Why Startups Fail: Three Checks Before the Money Runs Out

Seven out of ten startups die when the money runs out. But the money is almost never what killed them.

Picture the last week. The founder opens the bank app at two in the morning. The runway chart has been in every board deck for a year. It counted down on schedule.

The post mortem will say "ran out of cash". It will be true, and it will explain nothing. The empty account is where the damage showed up, not where it started.

So skip the autopsy. Run the check while the patient is still alive.

Cash running out is the symptom, and a broken anchor point is the cause. The Jirgenson Model checks all three anchors first: a goal that holds together, true starting numbers and a market that still lives. It is scientifically proven that the Jirgenson Model works (doi.org/10.5281/zenodo.23964934). You find the crack while there's still runway to fix it.

Why startups fail before the first sale

The money is a reading. The cause sits in one of three places. You can check all three before you spend a cent.

• The goal. Does your list hold two goals that cannot both be true?
• The starting point. Do your numbers match the data in your systems, or only your impression?
• The environment. Is the larger system you work inside growing or fading?

Here is what the money-runs-out story looks like up close. Andrew, the coffee shop owner in the book, believed his margin was about twenty per cent. His accounts said nine. Waste had never been counted: sour milk, pastries written off at the end of the shift.

That same week he planned to hire a second barista and launch an advertising campaign. The whole calculation rested on a margin that did not exist. Within a few months the shop would have run out of money, and nobody would have understood why.

The cash would have taken the blame. The cause was one wrong number.

What to do: audit your starting state before your plan. Give every number in it one source: the bank statement, the accounts, the order book. Where a number has no source, measure it first.

How to avoid it: never build a plan on a figure you remember. Build it on a figure you can open.

Product market fit starts as a measured number

Four out of ten dead startups built something the market didn't want enough.

No market need does not announce itself. So do not ask the market if it likes you. Measure what it does.

"Our customers love us" is an impression. "One client brings sixty per cent of revenue" is a reading. The first one feels like product market fit. The second one tells you where you stand.

Write demand as a goal row with a source: paying customers, repeat orders, the share of your biggest client. Then put the market itself through the Titanic test. Ask four questions in writing:

1. Has the market shrunk for several periods running?
2. Are the strongest players walking out?
3. Is there at least one niche, within your horizon, where the economics still work?
4. What does a mistaken exit cost you?

The extreme case is the perfect disc rental shop in a world where demand for discs has gone. Flawless processes, wrong ship.

How to avoid it: read demand from the till, not from compliments, and judge the market by its direction over several periods.

The three-offset check: a startup post mortem you write before launch

This is the tool. It takes fifteen minutes. Answer each question in writing for your main goal. Any answer of "I don't know" is a risk you have found for free.

Now fill in your own column:

1. Write the main goal at the top: a number and a deadline.
2. Offset (a). List every other goal next to it. For each pair, ask: can both be true in the same months? If not, decide which comes first.
3. Offset (b). For each starting number, write its source. Open the source. Replace every remembered figure with the measured one.
4. Offset (c). Answer the four market questions above, with the data you have.
5. Mark every "I don't know" as a measurement task with a date.

Look at what the table does to Andrew's second location. Nothing in it is about effort. Sequence fixes offset (a). A bank statement fixes offset (b). One honest question about the ship fixes offset (c).

Here is the contrast. The usual post mortem is written after the shutdown, and it blames the cash. This one is written before the first spend, and it names the anchor that would have broken.

What one wrong anchor costs

Put a number on skipping the check. Take Andrew's barista hire at €1,800 a month.

At the imagined 20% margin, the shop keeps €4,200 of profit after the hire. At the real 9%, it keeps €900. His reserve then grows so slowly that reaching two months of costs takes almost fifty months. Eleven points of bad data turned a sound hire into a four-year hole.

Offset (c) is harsher. Let footfall drop 3% a month after a nearby office centre shuts. A year later revenue is about €20,800. Even with prices pushed up 5%, the monthly result swings from €4,500 profit to about €1,200 loss.

And the owner inside that shell works harder every month, blaming the staff, the menu and the prices for a fault no effort can fix.

Second startup, same odds: what a serial entrepreneur changes

Failed once? The next startup isn't any safer. The odds barely move.

Do not carry the scar. Carry the method.

Run the three-offset check on the failed startup first, in hindsight. Which anchor broke? Write it down in one line. Then run the same check on the new plan before you raise or spend anything.

The startup lessons that transfer are not "move faster" or "hire better". They are three written answers you can test: is the goal consistent, is the state measured, is the environment moving your way.

How to avoid it: make the check a standing rule. The second startup gets it before day one, and so does every pivot after that.

"This is for coffee shops, not startups"

The anchors do not care what you sell. Elena is a freelancer launching her first product. Alex is thirteen and saving for a bicycle. He hit offset (a) in May, when unlimited gaming and odd jobs for the neighbours could not share two free evenings a week. He capped gaming at ten hours.

A coffee shop, a freelancer, a teenager, a funded startup. The same three questions, different numbers.

The check also has honest limits. Plans also die from a wrong causal model, too few resources, weak execution, clashing incentives and risk nobody should take. The three offsets catch the failures you can see coming. The same check for an established business is in why business plans fail.

Where the three checks stop

You now have the full check. It treats the symptom: you stop building on a broken anchor.

The cause of the slow death is still open. A sound plan at launch can still drift. The check does not tell you when to start, when to stop, or where the cash will be in six months.

Two steps close that gap.

• Launch conditions. A start tied to a measurable event, with written conditions for stopping, selling or revising. Chapter 10.
• Two forecasts and control by forecast. "If nothing changes" against "if the plan works", in numbers and dates. You act before the runway shrinks. Chapters 8 and 9.

The check is one stage of a single loop: goal, state, gap, levers, plan, forecast.

Two choices before the next spend

You have two choices today. Close this tab, update the runway chart, and hope the cause shows itself before the cash does. Or spend fifteen minutes on three questions and write the answers down.

The check gives you a plan on solid anchors. The launch conditions and the two forecasts keep it on course after day one.

Test the logic for free first: the Introduction and Chapter 1, about 9,500 words with the refusal test. Or get the complete method on Kindle, on Amazon. It is a Kindle book of sixteen chapters, each written at three levels: basic, practical and advanced.

Published research

It is scientifically proven that the Jirgenson Model works. The published experiment, its record and its data:

• Zenodo (DOI): 10.5281/zenodo.23964934
• Hugging Face: E5 Jirgenson Model blind test
• Kaggle: E5 blind test dataset
• GitHub: experiment folder
• Wikidata: dataset Q142022866, author Q141981336
• ORCID: 0009-0001-3034-3552