Don't Start a Business

2026 started, and the questions piled up again. Should I found something. Is this the timing. Should I leave my job. The shape changes; the ask does not.

I am only nine years in. I do not have a long shelf of wisdom. What I still have is one line.

Do not. I want to talk you out of it.

Below is the evidence. I separate what I feel from what I can cite. Business judgment usually fails not because the arithmetic is wrong, but because an assumption was treated as evidence.

A business is not making things

Making is fun. Watching something run that did not exist is a pleasure I cannot deny. That is still why I stay in this work.

A business changes character not when you ship, but when you start taking money for what you shipped.

• If the service dies at 2 a.m., you wake at 2 a.m.
• If a paying user wants a refund, you handle it whether or not you are in exams or on leave
• If the product returns nonsense, you explain it even when you do not fully understand it
• Tax, contracts, hiring, and conflict show up even when revenue is small

That response work runs almost every day for years. There is a river between building as a hobby and taking other people's money. Whether you should spend your scarce commitment crossing it here is a separate question.

Look at the numbers first

Korea's business demography statistics put the five-year survival rate of firms born in 2018 at 36.4%. One-year survival is 64.4%.¹ Two out of three are gone within five years.

Failure rarely looks like a movie ending. More often you limp for two or three ambiguous years and quit quietly when you are exhausted. The time and opportunity cost do not come back. The question before you start is not "Will I succeed?" It is "Can I live with a few ambiguous years?"

Success myths are everywhere. The personal and social cost of failure stays quiet. That imbalance is itself a risk. The spotlight prefers founders who shine.

Separate assumptions from evidence

Revenue sheets are usually optimistic. Reach × conversion × price looks fine on one line. The problem is not the multiplication. It is whether each cell is an assumption or evidence.

"5% of the whole market will buy" is a common assumption. Industry conversion rates are usually measured against people who already entered as free users, not against the total market. Mix those two and the plan becomes a wish.

Real conversion is low. Freemium-to-paid often sits in the low single digits, roughly 2.6% to 5.8% by category.² Count from visit and it gets tighter: about 5 paid customers per 1,000 visitors, near 0.5%.³

Look at cost too. After infrastructure, fees, and fixed costs, the hourly wage can fall below minimum wage. Then it is not growth. It is an expensive hobby.

Thin differentiation does not last. Incumbent platforms absorbing your features (Sherlocking) is an old pattern.⁴ Prompts, UI, and thin wrappers rarely buy a lasting premium.

Domain over youth

In research on high-growth founders, the average age in the top 0.1% was about 45.0. The pattern held when restricted to high-tech, hubs, and successful exits.⁵ Age itself is less the story than prior experience in that industry, which strongly predicted success.⁵

"Start while you're young" does not sit cleanly on that data. The same idea after domain depth is a different object. Sites and pitch decks can be rebuilt later. Field vision from a particular season of your career often cannot.

If I had been smarter / and yet I would start again

If I had been a little smarter, I would not have started a company. Opportunity cost, base rates, and downside make most starts hard to justify on paper.

One honest line remains. If I could go back ten years, I would start again.

Not because the idea was brilliant. Because I had good co-founders.

For three years they took no salary. Worse, they were willing to carry hundreds of millions of won in debt, and to borrow from family and friends, if that was what it took to pay employees. They were crazy enough to leave school or set down a lawyer title. They slept for years on cardboard in the office. They could share a moldy one-room flat without flinching.

The joy of growth outweighed the opportunity cost and the pain. Most of that joy was not from shipping alone. It was from facing the same direction with those people.

Harder than a shipping container

I used to think living in a shipping container had made me tough. Building a company was harder.

Outages, payroll, contracts, conflict, ego, uncertainty. Daily response becomes the job. Making is only a slice; most of it is enduring. Without people who will endure it with you, I will not recommend starting.

Choosing co-founders is heavier than choosing an idea. Whether you can share boxes, a moldy room, and debt comes before the numbers on a pitch deck.

Questions before you start

When I review a new project, I ask roughly this:

1. How many people know this problem as well as I do? Consumer friction and a problem someone already pays to solve are different market sizes.
2. Are customers already spending money on this? Replacing existing spend is hard; creating new spend is harder.
3. What remains in the worst case? Experience and portfolio lean green; legal liability and debt lean red.
4. Can I respond almost every day for years? That is often physics and calendar, not willpower.
5. Does this compound over the next decade? One-off technique fades; domain depth compounds.

Put assumptions and evidence in separate columns. Do not mix reach, free users, and paid. Convert optimistic net profit to an hourly wage. Write down what "ambiguous" means before you begin.

So

Do not ask. Do not start.

If a smarter path exists, it is usually that one: stay longer in the domain, build skill, and look again after the romance has cleared. If you want to make things, keep them as hobby, learning, and portfolio. Cross the paid-sales and entity-and-hiring thresholds slowly.

And still: if the people who would share boxes, mold, and debt are already beside you, that decision may not be one I can talk you out of. That is the only case in which I would start again.

Talk most people out of it with numbers. If you start again because of people, hold that bar higher than the idea.

Sources

1. Statistics Korea / National Data Office, business demography (provisional), 2024 release. Five-year survival for 2018 births: 36.4%. One-year survival for 2022 births: 64.4%. ↩

2. First Page Sage, freemium-to-paid conversion across SaaS samples (2021–2025). Category spread includes edtech near ~2.6% and RegTech near ~5.8%. ↩

3. Kyle Poyar with ChartMogul and ProductLed (B2B software). Visit → free → paid orders of magnitude and early time-to-value. ↩

4. “Sherlocking”: platforms absorbing third-party features (name from Apple Sherlock / Watson, 2002). Common moat list in the foundation-model era: proprietary data, workflow integration, distribution, brand/trust, network effects. ↩

5. Azoulay, P., Jones, B. F., Kim, J. D., & Miranda, J. (2020). "Age and High-Growth Entrepreneurship." American Economic Review: Insights. Top 0.1% mean founder age ~45.0. Industry experience strongly predicts success. ↩ ↩²