Why Do Businesses Fail? 38 Documented Failures, Cause by Cause

We keep 38 documented failures — businesses that launched, were written up, and did not work — out of 406 entries in our index. National failure percentages come from government data like the U.S. Bureau of Labor Statistics Business Employment Dynamics survival series; what we add is the cause-by-cause detail those percentages hide.

A percentage tells you the odds. It does not tell you which mistake ends which business — the only part you can act on.

Table of contents

• ·Why do businesses actually fail?
• ·What the official failure rates measure — and what they miss
• ·What actually killed the ones we documented
• ·How much is each of these numbers actually worth?
• ·Why a list without failures is worse than useless
• ·How to use this without fooling yourself
• ·Frequently asked questions

Why do businesses actually fail?

They fail for specific, nameable reasons: no pain worth paying to fix, churn eating the top line, a compliance shortcut that ends the channel, or a skill whose window closed. We document 38 such cases out of 406 indexed businesses, 23 of them one-person operations.

Only 6 of the 38 disclose a clean monthly figure. Their median is $8K/mo, range $29/mo to $400K/mo — the first surprise. Failure is not the same as earning nothing.

The rest disclose nothing financial, and we do not estimate for them. What they contribute is mechanism: what was built, what happened, what ended it. Grading rules: how we grade the evidence. Full set: all indexed ideas.

What the official failure rates measure — and what they miss

Official survival statistics come from BLS Business Employment Dynamics and the SBA Office of Advocacy small business FAQ. They track establishments that stop reporting employment: right for the odds, wrong for the reason.

The flattening problem is severe. Two businesses both counted as "failed" can be opposites: one ran out of money after spending $35,000 to find one customer; another peaked at ~$8,000/month and churned back down. Same bucket, completely different lessons.

Establishment-based data also misses most of what we index. A solo app that stops earning was never an establishment with payroll, so it never appears in the survival series, and 23 of our 38 failures are one-person businesses. Hence the division of labour we suggest: government data for how often, our 38 case files for how.

What actually killed the ones we documented

Six named cases: what was built, what it earned, what ended it. Most directories lack this section, because most directories delete their failures.

LeadLedger is the cleanest post-mortem here. Product one, an in-app email-capture SDK, sold exactly one licence at $300/year to a developer in England. Product two, a Facebook Lead Ads to CRM connector, priced at $29/month, $99/year or a $199 lifetime licence, sold off a single webinar. Total build cost about $35,000, split $15,000 and $20,000. The founder's diagnosis: nobody had a pain they would pay to fix.

Yorbie is the other shape — a peak, not held. ~$8,000/month gross volume in March, ~$6,700/month four weeks later, and $0 revenue for the first two months after the founder quit his big-tech job. Churn, acknowledged on camera.

The 97% audit supplies the scale figure: of 1,000+ tracked vibe-coded apps, 97% were dead, abandoned, breached, or under $500/mo after six months (creator-relayed). The survivors are dull — $1,400/month per shop for HVAC routing software. And Chai is the counter-case: $30M/yr, 21 people, a $450M valuation, from a founder who could not afford the GPT-3 bill and crowdfunded it — 80K sign-ups in one day crashed the servers.

How much is each of these numbers actually worth?

The evidence split across these 38: 5 third-party verified, 21 founder-reported, 6 creator-relayed, 6 unproven. Verified plus founder-reported is 68% — higher than the index average, because people describing their own failure have less reason to inflate.

This is the argument the whole site rests on. A founder-reported figure and a third-party verified figure are not the same claim, and quoting them identically launders the weaker one.

Cleo is the demonstration. $60K MRR in 53 days, unproven, published in a video that is itself a link in the team's launch funnel; their earlier product, Mentions, did $20K MRR in month one. Compare the AI faceless Facebook content page at $32.5K/mo on 314M views in 28 days — third-party verified, hard data. Both are numbers. Only one is a fact. Across the whole index the split is 57 verified / 184 founder-reported / 121 creator-relayed / 44 unproven.

Why a list without failures is worse than useless

Our index is 406 businesses and 38 are cautionary tales we keep deliberately. A directory that quietly deletes its failures is not neutral — it is selling survivorship bias while calling it research. What you lose when a list shows only winners:

• ·The base rate disappears. Without the 97% dead-app figure, "build an app" looks like a plan rather than a lottery with a known ticket price.
• ·The mechanism disappears. LeadLedger's $35,000-for-one-customer is a question you can test against your own idea. A winners-only list gives you nothing to test.
• ·Revenue starts to look like safety. Yorbie hit ~$8,000/month and still ended at $0 for two months. Rook made $47K in two months on a model with shutdown risk built in.
• ·The honest caveat is hidden. This index over-represents businesses that succeeded and chose to disclose. The $27K/mo whole-index median is a comparison tool, not an expectation.

For category context, see What Is a SaaS? Definition, Business Model, and Real Revenue Data and What Is a SaaS Company?. More failure analysis in the Risks & Rules hub.

How to use this without fooling yourself

1. 1.Run the LeadLedger test. Does the buyer have a pain they would pay to fix, or something merely useful to have? Getting that wrong cost $35,000 and one $300/year customer.
2. 2.Model churn, not just the peak. Yorbie's founder-reported drop from ~$8,000/month to ~$6,700/month in four weeks is what a plateau looks like from inside.
3. 3.Refuse unproven figures outright. 6 of these 38 are unproven, including Cleo's $60K MRR in 53 days published inside its own launch funnel. If nobody outside the business confirmed it, do not plan on it.
4. 4.Check whether the channel survives scrutiny. Rook's $47K over two months depended on leaving off the AI disclosure. Then write the plan — the SBA's business plan guide is a free structure.

Frequently asked questions

Why do 90% of businesses fail?

We do not publish that percentage and will not repeat it without a source. National survival rates come from BLS and SBA data. We contribute 38 documented failures out of 406 indexed businesses, with named causes — plus one relayed benchmark: 97% of 1,000+ tracked vibe-coded apps were dead, abandoned, breached, or under $500/mo after six months.

What is the main cause of business failure?

Building something nobody has a paid pain for. LeadLedger spent about $35,000 across two products and sold exactly one licence at $300/year for the first. Churn is second — Yorbie peaked at ~$8,000/month and fell to ~$6,700/month within four weeks.

What are the 8 reasons why people fail in the business?

We do not have a numbered list of eight, so we will not invent one. From these 38 failures, four causes are clearly evidenced: no paid pain (LeadLedger), churn after a peak (Yorbie), a compliance shortcut that ends the channel (Rook, $47K over two months), and positioning around prevention rather than outcome (the NoFap app, $6K/mo).

How reliable are these why do business fail figures?

More reliable than most groups we index: 5 third-party verified, 21 founder-reported, 6 creator-relayed, 6 unproven — 68% verified or founder-reported, against 59% across the whole index. Founders describing their own failures have less incentive to inflate, though 6 entries remain unproven.

How many of the 38 indexed businesses actually disclose a monthly number?

6 of 38. Their median is $8K/mo and the range runs $29/mo to $400K/mo — failure and zero revenue are not the same thing. The other 32 contribute mechanism rather than figures, and we do not estimate numbers they never stated.