This is a translation of Paul Graham's essay "How to Get Startup Ideas," posted to his blog in November 2012.
Paul Graham is the founder of Y Combinator, Silicon Valley's premier startup accelerator. YC graduates include Dropbox, Airbnb, Heroku, Cruise, and many other notable tech companies.
The combined market cap of YC graduates is estimated at over $100B.
This essay is fundamentally about how to recognize the right pitch to swing at — essential reading for any startup pursuing rapid growth. It's extremely important and deserves wide readership. It's long, roughly 20,000 characters in translation, and my translation is rough — but I hope many people in Japan's growing startup world will read it. I couldn't find a Japanese translation anywhere else, so I made my own. I've added my own commentary at the end.
How to Get Startup Ideas
The way to get startup ideas is not to try to think of startup ideas. It's to look for problems — preferably problems you yourself have.
The best startup ideas have three things in common. This applies to Microsoft, Apple, Yahoo, Google, and Facebook:
- Something the founder themselves wanted
- Something they could build themselves
- Something few others recognized as worth doing
Problems
Why is it so important to work on your own problems? Especially critical is confirming that the problem actually exists. If the problem is real, the case for tackling it is obvious. The most serious mistake startups make is trying to solve problems that don't exist anywhere.
I made that mistake once. In 1995 I started a company to put art galleries online. But galleries didn't want to be online. The idea didn't fit the art business. Why did I spend six months on such a foolish idea? Because I wasn't paying attention to users. I was developing and working from a model that reality didn't actually require. I didn't notice that my model was wrong. I was possessed by the world model and spent too much time writing software, blindly believing "they must be waiting for this!"
Why do so many founders build things nobody wants? Because they are trying to think of startup ideas. This is dangerous in two ways:
- This approach rarely produces good ideas, and
- It produces bad ideas that sound plausible enough to be mistakenly acted upon.
At YC we call these "made up" or "sitcom" ideas. Imagine a character in a TV show starts a startup — the writers have to invent something. But inventing good startup ideas is hard. They'll come up with something that seems okay, but unless they're remarkably lucky, it will turn out to be terrible.
For example, consider a social network for pet owners. It doesn't look wrong at first glance. Millions of people actually own pets. Many spend a lot of money on them. Surely many of those people want a site to talk to other pet owners. Maybe not all, but 2-3% of ordinary visitors could be millions of users. You could offer premium features to target them and monetize.
The danger of this kind of idea is that when you try to make it real with friends, they won't say "I would absolutely never use that." They'll say "yeah, I'd probably use it." When the service launches, many people will think it sounds plausible. Even if they don't want it themselves, they can imagine someone else wanting it. You add up the market reactions — and wind up with zero users.
Digging a well
A startup at launch needs to be something at least a few people actually want. Usually that initial user count is small — for a simple reason: if something were urgently needed by many people and buildable with the effort a startup can typically muster for version 1, it would almost certainly already exist. So you have to compromise in some dimension. A startup can build something many people want slightly, or something a few people want intensely. In this case, choose the latter. Not every idea of the latter type is a good startup idea, but almost every successful startup idea is the latter type.
Imagine a graph with x-axis as number of users who want the product, y-axis as the intensity of desire. Invert the y-axis scale and you can picture a company as a hole. Google is a giant chasm — hundreds of millions of people use it. A startup just getting started can't expect to excavate that many users immediately. So there are two choices about the shape of the hole you start with: wide and shallow, or narrow and deep like a well.
The most important graph I draw for startup founders.
"Made up" ideas are usually the former — many people have some interest in a social network for pet owners.
Almost all great startup ideas are the latter. When Microsoft built Altair Basic, it was one well. Only a few thousand people owned an Altair, but without the software they'd have had to program in machine language. Thirty years later, Facebook had the same shape — their first site was limited to Harvard students, just a few thousand users. But those thousands wanted Facebook intensely.
When you have a startup idea, ask:
- "Who wants this right now?"
- "Would they use even a ridiculously crude version, built by two people nobody's heard of?"
If you can't answer, the idea is probably bad.
