest death rate at roughly 58%, followed by consumer (~43%), consumer health and wellness (~42%), and retail (~40%). The consumer-facing categories cluster at the top of the failure table.

Which startup sectors survive the most?

The lowest death rates were in recruiting and talent (~15%), education (~19%), healthcare IT (~20%), and fintech (~21%). These are contract-based categories with high switching costs — businesses pay for them under contract and find them costly to replace, both of which correlate with survival.

Why do social and consumer startups fail so often?

Two forces. First, free consumer usage is a weak, leading signal of demand, whereas a signed business contract is a strong, lagging confirmation of it — so consumer companies are wrong about demand more often. Second, social and consumer wedges are easily commoditised or absorbed by large platforms, so even a working product can lose its defensibility quickly.

Does a low death rate mean I should build in that sector?

Not on its own. A low death rate signals that demand in that category has already been confirmed — which also means it's crowded — and it says nothing about your specific company. The strongest companies are built by founders drawn to a problem, not by founders optimising a survival statistic. Use the table to understand the terrain, not to pick your market.

These figures come from NUVC's accelerator outcome graph — more than 6,000 Y Combinator companies and every Startmate cohort, each linked to what happened next. Outcome data like this is what keeps a pitch-deck score honest: any model that has only studied the survivors has learned what lucky looks like, not what good looks like.