Track 13 · Founding · lesson 7

Most of them fail

13 min

Six lessons about how to do this well, and now the part that has to be said before anybody acts on them.

Most attempts do not work. Not most bad attempts — most attempts, including careful ones, run by capable people who did the customer conversations and computed the runway and made no obvious mistake.

This is not a warning. It is a base rate, and the difference between the two matters: a warning tells you not to do something, while a base rate tells you what to expect if you do.

The failure of a business is usually not a verdict on the person running it.

There is a small number of ways to end and most of them are only partly under anyone's control: the market was smaller than it looked, the cash ran out before the learning finished, the timing was wrong by two years in either direction, a larger competitor arrived, a co-founder relationship broke, or a customer who was a third of the revenue left.

Competence moves the odds. It does not move them to one.

What the mechanism looks like from inside

Suppose, illustratively, that a business needs three things to hold at once: a market large enough, a route to reach it, and enough cash to survive learning both. Assume each has a fair chance of holding on its own — say two chances in three.

Multiply them and roughly three attempts in ten survive all three. Nothing in that arithmetic requires anybody to be bad at anything. It is what happens when several independent conditions all have to be true.

Real attempts have more than three conditions and they are not independent, which pushes the number around in both directions. The shape of the argument survives: the number of ways to fail is larger than the number of ways to succeed, and each attempt only gets one draw.

Predict

Given a base rate like that, what does it imply about how to run the attempt?

Do not attempt it — the odds are against youStructure the attempt so a failure is recoverable, and expect to need more than oneCommit everything, since half measures are why most attempts failPick a safer type of business where the base rate is better

Two attempts, one of which you would have heard about

Case file

Composite, not a real person

A composite founder, built from several similar operations businesses

Six years in logistics operations, savings of about 25,000, no funding and no outside profile.

Starting cash: 25,000
Monthly personal cost: 1,600
Months to profit: 14
Customers at profit: 11

1Spent the first month on conversations rather than building, and found the problem was narrower than expected
2Ran the process manually for two customers, for a fee, before writing any software
3Kept personal costs near 1,600 a month, which turned the savings into about fifteen months
4Declined a loan secured on the family home in month seven
5Took a customer pre-payment for a year of service in month nine, which funded the build

Profitable in month fourteen with eleven customers, and still operating four years later at a size that supports three people.

What luck did: A larger competitor withdrew from this segment in month eight for reasons nothing to do with them, and two of the eleven customers arrived from that withdrawal.
Who else tried this: Several people ran almost this exact playbook in adjacent segments over the same period, and most of them found a market too small to support even one person.
Why it is not a template: The whole plan depended on already knowing eleven buyers by name after six years inside the industry, which is not something a plan can supply to someone who lacks it.

Now the same shape, with the coin landing the other way.

Case file

Composite, not a real person

A second composite founder, same industry, overlapping years

Seven years in logistics operations, savings of about 30,000, no funding and a similar contact list.

Starting cash: 30,000
Monthly personal cost: 2,000
Months to close: 16
Customers at close: 4

1Spent the first month on conversations rather than building, and found a problem that eight companies confirmed
2Built a first version over four months rather than running it by hand, because the process was harder to fake
3Kept personal costs near 2,000 a month, which turned the savings into about fifteen months
4Signed two customers in month six at a price that covered under half the costs
5Spent months nine to thirteen trying to raise the price, and lost one of the two customers doing it

Closed in month sixteen with four customers and no route to covering costs. Returned to employment at a slightly higher salary than before.

What luck did: The two largest prospects were acquired in the same quarter and both purchasing decisions were frozen for a year, which removed the only realistic path to breakeven.
Who else tried this: The founder who succeeded above made most of the same calls, and the visible difference between the two outcomes was decided by events neither of them influenced.
Why it is not a template: This is not a list of mistakes to avoid, because the four months of building and the low opening price were defensible decisions that would have worked in a market that stayed still.

Put the two side by side and the honest summary is uncomfortable. Similar people, similar preparation, similar decisions, opposite outcomes, and the largest single difference was what other companies happened to do in months eight and nine.

Only one of those two founders gets asked to speak about how they did it.

What the attempt actually costs, and what it returns

The cost is not the savings. The savings are the visible part.

Forgone income. Two years at a lower income than employment would have paid, compounding against everything in the arithmetic track. This is usually the largest line and it never appears on any plan.

Time, which is not recoverable. Two years at thirty and two years at fifty are different quantities, and only one of them is followed by thirty more.

Attention. Relationships, health and everything else that runs on the same scarce resource. This one is routinely dismissed and is the most commonly regretted.

Against that, three things survive a failure, and they are real:

Skills you cannot get any other way. Selling something you made, to a stranger, for money, is a skill with almost no substitute, and most people who have done it once find employment easier afterwards rather than harder.

A network built by doing rather than by attending. The eleven companies know who you are now.

Calibration. Having been wrong at your own expense changes how you evaluate other people's plans permanently, which is worth something in every job you take afterwards.

Check

Which framing of the decision is most consistent with everything in this track?

A bet with a wide distribution, worth taking at a size where the bad case leaves you able to take another oneA test of whether you are the kind of person who can do thisA near-certain failure that only the exceptional surviveA calculation to be avoided, because the numbers cannot be known in advance

Worth remembering

Most attempts fail, and the failure is usually not a verdict on the person running it.
Several conditions all have to hold at once, so the number of ways to fail exceeds the number of ways to succeed.
Published failure rates measure different things and should be treated as rough estimates rather than facts.
The largest cost of an attempt is forgone income over several years, which never appears on the plan.
Structuring the attempt so a failure is recoverable matters more than raising the chance of success by a few points.