The narrowness of the well itself doesn't matter. What matters is depth. Narrowness is a side effect of optimizing for depth (and speed). You'll almost always get narrowness in the bargain. In fact, the correlation between depth and narrowness is so strong that when you can see an idea strongly appealing to a specific group or type of user, that's a good sign.
But well-shaped demand is necessary but not sufficient for a good startup. If Zuckerberg had built something appealing only to Harvard students, it wouldn't have been a good startup idea. Facebook started in a small market — but one it could quickly escape. That's why Facebook was a good idea: universities are similar everywhere, so if you could build a Facebook for Harvard, you could spread it to all universities rapidly. Once one student joined, they'd invite everyone else.
Microsoft was similar: Altair Basic → Basic for other machines → other languages → operating systems → applications → IPO.
The self
How do you know if an idea has a path forward? How do you tell whether an idea will become part of a massive company versus a niche product? Usually you can't.
Airbnb's founders didn't recognize how large a market they were attacking at first. They had a far narrower initial conception: helping hosts rent floor space during conventions. They didn't realize the idea would scale — it expanded gradually, forcibly. What they knew at first was only that they were onto something. That's probably the same as what Bill Gates or Mark Zuckerberg knew.
If there's a niche path in an idea, it's visible from the start — though I sometimes see paths that aren't immediately obvious. That's one of YC's specialties. But no matter how experienced, there's a limit to the accuracy rate. The most important thing to understand about finding a path out of the initial idea is the meta-fact: it's hard.
If you can't predict whether an idea has a path, how do you choose? The truth is frustrating but interesting: if you're the right person, you'll have the right intuition. But when you feel something has value at the frontier of a rapidly changing field, that intuition is more likely to be correct. As Robert Pirsig wrote in Zen and the Art of Motorcycle Maintenance:
Do you want to know how to paint a perfect picture? It's simple. First make yourself perfect, then just paint naturally.
I've been thinking about that passage since I read it in high school. I'm not sure how useful his advice is for painting, but it applies perfectly here. Empirically, the way to have good startup ideas is to become the kind of person who has good startup ideas.
Being at the frontier doesn't mean you have to be pushing it forward — you can be at the frontier as a user. Zuckerberg saw Facebook as a good idea not so much because he was a programmer but because he used computers so intensely. In 2004, if you'd asked an average 40-year-old whether to publish personal information online, they'd have been shocked. But Mark was already living online. It was natural to him.
Paul Buchheit says people at the frontier of rapidly changing fields "live in the future." Combining that with Pirsig's line:
Live in the future, and build what's missing. — Paul Buchheit
That describes how most great startups began. Apple, Yahoo, Google, Facebook — none were initially thought of as companies. They grew from founders' personal projects after discovering a gap in the world.
When you look back at how successful founders came to their ideas, they received external stimulation into a prepared mind. Bill Gates and Paul Allen heard about the Altair and said "we could write a Basic interpreter for that." Drew Houston forgot his USB drive and thought "I need my files online." Many people heard about the Altair. Many USB sticks got forgotten. The reason the stimulation led those particular founders to start companies was that their experience had prepared them to notice the opportunity.
The verb I want to use about startup ideas is "notice," not "think up." At YC we call ideas that grow organically from founders' own experience "organic" ideas. Almost all successful startups begin that way.
This may not be what you wanted to hear. You may have been hoping for a recipe. Instead, I'm saying the key is a properly prepared mind. Unfortunately, that's the truth and the only recipe. At worst, it won't take a weekend — it could take a year.
If you're not on the frontier of a rapidly changing field, you should probably master something. For example, any reasonable, smart person could likely master programming in a year. Since successful startups consume at least 3–5 years of your life, a year of preparation is a perfectly rational investment — especially if you're looking for a co-founder.
You don't have to learn programming specifically to stay at the frontier of fast-changing domains. Other domains change quickly too. But learning to hack is valuable enough to be worth it regardless. As Marc Andreessen said, software is "eating the world" — and that trend will likely continue for decades.
Knowing how to hack means you can implement ideas when you have them. It's not absolutely necessary (Jeff Bezos can't program), but it's an advantage. When you think "this seems like a great idea," instead of just thinking it, you can also think "let's build a prototype tonight" — and do it.
