How to Find and Evaluate Startup Ideas That Actually Work

Most startups do not die because the founders lacked talent. They die because the founders picked the wrong idea.

This is not speculation. Y Combinator’s partners have analyzed hundreds of the most successful companies in their portfolio and found a consistent pattern: the quality of the initial idea was one of the strongest predictors of long-term success. Execution matters enormously, but certain ideas are structurally more likely to succeed than others. A mediocre team with a great idea will outperform a great team with a fundamentally flawed idea more often than most founders want to admit.

The good news is that finding a strong startup idea is not a matter of waiting for a lightning bolt of inspiration. It is a learnable skill. There are specific, repeatable patterns behind how the best companies got their start, and specific, identifiable traps that cause smart founders to waste years on ideas that were doomed from the beginning.

This article breaks down the framework: the most common mistakes founders make with ideas, 10 filters for evaluating whether an idea is worth pursuing, and 7 recipes for generating ideas that have the structural properties of winners.

The Most Common Idea Mistakes

The SITIO Trap: Solutions in Search of Problems

The single most common mistake among aspiring founders. SITIO stands for Solution In Search Of a Problem. You can spot it instantly: the founder leads with the technology, not the problem.

“We are going to use AI to…” or “We are building a blockchain platform that…” They have a hammer and they are desperately looking for a nail.

The best ideas start with a problem that is painful, urgent, and real. The technology is the means of solving it, not the reason for existing. Every time a founder leads with the solution, the odds that they have validated the problem drop dramatically. They built something technically interesting and are now trying to retroactively find people who care.

The fix is deceptively simple: talk to potential users before writing code. Ask what problems they have, not whether they like your solution. If you cannot describe the problem your startup solves without mentioning the technology, you are probably in SITIO territory.

Tarpit Ideas: The Quicksand That Swallows Founders

A tarpit idea is an idea that sounds good, that many smart people have had, that seems like it should work, but for some structural reason, nobody can make it work. Y Combinator’s Dalton Caldwell coined the analogy: like actual tar pits that look like fresh water, these ideas appear promising on the surface. Animals step in, get stuck, and their decomposition attracts more animals into the same trap.

Classic tarpit ideas include: a social network for parents, a better calendar app, a restaurant discovery app. These ideas are popular because they come from common consumer frustrations. Everyone has experienced the pain of managing a messy calendar or finding a good restaurant in a new city. The ideas feel personal and urgent.

But the reason nobody has made them work is usually structural. Network effects that are impossible to bootstrap. Incumbents with distribution advantages that no product quality can overcome. Markets where the willingness to pay is zero regardless of how good the product is. A tarpit idea looks like a solvable problem, but the problem is the business model, not the product.

The diagnostic is straightforward: search for your idea and see how many companies have tried the exact same thing and failed. If the graveyard is large, ask yourself what structural advantage you have that all previous attempts lacked. If the answer is “we will just build it better,” you are in a tarpit.

Jumping Without Evaluating

Some founders grab the first idea that excites them and start building immediately. They would not buy a car without comparing options, but they will invest years of their life in a startup without spending two weeks evaluating whether the idea has the structural properties of a winner.

Waiting for the Perfect Idea

The opposite mistake. Some founders never start because they are waiting for the fully-formed, obviously brilliant idea to arrive. In practice, ideas evolve. YouTube started as a video dating site in 2005 before pivoting to general video uploads when dating videos failed to attract users. Slack emerged from Tiny Speck, a video game company whose game Glitch failed in 2012, but whose internal communication tool became a billion-dollar product. The initial idea just needs to be good enough to start. The real idea often emerges through iteration and user feedback.

The 10-Filter Evaluation Framework

Once you have an idea, run it through these 10 questions. No idea will score perfectly on all 10, but ideas that fail on multiple filters should raise serious concerns.

1. Do You Have Founder-Market Fit?

This is the single most important filter. Are you the right team to work on this problem? Do you have deep expertise, lived experience, or a unique insight that gives you an advantage over every other team that could attempt this?

Among the most successful startups in history, founder-market fit is nearly universal. Airbnb’s founders Brian Chesky and Joe Gebbia were Rhode Island School of Design graduates who could not afford rent in San Francisco. In 2007, they rented out air mattresses during a sold-out design conference, charging guests eighty dollars per night. Stripe’s founders Patrick and John Collison had been building internet businesses since their teens and knew firsthand how painful it was to accept payments online. When they founded Stripe in 2010, integrating payments still took months of navigating bank relationships and archaic payment gateways. DoorDash’s founders, four Stanford students including Tony Xu and Stanley Tang, started in 2013 by literally delivering food themselves. They personally handled the first hundred deliveries to understand every friction point in the process.

If you are building a healthcare startup but have never worked in healthcare, never been a patient with the specific problem you are solving, and have no healthcare advisors, you have weak founder-market fit. It does not mean the idea is bad. It means you are starting with a handicap.

