How do you pick a startup idea worth working on?

The best startup ideas come from noticing a real problem you personally experience, not from brainstorming what might be a good business. Your idea needs to pass three tests: the problem genuinely exists, you can build the solution, and almost no one else has recognized the opportunity yet. Get all three right and you have something worth betting years of your life on.

Start with problems you live inside, not ideas you invent

The single biggest mistake early founders make is constructing an imaginary user and then building for that fiction. Paul Graham's point about idea generation is that the most durable startups — Microsoft, Apple, Google, Facebook — began with founders solving problems they themselves faced. This matters because when you are the user, you cannot fake whether the problem is real. You already know how painful it is, what the current workarounds cost you, and what a genuinely better solution would feel like.

The practical implication is that you should audit your own frustrations before you audit market reports. Spend a week writing down every time you think 'why does this still work this way?' across your professional and personal life. The problems that make you slightly angry — not just mildly inconvenienced — are the ones worth investigating. Mild annoyances produce weak startups. Deep, recurring pain produces companies.

The trap to avoid here is building something that fits a story you want to tell rather than a problem the market confirms. If you cannot find at least a handful of other people who have the same problem and are actively, awkwardly working around it right now, that is a signal the problem is not as real as it feels from the inside. Talk to ten people before you write a line of code.

Look for ideas that look wrong to most people but aren't

Paul Graham's framework for evaluating ideas as an investor surfaces a counterintuitive filter: the very best startup ideas often look like bad ideas at first glance. This happens because many genuinely good opportunities only become viable after some shift in the world — a new technology, a regulatory change, a behavioral norm that just crossed a tipping point — and most people are still reasoning from the old map. When experts dismiss your idea, it is worth asking whether they are experts on an earlier version of the world rather than the current one.

This does not mean contrarianism is a strategy. Most ideas that look bad are simply bad. The distinction is whether you can articulate specifically why the idea looked unworkable before and what has changed. 'Strangers will rent rooms in their homes to other strangers' sounded wrong until smartphones made trust and coordination cheap enough to make it work. If you cannot name the change that switched your idea from bad to good, you may just have a bad idea with motivated reasoning attached to it.

A useful exercise: write down your idea and the three strongest objections to it. Then write down the specific reason each objection no longer applies. If you cannot do that cleanly, the idea needs more pressure-testing before you commit.

Check whether you can get close to early users

One of the most underrated filters for idea quality is access. Paul Graham raises this directly: if you are building in a domain where you have no connections and cannot easily observe a small set of real users, you are starting in a significant hole. This is not just a go-to-market problem — it is an idea-quality problem. Your ability to learn fast enough to survive depends almost entirely on tight feedback loops with actual users in the early months.

Ask yourself: do I know five people who have this problem right now and would let me watch them struggle with it? If the answer is no, your path to validating and iterating the idea is going to be slow and expensive. This does not mean you cannot build for a domain where you lack connections, but it does mean you should factor that access deficit into your conviction level before you start.

The advantage small startups have over large companies is the ability to care disproportionately about individual users. Garry Tan has noted that founders sometimes undermine this advantage by performing bigness — imitating the indifference of established companies because it feels more professional. In reality, the founders who win early are the ones who do things that do not scale: personally onboarding every user, reading every piece of feedback, treating early customers like the critical signal sources they are. Your idea's quality will be revealed through that process, and your willingness to do it determines whether you get the data.

Apply the unfair advantage test before you commit

Once an idea passes the reality check — real problem, real users, real access — apply one final filter: why are you the right person to work on this? The answer does not have to be glamorous. It might be domain expertise, a specific technical skill, a professional network, or simply that you have lived the problem longer and more intensely than anyone else. But some answer has to exist.

This matters because good startup ideas attract competition once they are validated. If your only edge is that you saw the idea first, you are in a race where other teams with more resources can catch up. If your edge is that you spent eight years in the industry and know exactly how decisions are made at the buyer level, that is much harder to replicate in twelve months.

The unfair advantage test also helps you rule out ideas that are genuinely interesting but not right for you. An idea can be a real problem, look contrarian-but-correct, and have accessible early users — and still be the wrong idea for you specifically because someone else has structural advantages you cannot match. Intellectual honesty about this early saves years. Pick the intersection of 'real problem' and 'I have an edge here' and you have a foundation worth building on.

“The way to get startup ideas is not to try to think of startup ideas. It's to look for problems, preferably problems you have yourself.”

— Paul Graham, source

The one thing to do

Before writing a line of code, find ten real people who have the problem today and watch how they currently suffer through it — if you can't find them, find a different idea.

Frequently asked questions

What if I don't have any personal problems that seem like startup ideas?

Change your environment before you change your thinking. Work inside an industry for a year, talk to professionals in a domain with known inefficiencies, or build something small in an adjacent space and notice where the friction is. Genuine problems surface through proximity, not brainstorming sessions.

How do I know if my idea is too niche to build a real business on?

Niche is fine at the start — almost every large company began by serving a very small, specific group of users intensely. The question is whether your niche is a foothold into a larger market or a ceiling. Map out who else has the problem beyond your initial users and whether solving it well for the few naturally expands to the many.

Should I avoid ideas that already have competitors?

Competition is usually a sign the problem is real, not a reason to walk away. The more important question is whether existing solutions are genuinely bad and why — and whether your angle attacks the problem in a way competitors are structurally unable to copy.

How long should I spend evaluating an idea before starting to build?

Long enough to talk to ten or more potential users in real conversations, short enough that you haven't convinced yourself the idea is good through repetition rather than evidence. For most ideas, two to four weeks of focused customer discovery is enough to decide whether to build a prototype or move on.

Sources

More playbook answers · Growth Prophet home