Noticing
If you can see some aspects of the future, the way to notice startup ideas is to look for things that are missing. If you're at the frontier of a fast-changing field, things that are clearly missing will be obvious. What's non-obvious is that they are startup ideas. Just apply the filter "what's missing" — turn all other filters off. Especially "could this become a big company?" You'll have time for that test. But using that filter first will cause you to reject many good ideas while gravitating toward bad ones.
Invisible things take time to find. You have to almost trick yourself into looking past the surrounding ideas. But you know the idea is there. When the right problem is found, it will seem obvious in retrospect. What you have to do is turn off the filters that block idea exploration. The strongest filter is taking the current state of the world for granted. Almost all of us do that even in our most open-minded moments — if we stopped to question everything, we couldn't get from the bedroom to the front door.
But if you're hunting for startup ideas, you need to sacrifice some efficiency in accepting the status quo and question things:
- Why does my inbox overflow?
- Is it because I receive too many emails, or because it's hard to get them out?
- Why do I receive so many?
- What problems are people trying to solve by sending me email?
- Is there no better way to solve them?
Be especially alert to things that put you to sleep. The benefit of taking the status quo for granted isn't just efficiency — it also makes life more bearable. Even if we knew everything we'd gain in the next 50 years and somehow still lacked it, present life would feel quite constrained compared to someone returning from 50 years in the future. When something bothers you, it may actually be because you're already living in the future.
When you find the right problem, you should be able to describe it clearly, at least to yourself. When I started Viaweb, every online store was hand-built by a web designer creating individual HTML pages. As a programmer, it was obvious these sites needed to be generated by software.
Strangely, this means "coming up with startup ideas is a matter of seeing the obvious." The next sentence suggests how strange this process is: you're trying to see something obvious that you've never seen before.
What you should do is relax your mind and not attack the problem head-on — like sitting down and trying to think of ideas. The best plan might be to keep a background process running, continuously looking for things that seem missing. Work on hard problems driven by curiosity, paying attention to gaps and anomalies, with a second self watching over your shoulder.
Give yourself time. You control your mind, but you can't control the stimuli that give you ideas. If Bill Gates and Paul Allen had imposed a one-month deadline to come up with a startup idea, would they have chosen it the month before the Altair appeared? Probably they'd have worked on a less promising idea. Drew Houston worked on a less promising idea (SAT prep) before Dropbox. But Dropbox was far better — absolutely, and as a match for his skills.
A good way to trick yourself into noticing ideas is to work on projects you find cool. That naturally tends to produce things that are missing. You won't want to build things that already exist.
Just as trying to think of bad ideas produces ideas that look like they might work but actually don't, working on something that might be dismissed as a "toy" often produces ideas that turn out to be good. When something is described as a toy, it means it has everything required of an idea except "being important" — it's cool, users love it.
If you're living in the future and building cool things users love, that may turn out to be more important than outsiders would guess. Microcomputers were toys when Apple and Microsoft started working on them. BackRub (Google's original name) looked like an unimportant research project. Facebook was a tool for undergrads to flirt.
At YC, when we meet startups working on ideas we might have dismissed as toys, we get excited. It's a sign that the idea is good.
If you can take the long view (it's not always possible, but nearly always), you'll find it easier to "live in the future and build what's missing."
Live in the future and build interesting things.
School
I advise college students to "start doing rather than studying entrepreneurship." The best founders' examples make that clear. What you should spend your time on at university is driving yourself toward the future. University is the best place to do that. Sacrificing the opportunity to tackle hard problems — as someone who can have organic startup ideas — is a big waste. Especially if you're not actually going to learn in class.
Domain collision is a particularly fertile source of ideas. When you know a lot about programming and start learning another field, you'll likely encounter problems that software can solve. In fact, good problems in other domains are twice as likely to be found. (a) People in that domain are less likely to have already solved it with software. (b) You can't take the status quo for granted because you don't know it.
