Startups That Failed After Making Money: 38 Cautionary Tales With Receipts

ProvenStartups files 38 cases as cautionary tales, and the useful ones are not the businesses that never worked — they worked first. Six document a peak and then a fall you can date. One documents the opposite mistake: selling at $10M ARR while growing almost 100% a year.

Contents

• ·How much did these startups make before it went wrong?
• ·Which failed startups made the most money?
• ·What actually kills a business that already makes money?
• ·Who should not copy any of this?
• ·Frequently asked questions

How much did these startups make before it went wrong?

Real money, at every scale. Calming Blankets did over $10M in sales and $1.5M profit in year one 🗣. Botkeeper passed $1M eight months into year two 🗣. 27 Talent billed about £550K in 2023 and again in 2024 🗣. Revenue prevented none of what followed.

Grades matter more here than anywhere else in our risks and rules hub: ✅ an outside source confirmed the figure, 🗣 the founder said it on the record, 📎 a creator relayed it, 🔮 a price tag and nothing else. This set is 5 ✅, 21 🗣, 6 📎 and 6 🔮 — heavily self-reported, because collapses rarely have outside witnesses.

Cautionary Tale is an editorial label, not a death certificate: several of these businesses are alive and simply carry an exposure worth seeing (public subset here). Eight publish no money figure at all; the table takes the strongest fifteen of the rest. Base rates live in the BLS survival series; these cases add the mechanism, cause by cause in why businesses fail.

Which failed startups made the most money?

The biggest number belongs to a business that has not failed: Chai, at $30M/yr with 21 people and a $450M valuation ✅, filed for the gray-zone content it runs on. The largest real collapse is Botkeeper — $100M+ raised, 500-600 people at peak, then a third of revenue gone in one quarter 🗣.

Sorted by evidence grade, then case rank. The third column names the fall where there was one, the documented exposure where there was not.

Do not average that column: Chai's $30M/yr is a funded outlier, and 97% is a failure rate across 1,000+ apps, not revenue.

What actually kills a business that already makes money?

Six modes, and they cost different amounts. Platform dependence costs the whole asset overnight. Concentration costs a third of revenue in a quarter. One channel costs the growth you planned around. Lost focus cost one founder over $1M in twelve months. Pricing costs years. Selling early costs the difference between an exit and a company.

Platform dependence: what happens when the payout is not yours?

You lose the business at the platform's convenience. Adam Lyttle has $1M+ in lifetime App Store revenue and 2M+ downloads across six years 🗣, and his 2026 position is that the Apple developer account is the asset — after developers who put a trending model name in a keyword field were terminated with no warning and no second chance.

The Facebook page earning $32.5K/mo off 314M views in 28 days ✅ is the same bet with a shorter fuse: the payer is a bonus program, not a customer. The audited extreme is the offerwall economy — $0 withdrawn after roughly 40 hours, near $0.33/hour 📎. Read the App Review Guidelines before calling any of it an asset.

Concentration: how few customers is too few?

Fewer than you think. Botkeeper served hundreds of accounting firms and still lost a third of its revenue in one quarter when eight clients cancelled expansions 🗣 — all eight renewed on calendar-year schedules, so one January undid ten years and $100M+ of funding, and the company announced its own shutdown before an acquisition rescued it.

LeadLedger is the miniature: exactly one licence at $300/year against about $35,000 of build cost 🗣. If eight buyers can move a third of your revenue, you have clients, not a market.

One channel: why does a working channel hide a broken business?

Because acquisition says nothing about retention. Yorbie peaked near $8,000/month in March and slid to about $6,700 within four weeks on churn 🗣, after two months at $0. The channel kept working the whole time; the business leaked out the back.

NaturalWrite did $100K in three months off one TikTok account 🗣, and Quick Shorts took $500+ on launch day from a single 2.3M-view Reel 🗣. Both are real; neither proves a second month.

Burnout and focus: what does a second brand cost the first one?

Calming Blankets priced it exactly: $10M+ in sales and $1.5M profit in year one, $2.5M profit in year two, one month at $500K profit — then a loss of over $1M across the following twelve months 🗣. The business never broke; the founder's attention did.

That was after ten more brands, about 100 staff at peak, and roughly an hour a week left for the brand paying for it. He also turned down an offer of over $10M. The tell is dilution, not exhaustion — and it stays invisible for about a year.

Pricing: which price tag caps the ceiling before you notice?

The generous one, set early and never revisited. Botkeeper sold licences at $69 with volume rates below $40 🗣; LeadLedger offered $29/month, $99/year or a $199 lifetime licence 🗣; Topical Map AI opened with a $39 lifetime deal that "made a couple of grand" 🗣.

Cheap pricing does not lose money on day one; it removes the margin a bad quarter eats. Kyle Gawley shows the tax in years: $400 total in 2018, $18,000 across 2019, $5,000/mo in 2020, $10,000/mo by 2022 🗣.

Sold too early: when is the exit the mistake?

When the business is already hard to kill and you cannot see it. EchoSign was at $10M ARR, growing almost 100% a year and cash-flow positive on $8M raised, when its founder sold to Adobe in 2011 for nine figures — then grew it from $10M to $100M inside Adobe 🗣.

DocuSign was worth $10 billion at the time of his talk. Topical Map AI is the small version: ~$120K gross over 18 months and a $10K/mo run rate, then nine months of no marketing and a five-figure exit 🗣. Buyers read the graph before the name.

Who should not copy any of this?

Anyone who cannot name their second channel, their second-largest customer, and their fixed monthly cost — those three answers decide whether a good year survives a bad quarter. If they are "one", "there isn't one" and "I haven't counted", the channel is running the business, not you.

• ·Price so a good month covers a bad quarter. A $39 lifetime deal cannot.
• ·Treat platform payouts as rented revenue; build channel two while channel one still pays.
• ·Do not sell while growth is near 100% and churn is net negative.

More in the risks and rules hub: what the official failure percentages measure, and the 82 vibe-coded apps the 97% audit was arguing with.

Frequently asked questions

Do most startups fail even after they start making money?

Often enough that revenue should never be read as safety. One consultant tracked 1,000+ vibe-coded apps over 13 months and reported 97% dead, abandoned, breached or under $500/month after six months. That is one category and one person's sample; our 38 are curated, not sampled.

What is the most common reason a profitable startup fails?

Concentration meeting fixed cost. Botkeeper lost a third of its revenue in one quarter because eight clients renewed on the same calendar, despite serving hundreds of firms. 27 Talent carried £50-60K of monthly overheads into a January that billed £8,500, then filed for creditors' voluntary liquidation on 19 February.

Is selling a startup early always a mistake?

No. EchoSign's founder regrets a nine-figure exit at $10M ARR while growing almost 100% a year. His earlier company sold for $50M on $6M of first-year revenue, but he had already sold 80% in the seed round. What matters is how much you still own, and whether growth and churn say you are hard to kill.

How much can I trust these numbers?

Twenty-one of the 38 are founder-reported, six are relayed by a creator, and only five are third-party verified. A grade says where a figure came from, not how large or durable it is. Collapse figures are almost always self-reported — nobody audits a shutdown. Read direction and order of magnitude; treat exact amounts as claims.