How to Get and Evaluate Startup Ideas | A Keynote and Personal Reflection
A practical framework from Y Combinator's Jared Friedman for how to generate and evaluate startup ideas. Covers the 4 most common founder mistakes, 10 key evaluation questions, 3 counterintuitive signs of a good idea, and 7 proven recipes.
Opening
During my daily plank exercise routine, I ran into a great video from Y Combinator's Startup School, titled "How to Get and Evaluate Startup Ideas | Startup School" that teach founders how to generate and evaluate startup ideas. The original concept I had were more crude and not as sophisticated as the ones in the video. Recently, I also ran into multiple startup ideas (being interviewed or consulted by founders) that I wish they have known the frameworks in the video. By putting this talk into words and make the content more accessible, I hope to help more founders and potential founders to generate and evaluate startup ideas effectively.
If you are here with a potential startup idea, congratulations! The best way to use this blog post is to copy the whole article into a markdown file, and ask your LLM to attack your startup idea with the frameworks in the article. This makes your idea more structured and defensible in front of potential co-founders, investors, and future employers.
The Four Most Common Mistakes
Jared opens by identifying four ways good founders end up with bad ideas — the first two are about picking the wrong type of idea, and the last two are about how you approach the search itself.
[CM-1] Not Solving a Real Problem (SISP)
The most common mistake is working backward from a technology to find a problem — a "Solution In Search of a Problem" (SISP) — instead of starting with a real problem people genuinely care about. The failure pattern looks like: "AI is cool. What could I apply it to?" You end up with a superficially plausible problem rather than a real one. If people don't genuinely care about the problem, they won't care about your solution. The fix: fall in love with a real problem first. Good starting problems are specific and tractable — not abstract societal challenges like global poverty, but something narrow enough for a startup to genuinely address.
[CM-2] Getting Stuck on a Tarpit Idea
Tarpit ideas are a recurring class of startup concepts that look attractive from a distance but have hidden structural reasons why they have never succeeded, trapping founders in months of fruitless work. The canonical example: an app to help friends plan meetups more efficiently — submitted to YC for 20+ years with no success. These ideas are tantalizing because they address widespread problems nearly every founder encounters, but an invisible structural barrier prevents any solution from working. Before pursuing a tarpit idea: (1) Google it to find everyone who has tried it before, (2) talk to those founders, and (3) name the specific structural barrier that stopped them. If you cannot name the barrier, you probably don't have a plan to overcome it.
[CM-3] Picking the First Idea Without Evaluating It
Jumping into the first idea that comes to mind — without stopping to assess whether it could be a real business — burns months of founder energy on a bad foundation. The fix is not to be so rigorous that you become paralyzed; that is the opposite failure mode. Evaluate the idea enough to confirm it has promising qualities using the 10 key questions below. No startup launches with exactly the idea it succeeds with anyway; what matters is that the starting idea is interesting enough to morph in the right direction.
[CM-4] Waiting for the Perfect Idea
Founders who sit waiting for a perfect startup idea before they begin never actually start a company, because no startup idea is perfect at inception. The healthy position is somewhere in the middle of a spectrum between picking blindly and waiting indefinitely for certainty. Think of your idea as a good starting point, not a final answer — one with enough interesting qualities to be worth pursuing.
Ten Key Questions to Evaluate Your Startup Idea
Use this as a structured checklist. A strong idea doesn't need to pass all ten, but each "no" is a risk to understand and account for.
[KQ-01] Do You Have Founder Market Fit?
If Jared had to pick one most important criterion, this is it: are you and your team the right people to be working on this specific idea? PlanGrid example: Tracy had construction industry knowledge, Ralph had the developer skills to build an iPad app — together, they were the obviously right team. Reframe your search from "is this a good idea?" to "is this a good idea for my team?" An idea that is excellent for someone else but not matched to your team's strengths is not a good idea for you. Start your search by asking what you are exceptionally equipped to build.
[KQ-02] How Big Is the Market?
You need either a large market today ($1B+) or a market that is small now but growing rapidly enough that it will be large by the time you scale. Coinbase launched in 2012 when the Bitcoin trading market was minuscule — but if Bitcoin succeeded as people hoped, this would eventually be a billion-dollar market. Market size at time of entry matters less than market size at scale.
[KQ-03] How Acute Is This Problem?
The strongest problems have no good existing solution. Brex example: before Brex, YC startups literally could not get a corporate credit card — no bank would issue one to an early-stage company. The alternative to Brex was not a worse credit card; it was nothing at all. Look for situations where users are working around the problem with painful hacks, or where an entire workflow breaks down without a solution. The more unavoidable and painful the problem, the less convincing you need to do when you approach users.
