Startup Failure Rate: Steer by Forecast, Not by the Calendar

Three out of four venture-backed startups never give investors their money back.

Picture the monthly investor update. Twelve slides. Revenue against plan, burn against plan, a chart that bends upward from next quarter. Every number on it describes last month.

You spend two days polishing the deck, and it still cannot tell you where the cash will be in July.

That is the gap to close. Not a lack of effort. A way of steering that only ever looks in the mirror.

A deck that only reads last month can't steer next quarter. The Jirgenson Model adds the missing view: real numbers, a rolling forecast and levers you pull before the runway ends. It is scientifically proven that the Jirgenson Model works (doi.org/10.5281/zenodo.23964934). You stop reporting the past and start steering the next move.

Venture capital returns and the goal you defend last

That first number belongs to investors. Startup statistics and startup survival rate tables describe the crowd. Your job is narrower and harder: know what you defend to the last.

Write a goal table before the term sheet, not after. No more than nine rows, each with a number and a deadline. Then run the refusal test: if things got bad, which goal goes first? Which is the last one left?

Do the same, quietly, for the people across the table. Compare what they say with what they measure and move. Whatever someone defends to the last is their goal number one, announced or not.

How to avoid it: agree the order of goals while everyone is calm. In a bad quarter, that order decides. Without it, whoever shouts loudest does.

Seed stage startups and the unicorn in the plan

Two out of three seed-funded startups end up dead or barely alive. One in a hundred becomes a unicorn.

Do not build the plan on the unicorn startup, the one in a hundred. Build two forecasts instead: "if nothing changes" and "if the plan works", each with figures and dates.

Then write each one as a range with a condition. A single number is almost always wrong. "The reserve reaches two months between month 10 and month 12, provided the margin holds above 14%." That is a forecast you can test. "We hit the target by Q3" is a hope.

The lower the odds that the result you want comes on its own, the stronger the control you need. So compare scenarios: the chance of reaching the goal under each action, its cost and its risk.

How to avoid it: let the unicorn stay a scenario, never the base case.

Startup failure rate vs your steering: plan, deviation, forecast

This is the tool. Every startup steers in one of three ways, and the way you choose sets how fast you can respond to the unexpected.

Picture three drivers on one road. The first has locked the accelerator. The second brakes when the brake lights ahead come on. The third watches three cars ahead and eases off before anyone brakes.

Now the same three on Andrew's coffee shop counter, with a hundred croissants.

Waste fell, and so did shortages. Andrew then used the same habit on his cash reserve, milk orders and shift schedule.

Now run four rules on two weeks of croissant sales. Here is the total miss over the second week:

The fastest rule did worst. It chased noise. The rule that knew the pattern won.

Now build yours. It takes fifteen minutes.

1. Name your scheme. For cash, which do you use today: plan, deviation or forecast? Write it down honestly.
2. Find the pattern. Record your key number by week, by month, by season. Andrew needed a few weeks of sales by day and weather.
3. Draw a rolling forecast. Cash, enquiries and workload for the next three months. Update it every week or every month.
4. Pick a leading indicator. Elena, the freelancer in the book, checks it every Monday. 280 visitors at a 1.5% conversion rate is 4.2 sales a week, about 17 in four weeks. Her goal is 21. She sees the gap of four before the month ends, so she can rewrite the page headline now.
5. Respond in the right order. First build the pattern into the forecast. Then decide which deviation deserves a response. Only then speed up.

Here is the contrast. The usual way waits for the month-end report and learns about the shortfall when it is too late to fix. The forecast way sees the same shortfall in week one.

Startup shutdown numbers and the current you sail in

Startup shutdowns have climbed every year since 2022. This year is on track to be the worst yet.

Do not wait for that wave to show up in your own accounts. Respond to the measured cause, not to the damage.

A yacht in a crosswind that pushes it ten degrees east can hold due north and correct all the way. Or it can measure the drift and set 350 degrees from the start. The heater that switches on when the outside temperature drops works the same way. It does not wait for the room to cool.

So list the outside factors you can measure early: the funding climate in your sector, your customers' budgets, your suppliers' prices. Put each one into the forecast as an assumption you check.

How to avoid it: read the climate as an input to the forecast, never as a surprise in the report.

Startup runway: from the last round to the last month

From the last funding round to total shutdown? Less than two years.

Start the runway calculation the day the money lands, not the day it runs low. Two numbers, not one: how many months of costs you hold, and when the money runs out at today's pace. The full method is in months of runway.

Then draw a red line and say it out loud early. Andrew drew his at one month of reserve. That is how long he needs to ask for a deferral, find money and decide. Under a month, none of the three fits.

Name the risk responsibly, and back it with an action the same day. "We will go bust in three months", said to the team with no plan, makes people leave and the reserve fall faster. The same forecast with an action attached moves the levers sooner.

How to avoid it: forecast the month you cross the red line, and act before it.

What steering by the calendar costs

Put a price on six more months of looking backwards. At a 9% margin, Andrew's reserve reaches two months of costs at month 16. With his plan working, at month 10. If nothing changes, the reserve stands at about 0.8 months by the end of month 3: the red zone.

And the croissants keep the score in plain view. Thirty in the bin every Tuesday. Thirty guests sent away every Saturday.

A system that does not read reality cannot learn from it. The owner keeps calling it bad luck and looking for someone to blame.

"A startup is too uncertain to forecast"

That is the point of forecasting it. A forecast is not a prophecy. It is a hypothesis you test against the facts every month, and the gap teaches you faster than any report.

The scheme has honest limits. It needs a model and discipline. With a poor model it can do worse than simple feedback, so you keep the feedback too. Forecasts err, and the schemes work best combined.

Where the forecast stops

You now have the tool, and it is complete. It treats the symptom: you stop steering by last month.

The cause of most late decisions is still in place. The forecast shows a miss coming. It does not set the line where a miss turns from yellow to red, or tell you when a growth push may start.

Two steps close that gap.

• Zones and the red protocol. The exact lines between green, yellow and red, and the order of action once cash goes red. Chapter 4.
• Steady mode and manoeuvre. When to change the plan and when to change the goal, and the launch conditions for a growth push. Chapter 10.

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

Two choices for the next board meeting

You have two choices this month. Polish the deck again, report last month, and hope the next quarter bends the way the chart says. Or draw a three-month rolling forecast tonight and steer by it on Monday.

The forecast lets you see the miss coming. The zones and the manoeuvre rules tell you what to do when it arrives.

Test the logic for free: read the Introduction and Chapter 1, refusal test included. Or set up the whole loop with the full control system, on Amazon. The system comes as a Kindle book, and each chapter ends with a task you can try the same week.

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