How to Get Startup Ideas

How to Get Startup Ideas. An essay by Paul Graham, from November 2012.

Look for problems

How do you get a startup idea? Graham's answer is counterintuitive: don't try to think of one. Look for problems instead, ideally problems you have yourself.

The best ideas share three traits: the founders want it themselves, they can build it, and few others see that it is worth doing.

Microsoft, Apple, Yahoo, Google and Facebook all started that way.

Problems nobody has

The most common startup mistake is solving a problem nobody has. Graham made it himself: in 1995 he spent six months putting art galleries online. Galleries had no interest.

Why does it happen? Founders sit down to think up an idea, and get what YC calls a sitcom idea: what a TV writer would invent for a character. Plausible, but bad.

Take a social network for pet owners. Nobody says they'd never use it. They can picture other people using it. Add that up across the world and you get zero users.

Dig a well

When you launch, some users must need it urgently. That group is usually small. So choose: something many people want a little, or something a few people want badly.

Picture demand as a hole. Google is an immense crater. A new startup can't dig that much. It can go broad and shallow, or narrow and deep, like a well. Dig the well.

Microsoft's first product was Basic for the Altair: a couple thousand owners who needed it badly. Facebook began with a few thousand Harvard students who wanted it a lot.

So ask: who wants this right now, enough to use a crappy first version from two strangers? If you can't answer, the idea is probably bad.

A path out

A well isn't enough. Something only Harvard students could want is a niche product. Facebook had a fast path out: all colleges are alike, and after them, everyone.

Microsoft climbed the same way: Altair Basic first, then Basic on other computers, more languages, an operating system, applications.

But paths out are hard to see. Airbnb began as floor space for rent during conventions. The founders only knew they were onto something. That is usually all anyone knows.

Live in the future

If you can't see the path, how do you choose? Graham's honest answer: the right sort of person has the right hunches. At a field's leading edge, hunches tend to be right.

You can be at the edge as a user. Facebook seemed natural to Zuckerberg because he already lived online. Most forty-year-olds in 2004 would have been horrified.

Graham's recipe: “Live in the future, then build what's missing.” Apple, Yahoo, Google and Facebook were not even meant to be companies at first.

Ideas come from a stimulus hitting a prepared mind. Lots of people heard about the Altair. Gates and Allen thought of writing Basic for it. The verb is notice, not think up.

Becoming that person

That's not the recipe people hope for. It takes a year, not a weekend. But anyone reasonably smart can reach an edge of programming, say mobile apps, in about a year.

A startup eats three to five years of your life, so a year of preparation is a reasonable investment. Learning to program isn't required, but for now it's sufficient.

Turn off the filters

Once you're living in the future, look for what's missing. And switch off every other filter, especially could this become a big company? Ask that later.

The strongest filter is taking the world for granted. Why is your inbox overflowing? Why do you keep emails after reading them? Is an inbox even the right tool?

Pay attention to things that chafe. If you knew everything we will have in fifty years, today would feel constraining. Annoyance can mean you are living in the future.

The right problem seems obvious, at least to you. So work on hard problems out of curiosity, while a second self looks over your shoulder, noting the gaps.

Toys

Give it time. Had Gates and Allen demanded an idea within a month, they might have picked one before the Altair appeared. Houston tried an SAT prep startup before Dropbox.

A good trick: work on projects that seem cool. You'll naturally build what's missing, because copying what exists isn't fun.

Things dismissed as toys often become great startups: cool, loved by users, just not important. Microcomputers were for hobbyists. BackRub was a science project.

At YC, a toy that users love counts as positive evidence. So the recipe upgrades: get to the future, then build what seems interesting.

What to do in college

Graham's advice for students: don't study entrepreneurship. You learn that by doing it. Use college to ratchet yourself into the future.

The clash of domains is a rich source. A CS major should take genetics, or work at a biotech company, and spot problems software could solve that nobody there has solved.

Or just build things, with other students. Microsoft and Facebook both began in January, during reading period. Organic ideas plus organic teams: the best combination.

Competitors

Because good ideas seem obvious, you'll feel you're late. That feeling is a sign of a good idea. Startups are almost never killed by competitors.

Users who urgently need what no rival offers are your beachhead. What matters is who is in it: people doing today what everyone does later, like owning the newest phones.

A crowded market is good news: demand exists and nothing is good enough. Enter with a thesis about what incumbents overlook. Google did: better search meant users left faster.

Schleps and unsexy ideas

Two more filters to switch off: the schlep filter and the unsexy filter. Programmers want to write brilliant code and get paid, so the convenient ideas are picked clean.

Stripe is the striking example. Thousands of programmers knew payments were painful, but shrank from dealing with them. Stripe did not, and users were desperately waiting.

The schlep filter is mostly an illusion: any good idea has as many schleps. The unsexy filter is less useless, since what you find sexy is somewhat correlated with value.

If you need an idea now

Sometimes you need an idea now, say your first one failed. Then be disciplined: you'll see many ideas, mostly bad. Beware the feeling of inspiration; made-up ideas fake it.

Look where you have expertise. A database expert shouldn't build a chat app for teenagers. Your standards are only high where you know things.

Start with things you need. Did people at your old job wish someone would build a certain thing? That is demand. What is unusual about you? Being young counts.

Then other people's unmet needs. Ask everyone about gaps they see. Act like a consultant for one user. One founder became a waiter to write restaurant software.

More tricks

Go further: seek out unsexy ideas and schleps. Nobody finds the next Twitter by looking for it. Build something unsexy that people pay for.

Look at industries that are dying, or deserve to. What will people later say replaced journalism? Replacements come in from the side, on an axis nobody expected.

Look for small markets the big players disdain. HP turned down the Apple I partly because it used a TV as a monitor. Far too déclassé.

And look for waves: gene sequencing and 3D printing are getting cheaper at Moore's Law speed. What becomes possible in the world a few years out?

You are the wave

But recipes are plan B. Hunting for waves only imitates the organic way. At the leading edge, you are the wave.

Finding ideas is subtle, which is why most people fail. Thinking them up yields plausible bad ones. With the right background, good ideas look obvious, but only over time.

So keep time free to build things because they interest you. Get to the future early, and build whatever seems interesting there. That is the real recipe.

In short

Don't think up ideas. Notice problems, ideally your own.

Good ideas are wells: a few people who want it badly, with a path out.

Live in the future, then build what is missing, or just what interests you.

Switch off the filters: big company, schlep, unsexy, and taking the world for granted.

The original is long, and better than this. Read it: paulgraham.com/startupideas.html