2. How Big Could This Market Become?

Not “is the market big today” but “could this market become big?” When Brian Armstrong founded Coinbase in June 2012 as a former Airbnb engineer, the Bitcoin trading market was tiny. But if cryptocurrency succeeded the way early advocates believed, it was obviously going to be a massive market. The founders were betting on the trajectory, not the current state.

The most dangerous version of this question is when founders convince themselves that a small, stagnant market will suddenly grow because of their product. Markets grow because of external forces, including regulatory changes, technology shifts, and demographic trends, not because one startup built a slightly better product.

3. How Acute Is the Problem?

The most common mistake is working on something that is not really a problem, or not a problem people care enough about to change their behavior. The best indicator of acuteness: what is the alternative to your solution?

Consider Brex, founded in 2017 by Henrique Dubugras and Pedro Franceschi. Before Brex, if a startup wanted a corporate credit card, traditional banks would not issue one because startups lacked established credit history. The alternative was literally nothing. That is an acute problem. If the alternative to your product is “the customer continues doing what they are already doing and is mildly inconvenienced,” your problem is not acute enough.

4. Do You Have Competition?

Most founders think competition is bad. It is usually a good sign. Competition means there is a real market with real demand. The question is not “does someone else do this?” but “is the existing solution good enough that people will not switch?”

Having competition also means you need a genuine new insight, something the incumbents do not know or cannot execute on. If your only differentiator is “we will build it better,” you are in a fight you are likely to lose.

5. Do You Want This Personally?

Are you building something you would use yourself? This creates the tightest possible feedback loop: you are your own first user. You can test the product on yourself. You notice rough edges that a non-user founder would miss. Many of the best companies were built by founders scratching their own itch. Dropbox, Slack, GitHub, and Notion all started this way.

6. Did This Only Recently Become Possible or Necessary?

Is there something new that makes your idea possible now: new technology, new regulation, a platform shift, a behavioral change? When the iPhone launched, a wave of mobile-first startups became possible. When COVID hit, remote work infrastructure became urgent overnight. When large language models became commercially available, AI application companies exploded. Y Combinator’s Spring 2025 batch dedicated over fifty percent of its slots to agentic AI startups, reflecting this exact dynamic.

Ideas that ride a recent change have a structural advantage: the market is forming, incumbents are slow to react, and early movers can establish themselves before competition arrives.

7. Are There Good Proxies?

A proxy is a large, successful company that is similar to your idea in some way, usually the same model applied in a different market. Before DoorDash, Uber had proven the gig-economy delivery model for rides. DoorDash could point to Uber and say: we are doing for food what Uber did for transportation. Proxies reduce perceived risk. They are evidence that the model works somewhere.

8. Would You Want to Work on This for Years?

Startups take seven to ten years. The overnight success stories are companies that were grinding for years before anyone noticed them. If you are choosing an idea purely because you think it will make money, you will quit when things get hard, and things always get hard.

The founders who survive are the ones who are genuinely fascinated by the problem, not just the potential payoff.

9. Is This a Scalable Business?

Can this business grow without adding headcount proportionally? Software is inherently scalable. You write code once and it can serve millions of users. Consulting is not scalable. Every new client requires more people, more hours, more coordination. The scalability filter eliminates a large number of ideas that might be good businesses but will never be good startups.

10. Is This a Fertile Idea Space?

Some categories of ideas produce winners at a much higher rate than others. B2B software tends to be fertile. The problems are clear, the willingness to pay is high, and the competition is often sleepy incumbents with outdated products. Consumer social networking is less fertile. Winner-take-all dynamics mean one winner and hundreds of losers, and user acquisition costs make it nearly impossible to break through without viral mechanics.

This does not mean you should only work in B2B. It means you should be honest about the hit rate in your category and adjust your expectations accordingly.

The Boring Idea Advantage

One of the most counterintuitive patterns in startup success: boring ideas have a higher hit rate than exciting ones.

Gusto makes payroll software. Founded in 2012 as ZenPayroll by Josh Reeves, Tomer London, and Edward Kim, the company came out of Y Combinator’s Winter 2012 batch. Nobody wakes up thrilled about payroll processing. But thousands of smart people must have noticed that payroll software for small businesses was terrible. Because the problem was boring, nobody tried to fix it until the Gusto founders came along. The company is now valued at roughly nine billion dollars.

Boring ideas get left on the table for years. They accumulate in the corners of industries where nobody wants to look, gathering dust while ambitious founders chase flashy consumer apps and social networks. The competition for boring ideas is structurally lower because most founders are drawn to ideas that sound impressive at dinner parties.

And here is the thing most founders do not realize until they are deep in the work: the day-to-day reality of running a startup is mostly the same regardless of the idea. You write code. You fix bugs. You talk to users. You try to grow. Whether you are building a payroll system or a social media platform, you are spending most of your time on the same types of tasks. The idea determines your daily experience far less than people think.