If you're a CS major trying to start a startup, instead of taking an entrepreneurship class, take a genetics class. Or intern at a biotech company. CS students usually do summer internships at computer hardware or software companies. But if you want to find startup ideas, interning in an unrelated field might be better.
Or skip extra classes and just focus on building things. It's not a coincidence that both Microsoft and Facebook launched in January — at Harvard, January is a reading period with no classes.
But you don't need to build something that will become a startup. That's premature optimization. Just build things. Even better, build with other students. University is a good place to pursue the future not just because of classes — you're surrounded by other people trying to do the same thing. Working on projects with them gives you not just organic ideas but organic founding teams. In my experience, that's the best combination.
Be cautious about research. PhD theses are very unlikely to produce good startup ideas. For some reason, the more a project has to count as research, the less likely it is to become a startup. The set of ideas qualifying as research is very narrow, making it unlikely such a project would solve a user's problem. When a student or professor builds something as a side project, there's pressure to solve users' problems — and the energy freed from research constraints may help too.
Competition
Good ideas will seem obvious, so when you have one, you tend to feel like you're too late. Don't give in to that feeling. Your worry about being late is actually one sign that it's a good idea. A ten-minute web search is usually enough. Even if you find other people trying to do the same thing, you're probably not too late. Startups are rarely killed by competitor startups. So unless you find a competitor with lock-in that prevents users from choosing you, don't discard the idea.
If you're unsure, ask users. The question of whether you're late is the same as asking whether someone urgently needs what you're trying to do. If competitors haven't addressed it and some users urgently need it, you have traction. The real question is whether that traction is large enough. More important questions:
- Who is doing this?
- Will the number of people doing this keep growing in the future?
For example, if you're building something that "only works on phones but only the latest models," that's probably big enough traction for starters.
Engage with competitors rather than running from them. Inexperienced founders usually give competitors far more credit than they deserve. Whether you succeed depends far more on you than your competitors. A good idea with competitors beats a bad idea with no competitors.
Don't worry about entering a "crowded market" as long as you have a hypothesis about what everyone else is missing. Actually, that's a promising starting point. Google was such an idea. You need to be able to articulate in words what dimension the incumbents are overlooking. Ideally, you can say "the incumbents didn't have the courage to follow through on their own insight" — and your plan is what they would have done if they had.
A crowded market is actually a good sign: it shows there's demand, and none of the existing solutions are good enough. Successful startups enter markets with existing competitors but either have a secret weapon to take all the users (like Google) or enter what looks small but is actually large (like Microsoft).
Filters
Two more filters to turn off when looking for startup ideas: the unsexy filter and the schlep filter.
Most programmers want to start a startup by writing great code, pushing it to a server, and having users pay lots of money — without dealing with messy problems or getting tangled up with the real world. This is reasonable — those things slow you down. But this thinking is so widespread that it's left less room for useful startup ideas. If you wander toward slightly annoying or boring ideas, you might find ideas worth starting startups on.
The schlep filter is so dangerous I've written a separate essay on what I call "schlep blindness." I used Stripe as an example of a startup that benefited from turning this filter off — and it's pretty clear. Thousands of programmers were in a position to see this idea. Thousands of programmers knew how painful it was to process payments before Stripe. But when looking for startup ideas, they unconsciously balked at having to deal with payments and didn't look directly at the idea. Dealing with payments is boring for Stripe too — but not unbearable. In fact, the fear of dealing with payments kept most people away, so Stripe could start relatively smoothly in other areas like user acquisition. They knew what users wanted, so they didn't need many user interviews.
The unsexy filter is similar but involves problems you feel contempt for rather than fear. We overcame it to start Viaweb. We were interested in our software's architecture, but not e-commerce itself. We understood the problem needed solving.
Turning off the schlep filter is more important than turning off the unsexy filter. Because schlep is more likely to be an illusion — and a worse form of self-indulgence. Starting a successful startup will be a schlep regardless. Even if the product isn't boring, you'll have to deal with investors, hiring, firing, and much more. So if you have an idea you find cool but boring, don't worry. That's how good ideas are.