[KQ-04] Do You Have Competition?
Counterintuitively, the presence of competitors is usually a good sign — it confirms the problem is real and a market exists. No competition often means no one actually wants the product. What matters is whether you have a genuine insight the existing players are all missing. Entering a competitive space with a sharp differentiation is often an advantage; entering without a clear insight is the actual problem.
[KQ-05] Do You Want This? Do People You Know Want It?
If neither you nor anyone you personally know would use this product, that's a serious signal that maybe nobody wants it. This is a fast gut-check, not a definitive market study. Founders who personally experience the problem they're solving have a natural feedback loop — they know firsthand when the product is getting better or worse, and they're motivated in a way that sustains the grind.
[KQ-06] Did This Only Recently Become Possible or Necessary?
A change in the world — new technology, a regulatory shift, a new behavior pattern — often creates a startup opening that couldn't have existed before. Checkr example: the rise of DoorDash, Instacart, and Uber created sudden demand for high-volume, fast background checks on gig workers. Existing background check companies were built for slow enterprise HR workflows and weren't suited for this new use case. Ask: what has changed recently that makes this problem newly solvable or newly urgent? If the answer is "nothing," it's worth asking why nobody else has built this already.
[KQ-07] Is There a Proxy for Your Idea?
A proxy is a successful company doing something similar in a different market — it proves the model works without requiring you to prove first principles from scratch. Rappi used DoorDash as a proxy: DoorDash's US success was concrete evidence that on-demand food delivery could be a billion-dollar business, and Rappi applied that model to Latin America where no equivalent existed. A good proxy answers the investor question "does this model work?" with real evidence.
[KQ-08] Is This an Idea You'd Want to Work on for Years?
Passion at the start is a positive signal, but absence of passion is not disqualifying — it often develops as the business gets traction. Many of the best startup ideas are in "boring" spaces like payroll software, tax tools, or compliance platforms where initial enthusiasm is low but business fundamentals are strong. Once you're actually running a successful business, you tend to become passionate about it regardless of how exciting it sounded at the start. Don't let absence of initial excitement veto an otherwise strong idea.
[KQ-09] Is This a Scalable Business?
Pure software scales infinitely — if you're building pure software, check this one off. The trap is businesses that feel like software companies but hide a service dependency: agencies, dev shops, or platforms where each customer requires significant human work to deliver value. If revenue grows proportionally to headcount rather than decoupling from it, you're building a services business, not a scalable startup.
[KQ-10] Is This a Good Idea Space?
An idea space is a class of closely related startup ideas — like fintech infrastructure or vertical SaaS. Different spaces have wildly different hit rates. Over the last decade, fintech infrastructure and vertical SaaS produced an astonishingly high proportion of billion-dollar companies; consumer hardware, social networks, and ad tech had far lower success rates. A good idea space matters for two reasons: first, if your initial idea is wrong, a fertile space lets you drift into adjacent ideas without starting over; second, idea spaces flip over time, so pick one you expect to remain productive. Fivetran illustrates this: after two failed pivots within data analysis, they eventually stumbled into a product companies actually wanted — because they kept shopping in a fertile space.
Three Counterintuitive Signs of a Good Idea
Most founders shy away from these three types of ideas, which is precisely why they stay on the table for smarter founders to claim.
[GI-1] Ideas That Are Hard to Get Started (Schlep Blindness)
Paul Graham wrote about this in his essay "Schlep Blindness." Stripe example: thousands of developers knew that integrating credit card payments was broken, but not one of them even tried to build what became Stripe. Getting started required negotiating deals with banks and mastering arcane payment infrastructure — things that felt so hard that they filtered out everyone except the Stripe founders. That filtering left a $100B opportunity unclaimed. If an idea requires an uncomfortable but one-time schlep to get off the ground, that difficulty is protecting you from competition. Ask "what's the schlep here?" before dismissing an idea as too hard to start.
[GI-2] Ideas in a Boring Space
Gusto example: payroll software was obviously broken and widely used, but its dullness caused founder after founder to ignore it for years. Boring ideas have a higher hit rate than exciting ones (restaurant discovery apps, music recommendation apps) because they are systematically under-explored. Here's the key insight: the day-to-day reality of building a startup — coding, debugging, talking to users, grinding on execution — is largely the same regardless of how glamorous the original idea sounds. Initial excitement fades within months; what remains is execution. If the business fundamentals are strong in a boring space, that is exactly where you want to be.