7 Recipes for Generating Startup Ideas

Recipe 1: Mine Your Own Experience

Go through every job you have ever had, every internship, every life experience. Think carefully: what problems did you encounter? What did you learn that other people do not know? What problems or opportunities were you in a special position to see?

The founders of Vetcove, Alex and Mitchell Kates, grew up working in their father’s equine veterinary practice in New Jersey. They saw firsthand how veterinarians had no good way to compare prices or order supplies online. A genuine problem with minimal competition, sitting right in front of them. Nobody outside the industry would have noticed it. But anyone inside the industry saw it every day. Vetcove now serves more than eighteen thousand veterinary hospitals.

This is the highest-yield recipe for a reason. You already have the domain knowledge and the founder-market fit. You just need to recognize that the frustration you experienced is a business opportunity.

Recipe 2: Build What You Wish Existed

Drew Houston was riding a bus from Boston to New York while studying at MIT and realized he had forgotten his USB drive. He could not work on his files during the trip. Instead of just being frustrated, he built Dropbox, a tool to sync files across computers. This recipe works because you are solving your own problem. You know the pain is real because you feel it yourself, and you can evaluate the solution because you are the target user.

Recipe 3: Listen to Other People’s Frustrations

Same principle, different lens. Pay attention when friends, colleagues, or family members complain about something. The frustration they express is often the starting point of a viable business. They just do not see it that way because they are users, not builders.

Recipe 4: Ride Recent Changes

New technology, new regulation, new platform shifts: what do these make possible that was not possible before? After large language models became commercially available, a wave of AI application companies became viable almost overnight. After remote work became permanent for millions of workers, opportunities in remote-work infrastructure exploded.

The advantage of riding a change: the market is forming in real-time, incumbents are slow to react, and talent is excited to work on the new thing.

Recipe 5: Import a Proven Model to a New Market

Take a business model that works in one industry and apply it in another. DoorDash applied the Uber model to food delivery. Instacart, founded by Apoorva Mehta in 2012, applied it to grocery delivery. Faire, founded in 2017 by Max Rhodes and three former Square colleagues, applied a B2B marketplace model to wholesale for independent retailers.

This recipe reduces risk because the model is proven. You are betting on execution in a new context, not on whether the model works at all.

Recipe 6: Find Broken Industries

Talk to people in different industries. Ask them what is broken. Look for industries that have not been meaningfully updated in decades, where the tools are ancient, the processes are manual, and the incumbents are complacent.

Industries that are still running on fax machines, spreadsheets, and phone calls are prime targets. The technology to fix them usually already exists. The gap is not technical. It is that nobody with the right combination of industry knowledge and technical ability has bothered to build the solution.

Recipe 7: Reimagine Today’s Giants

If Salesforce were starting from scratch today, what would CRM look like? If Google were starting today, what would search look like? This thought experiment helps you imagine the next generation of software, products built with current technology and current user expectations, unconstrained by legacy code and legacy thinking.

Organic vs. Manufactured: What the Data Shows

Perhaps the most striking finding from Y Combinator’s analysis of how their most successful companies got started: the majority of ideas were organic, meaning the founders noticed a problem through their own experience. A smaller portion were manufactured, meaning the founders deliberately sat down to brainstorm or chose an idea from a list.

Among the very top companies, the ones worth ten billion dollars or more, the organic rate is even higher. Jared Friedman, a YC Managing Director, estimates that roughly seventy percent of the best YC companies started with organic ideas.

This does not mean brainstorming is useless. It means you should bias your brainstorming toward areas where you have deep personal experience. Use the recipes above to generate candidates, but always filter them through founder-market fit. The best ideas come from founders who are deeply embedded in the problem space, not from reading trend reports over coffee.

The Unfair Advantage Test

When evaluating your final list of ideas, ask yourself: do I have at least one unfair advantage? Either you know this space better than almost anyone, or you are a world-class builder, or you have early access to some new technology, or you have a distribution channel that competitors lack.

As Dalton Caldwell, YC’s Managing Director, has observed from working with over a thousand startups: the most common failure pattern is when founders feel pressure to work on fashionable ideas instead of the ideas sitting right under their nose, where they are genuinely the world’s expert. The founders who succeed tend to be the ones who resist fashion and follow their expertise.

If you cannot identify any unfair advantage, it does not mean the idea is bad. It means the competition will be pure execution, and in a pure execution race with no structural advantage, the odds are against any individual team.

Start with the problem. Make sure the problem is real, that people are willing to pay for a solution, and that you are the right team to solve it. Then build, talk to users, and iterate. The idea will evolve. But starting from a structurally sound foundation makes everything that follows dramatically easier.

Sources & Further Reading

• How to Get Startup Ideas — Paul Graham
• Organic Startup Ideas — Paul Graham
• How to Get and Evaluate Startup Ideas — Jared Friedman, YC Startup School (YouTube)
• Tarpit Ideas: The Sequel — Y Combinator Library
• Lessons from 1,000+ YC Startups — Dalton Caldwell on Lenny’s Podcast
• How to Get Startup Ideas — Y Combinator Library