The short definition
A startup is a company that doesn't yet know how it will make money repeatably, and is deliberately organised to find out. That's it. The hoodies, the office, the funding round, the word "tech" — none of those are part of the definition, and every one of them can be present in a company that isn't a startup.
The useful contrast is a new bakery. A bakery opening on your street is a new business, and a hard one, but the model is known: buy flour, bake bread, sell it at a margin, repeat. Nobody has to discover whether people want bread. A startup is in the opposite position — the product, the customer, the price, and the way the two meet are all still open questions.
The two things you're searching for
Everything a startup does early on serves one of two searches. Confusing them is the most common way early founders waste months.
- A problem someone will pay to removeNot a problem people agree is annoying — a problem they already spend money, time, or workarounds on. Payment is the signal; agreement is not.
- A way to reach those people repeatablyFinding one customer proves the problem exists. Finding the tenth the same way you found the first proves you have a channel. A business needs both.
When both are found and hold up under repetition, the search is over. What's left is execution: doing the known thing more, better, and at lower cost. That transition is the moment a startup quietly stops being one.
A worked example
Why growth is evidence, not the goal
Startups are associated with fast growth, and that association gets the causality backwards. Growth isn't the objective — it's the most legible evidence that the two searches succeeded. If you've found a real problem and a repeatable way to reach the people who have it, growth is what that looks like from the outside.
Chasing the evidence directly produces the familiar failure: numbers that move because you paid for them, and stop the moment you stop paying. That's a purchased graph, not a business model.
| Signal | Bought growth | Found model |
|---|---|---|
| Where users come from | One paid channel | A channel you can explain and repeat |
| What happens when spend stops | Growth stops | Growth slows, doesn't vanish |
| Why users stay | Discount or novelty | The problem is still solved |
| What you learn | Cost per install | Which problem is worth solving |
How to use this now
- Write down the model in one sentence"We charge [who] [how much] to [remove what work]." If you can't fill in all three blanks, you're in the search phase — which is fine, but name it.
- Mark which parts are known and which are guessesMost early sentences have exactly one known part and two guesses. Knowing which is which tells you what to work on this week.
- Pick the riskiest guess and test that oneNot the easiest to test. The one that, if wrong, makes the rest irrelevant.
Common mistakes
- Using "startup" as a synonym for "new" or "small" — a new business with a known model is a business, and it's often the easier path
- Treating funding as a milestone. It's a tool for going faster at a search, and it makes a wrong answer more expensive
- Building for twelve months before showing anyone, then calling the reveal a launch
- Confusing interest with demand. "I'd use that" costs nothing to say
Key takeaways
- A startup is a company searching for a repeatable business model, not a small company
- The search has two halves: a problem people pay to remove, and a repeatable way to reach them
- Growth is evidence the search worked, not the thing to aim at directly
- When both halves are found and hold, you're running a business — and that's the goal
Try it yourself
Write your model as one sentence: "We charge ___ ___ to ___." Then underline every part you have not personally verified with a real person outside your household. That underlined text is your actual to-do list for the next two weeks.