How to find and evaluate startup ideas

"What you want is to fall in love with a problem."

— Jared Friedman

Friedman cofounded Scribd, the digital library, as its CTO; Y Combinator funded it in 2006, and in 2015 he joined YC as a partner. Today he is a Group Partner, and in this Startup School talk he summarizes what YC has learned about ideas: how those of its most valuable companies were born, the mistakes that repeat in applications and the lessons from pivots. No one knows for sure which idea will work, and execution matters more, but some ideas have a higher probability of success than others. The link starts at minute 19:42, where the recipes for generating ideas begin; the first part, with the common mistakes and the ten questions, is also worth it.

Friedman starts with the most common mistakes:

• A solution in search of a problem. Fixating on a technology, today artificial intelligence, and then looking for someone to apply it to produces problems nobody cares about. The remedy is to start from a concrete problem.
• The tar pit. There are ideas that seem obvious and attractive, like an app for making plans with friends, but hide structural barriers that have stopped everyone who tried. Before jumping in, find out who tried it and understand why they didn't make it.
• Sticking with the first idea without evaluating it, or the opposite, waiting for the perfect idea, which doesn't exist.

The ten questions for evaluating an idea

1. Are you the right team? Founder-market fit is usually the most important criterion (PlanGrid: a founder with construction experience and a programmer cofounder).
2. How big is the market? Big today, or small today but growing fast (Coinbase started when the Bitcoin market was tiny).
3. How acute is the problem? If the user has no good alternative, it's acute (Brex: startups couldn't get corporate cards).
4. Do you have competition? Competition is almost always a good sign: it confirms there is a market.
5. Do you want it yourself? Do you or the people you know need it? If not, you're guessing.
6. Why now? What changed in technology, regulation or behavior that makes it possible or necessary (Checkr and the demand from the gig economy).
7. Is there a reference model? A company that already works in another country or another market (Rappi taking DoorDash as its reference).
8. Is it a long-term interest? Could you work on this for many years? Passion often comes later, when the business starts working.
9. Is it scalable? Software scales; services hidden inside the product don't.
10. Is it a good idea space? If your specific idea fails, are there good pivots nearby? (Fivetran found its product after two pivots within data analytics.)

Friedman adds three signals that look bad and are good. Being hard to get started scares others away (Stripe, because nobody wanted to negotiate with banks). The space being boring means less competition and a higher chance of getting it right (Gusto, in payroll). And the existence of competitors proves the problem is real, and a better solution can win (Dropbox, with its automatic file sync).

The seven recipes for generating ideas

1. Start with what your team knows how to do. What is your team unusually good at? (Rezi: its founders looked in real estate and finance, where they had experience.)
2. Start from a problem you lived. What do you see broken that others don't? (Vetcove: the founders saw how a veterinarian in their family bought supplies.)
3. Think about what you wish existed. It's the recipe with the highest tar-pit risk; first check why it doesn't exist (DoorDash was born from Stanford students who wanted food delivered).
4. Look for recent changes in regulation, technology or behavior (Gather Town pivoted when the pandemic changed how people met).
5. Look for variants of successful companies, a proven model taken to another region or vertical (Nuvocargo, "the Flexport of Latin America").
6. Talk to people about their problems. Pick a space, go out and interview users and map their pains (AtoB: its founders went to truck stops, talked to drivers and found the fuel-card problem).
7. Look for big, broken industries. And as an extra recipe, find a cofounder who already has an idea.

The fact that dismantles the brainstorming myth is that at least 70% of YC's hundred most valuable companies found their idea organically, not by sitting down to think up startup ideas. The way to prepare is to become an expert in a valuable field, work at a startup to see broken processes or build things for fun.

Friedman's close is direct: you can analyze an idea forever and still be wrong. If you have doubts, launch as soon as possible and find out. Even if your idea is good, it will change along the way.

How to apply it? Run your idea, or that of a team you're evaluating, through the ten questions on one sheet, one line per answer. If team, market or acuteness of the problem come out weak, stop. And before falling in love with the solution, find out who tried it before and why they didn't make it.

As a founder it will serve you as a filter before committing years to an idea. It will also help you evaluate the teams and projects you're asked to invest in or collaborate with.