Anatomy of 11 Closures — Platforms and "Profitable but Unsustainable" Kill More Small Businesses Than Running Out of Money

This site’s published cases lean heavily toward businesses that succeeded. Out of 392, only 11 could be recorded as closed or withdrawn, under 3%. That is far removed from the real mortality rate of businesses, and this very skew is the survivorship bias a reader of revenue-case media should correct for first. Even so, lining up the causes of death across these 11 reveals a distribution that differs from the folk wisdom.

The 11 cases

Classifying the causes of death

Sorted by “why it ended,” the 11 fall into roughly five groups.

1) External rules and platforms (2 cases). The Reddit research tool closed while profitable, after commercial-license negotiations over the Data API fell through. The AI article-writing SaaS collapse was likewise triggered by a $50,000 fraud loss and payment-related terms changes. In this group the cause of death is not revenue or demand but rules held by another company, plus outside malice, and success numbers like $35K or $15K MRR provided no defense.

2) Market and demand never materializing (3 cases). The health AI reached 4 paying customers in 7 years. The SEO tool peaked at 1,000 users on beta day one. The Go game ran 10 years with zero revenue. All are the “could build it, nobody would pay” type, where the fatal wound is not a single blow but the prolonged stretch of not growing. Note the years to withdrawal, 7 and 10 years mean most of the loss accrued not in money but in opportunity cost.

3) P&L and time-for-money judgment calls (3 cases). The rental spaces locked in roughly ¥2.2M of losses over 29 months and withdrew. The Python automation side business was earning ¥94K a month and was folded anyway, judged not worth the hours. The logo generator ran stable at $3,000/month for 5 years and was ended over the structural fatigue of a pay-once model that demanded 100 new buyers every month. In this group, a small profit, not only a loss, is grounds for exit. The cost of continuing sits outside the books.

4) External shocks and accidents (2 cases). The food truck had just reached ¥380K/month when COVID movement restrictions forced it to stop. The direct trigger that ended the fan-game platform with 1M monthly UU was a code accident. One line capping query results (.limit(20)) went missing. But the essence of the latter is not the accident. With no revenue, the service had to survive on an ¥8,000/month server, and the option of paying for scale simply did not exist the moment load spiked. For a popular service with no revenue, the first outage tends to be the last.

5) Quitting while ahead (1 case). The AI avatar generator earned $150K in its first week, watching capital-rich competitors turn the market into an ad-spend war, the founder shut it down early by choice. Not even a failure, an ending rarely recorded: stepping off while winning.

The numbers behind the endings

Across the 8 cases where monthly revenue at closure is known, the distribution spans ¥0 to about ¥5.25M (MRR $35K). At least 3 of the 11 were profitable or high-earning at the moment they closed. The intuition that collapse means the money ran out is, in this sample, the minority pattern.

Operating spans run from 4 months to 10 years, with the median in a band of roughly 2–4 years. Fast deaths cluster in the external-shock group. Slow deaths in the no-demand group. Speed of withdrawal measures not the size of the failure but the speed of the decision. Ten years at zero revenue likely cost more, opportunity cost included, than cutting losses at ¥2.2M after 29 months.

What generalizes, and where it stops

What these 11 support generalizing: 1) a business standing on another company’s rules becomes a negotiation target as its success numbers grow, its defenses weaken rather than strengthen (a structure seen repeatedly in the platform-risk cases), 2) exit criteria should be set in advance to include not just loss amounts but time-for-money return and structural fatigue, 3) “folding while profitable” is not failure but a decision.

The limit lies in the sample. These 11 are confined to operators honest enough to publish their withdrawal. The great majority that vanished quietly do not appear here. The 3% share of closed cases is not a real-world mortality rate but a measurement of disclosure bias. Acknowledging that skew, this site will keep prioritizing records of withdrawal. For the overall distribution see the data page. For the other exit, selling, see the list of exit cases.