Our Own Record #1 | We Started Four Businesses and Three of Them Broke — Why This Site Is Publishing Its Own Numbers

This site collects and records numbers that other people have made public: monthly revenue, sale prices, years in operation, acquisition channels. After lining up 343 case studies, one thing started to bother us. We had never once published our own numbers.

We annotate other people’s businesses with “n = how many” and “the median is X” while staying anonymous ourselves. That doesn’t hold up, so we are starting this series. Over four installments, we will record the four businesses this editorial team actually built, starting with the ones that didn’t work.

To be clear up front: none of these four is this publication. They are businesses the writer built through day-job or personal work. They are not Small Start’s own operation. The figures come from each business’s own repository and analytics.

The four businesses, and where they stand

The bottom three were all started within the same 17 days in late May 2026. Same period, same person, same enthusiasm. Two months later, only one was still moving.

About withholding the names

Let us state the unfavorable condition up front: none of the four businesses will be named in this series. For the one still operating, we are choosing the timing of the reveal separately. For the ones that stopped, other parties are involved.

This is a bad deal for the reader. The condition this site imposes on every published case, “follow the source links and you can verify the numbers”, is one our own articles fail to meet. We write anyway, knowingly. The numbers live in places you cannot see from outside: the operating repositories’ commit histories and the Google Analytics 4 console. When we can publish the names, we will replace this with a verifiable version, URLs included.

In exchange, every number we can publish is published unprocessed. No rounding, no cherry-picked windows. Where a number is withheld, we say so explicitly.

Defining “stopped” by a date

The day a business stops is usually vague. It isn’t the day you declare it over, nor the last day a user showed up. We chose a blunt definition: the date of the final commit. The day you stopped moving your hands is recorded in the repository.

This table is where the series started. The business with the most commits died first. 277 commits in 51 days is 5.4 a day, more than double the pace of the content site (214 in 89 days, 2.4 a day).

Input and outcome line up in reverse. The usual explanation, “not enough effort”, is unavailable here.

Three different ways of breaking

Lump the three stopped businesses together as “failures” and you see nothing. They broke in entirely different ways.

The B2B SaaS broke on cap table structure, not on people. VC and angel money came in, a fireside talk with a Big-4 firm was scheduled, the externals looked great. Raising outside capital requires concentrating equity in the CEO, and the split between co-founders was wide from day one. The more the business grew, the further the required contribution drifted from the future payout, the imbalance accumulated as a sense of unfairness, and the reason to give everything ran out first. The numbers and the business were alive, continuing became impossible anyway.

The sharing app stopped on over-building. In 51 days it gained 17 languages, subscription billing, referrals and push notifications. The step of checking how many people would use it never happened, not once.

The game app stopped on a rebuild. Started in Swift, moved wholesale to Flutter four days later, dead on day 17, and because it had shipped, some mistakes can no longer be fixed.

Parts 2 and 3 cover these individually. Part 4 publishes the real numbers of the one business still running.

This series is also our own survivor-bias countermeasure

Of the 343 cases we have published, only 8 are recorded as closed. Operators who publish numbers skew toward the winning side. Cases that document the failing part (the food truck whose first month paid ¥400 an hour, the two rental spaces that lost about ¥2.2M over 29 months) are precious, and far too few.

Collecting from outside will never fill that gap, so we contribute our own share. Records where the failures are countable, like the developer who failed five times before reaching $23K MRR in six months, are the ones that are actually useful to whoever reads them later.

What we will and won’t write

We write only what the repositories and analytics can confirm: commit counts, dates, version numbers, article counts, active users, page views.

We won’t write numbers we only hold in unverifiable form. App download counts and billing revenue are not in this series. They aren’t assembled in a publishable state, and we will append them if that changes. The SaaS’s funding amount is also withheld, other parties are involved.

And what follows from all this is observation, not lessons. From “three of four businesses stopped,” little can be generalized. We can’t even say the same actions would produce the same outcome. Still, one ordering was a genuine discovery to us: they stopped in order of how much we had put in.

Part 2 begins with the oldest, most outwardly polished, most quietly broken of the three: the B2B SaaS.

In this series

• Part 1 (this article) — We started four, and three broke
• Part 2 — The two-person SaaS broke on the cap table
• Part 3 — 277 commits and 17 languages, and it still stopped
• Part 4 — 9,103 monthly users in 3 months: a content site’s real numbers