The unsexy filter, though still a source of error, isn't completely useless. If you're at the frontier of a rapidly changing field, your intuition about sexy correlates somewhat with what's actually valuable — especially as you get older and more experienced. And finding a sexy idea may help you work harder.
A recipe
The best way to find startup ideas is to become someone who has them, then build what interests you. But sometimes you don't have time. Sometimes you need ideas immediately — for example, when you've already started a startup and your first idea turned out to be bad.
The rest of this essay describes tricks for getting startup ideas on demand. Empirically, the organic strategy works better — but you can still find ideas this way. You have to work harder. With the organic method, an idea forces itself on you as evidence that something is genuinely missing. But when you consciously search for startup ideas, you need to replace that natural constraint with self-discipline. More ideas become visible — but most are bad and need filtering.
One of the biggest dangers of not using the organic method is that organic ideas feel like inspiration. There are many stories of successful startups where founders had crazy ideas but just knew they were promising. If that's how you feel when you think of a startup idea, it's probably wrong.
When looking for ideas, look where you have expertise. If you're a database expert, don't make a chat app for teenagers (unless you are one). It might be a good idea, but you can't trust your judgment, so ignore it. Look for ideas involving databases, things you can evaluate the quality of. Is it hard to think of good database ideas? That's because your expertise raises your standards. Your chat app idea is mediocre, but you're a victim of the Dunning-Kruger effect.
The place to start looking for ideas is what you need. It has to be something you need.
One good method: think back on your previous jobs. Was there anything you kept thinking "why doesn't anyone build X? I'd buy it immediately." If that thought spread as gossip among your colleagues, you probably have an idea. You know the demand exists, and people don't say "it's impossible to build."
More generally, look for things about you that are unusual — ways you differ from most people. You're probably not alone. If you differ from others in a way that seems to be the direction more people will move, that's especially good.
If you've pivoted from a previous idea, that previous idea is your unusual thing. Did you discover any needs while working on it? Some famous startups began this way. Hotmail started as a way to discuss a previous startup idea without employers seeing.
A particularly rare and promising approach: being young. Some of the most valuable new ideas take root first among teenagers and young adults. Young founders are at a disadvantage in some ways, but they're the only ones who truly understand their peers. It would have been very hard for someone who wasn't a college student to start Facebook. If you're a young founder (23 or under), what do you and your friends want that current technology doesn't provide?
The next best to your own unmet needs is someone else's. Talk to everyone about the gaps you see in the world. What's missing? What do they wish they could do? What's boring or annoying? Keep conversations general. You're not looking for startup ideas — you're looking for something to spark thought. You'll notice problems they haven't recognized, because you know how to solve them.
When you find an unmet need that isn't your own, it may be fuzzy at first. The person may not know exactly what they need. In that case, I often recommend acting like a consultant — work to solve this one user's problem, and the right path will become clear. Since people's problems are similar, most of the code you write will be reusable — and the part that isn't is a small price to pay for confirming you've hit the bottom of the well.
One way to solve others' problems well is to make their problems your own. When Rajat Suri of E la Carte decided to write software for restaurants, he started working as a waiter to learn how restaurants work. That may sound extreme — but startups are extreme. We love founders who do things like that.
Actually, my strategy for people who need new ideas isn't just to turn off schlep and unsexy filters — it's to actively look for ideas that are schlep-y or unattractive. Don't start Twitter-like ideas. Those are extremely rare and you won't find them by searching. Build something unattractive that people will pay for.
(Translator's note: this is a translation of approximately the first 2/3 of the full essay; my notes and commentary follow below)
My thoughts
I translated this essay because it most clearly articulates what I believe about startup ideas — specifically, the concept of "organic ideas." The ideas I've pursued myself were all organic: problems I personally experienced, that I couldn't ignore, that I happened to be able to build something for.
The concept Paul calls "live in the future, build what's missing" resonates deeply with me. The product Tabery, which 10X launched, came from exactly this process: recognizing a real daily pain (meal planning and shopping), starting small with a Slack bot I built for my own household, and discovering that the same pain was universal.
What I take from this essay most strongly: the best ideas are already inside you. You just need to stop trying to "think of an idea" and start living so deeply in your domain that the gaps become visible.