[GI-3] Ideas With Existing Competitors
Dropbox entered a market with approximately 20 existing cloud storage companies. A naive reading: saturated, bad idea. A savvy reading: 20 companies exist and most people still don't use any of them — which means the problem is genuine and none of the incumbents have actually solved it. Drew's specific insight was that all existing products required manually uploading files through a website. Integrating directly with the OS to sync files automatically was a step-function improvement in convenience. The formula: competitors prove the problem exists; your specific insight proves you can solve it better than they have.
How to Come Up With Startup Ideas
The best startup ideas are noticed organically, not generated deliberately — at least 70% of YC's top 100 companies discovered their ideas this way, not by sitting down to brainstorm. When founders deliberately try to think of startup ideas, they tend to produce the same tarpit ideas everyone else produces. Organic ideas surface from real problems the founder has personally experienced or observed from a position of expertise.
Three ways to set yourself up for organic ideas over the long term: (1) become an expert in something valuable by working at the frontier of a field; (2) go work at a startup — you will gain domain expertise in what that company does, putting you in a position to see adjacent problems; (3) if you are a programmer, build things you find personally interesting even if they are not obviously businesses. Replit started this way — something founder Amjad built for himself before it became a company.
Seven Recipes for Generating Startup Ideas
Listed in order of how likely each is to produce a genuinely good idea. Start with the first ones.
[R-1] Start With What Your Team Is Especially Good At
Begin your idea search within the domains where you already have deep expertise — any idea you find there automatically has founder market fit built in. Rezi example: founders with backgrounds in real estate and debt financing only looked at ideas at the intersection of real estate and fintech — a space that had already produced multiple billion-dollar companies. Because they knew the space deeply, their search was quick and they converged on a strong idea fast. If you are a young founder without deep domain expertise yet, this recipe may not apply — try the ones further down the list.
[R-2] Start With a Problem You Have Personally Encountered
Problems you have lived firsthand are more likely to be real, and an unusual vantage point means the opportunity may have been invisible to everyone else. VetCove example: the founders' father was a veterinarian who ordered supplies by calling 1-800 numbers. Thousands of vets knew this was broken, but vets don't start tech companies — and programmers in Silicon Valley had no idea this problem existed. The intersection of "person who knows the problem" and "person who can build the solution" rarely overlaps naturally. If you sit at that intersection due to family background, a past job, or an unusual life experience, that is a significant signal to explore. Practical instruction: for each founder on your team, go through every job, internship, and life experience and ask — what problems did I encounter? What do I know that others don't? What opportunities was I in a special position to see?
[R-3] Think of Things You Personally Wish Existed
DoorDash example: the founders were Stanford undergrads who wanted to order food from local restaurants and have it delivered to their dorm. Before DoorDash, they couldn't. So they built it. This is the most intuitive recipe and the one founders default to most often — but it is also the recipe most likely to lead into tarpit ideas, because people wish for things that have structural reasons for not existing. Before acting on it, ask explicitly: "Is there a reason this doesn't exist yet?" If you can't answer that question confidently, dig before you build.
[R-4] Look for Things in the World That Have Changed Recently
A major change — new technology, a regulatory shift, a behavior shift like COVID — invalidates old assumptions and creates new openings that didn't exist before. Gathertown example: the founders were working on an idea that wasn't going well when the pandemic hit in 2020. They recognized that a massive behavioral shift had just occurred and that new categories of software were now needed, and pivoted accordingly. Ask: what assumptions about how people live or work have just been invalidated, and what products become newly possible or necessary as a result?
[R-5] Look for Recently Successful Companies and Find New Variants
If a business model has just proven to work in one context, ask whether you can apply it to a different geography, vertical, or customer segment. Nuvocargo example: founder Deepak explicitly modeled it as "Flexport for Latin America" — specifically for US-Mexico cross-border freight. He picked it for analytical reasons: large market, proven proxy in Flexport, and personal connections that would let him get started even without deep domain expertise in freight. This recipe works especially well for founders who are operationally strong and willing to learn a new domain.
[R-6] Talk to People and Ask Them What Problems They Have
Systematically interview people within a promising idea space to surface real problems rather than generating ideas in isolation. A2B example: the founders were young with little domain expertise, but they identified the trucking industry as a potentially fertile space — large, not yet disrupted by software — and spent an entire YC batch talking to truckers, fleet operators, and founders in adjacent areas. They eventually discovered the fuel-card problem and built A2B. Best used when combined with a pre-selected idea space: don't talk to everyone, talk to people in a specific domain you have already decided is worth exploring, and talk to other founders who know the space, not just potential customers.
[R-7] Look for Big Industries That Seem Broken
Any large industry that is clearly malfunctioning is likely ripe for disruption — the combination of scale and obvious dysfunction is a reliable signal that a real, unsolved problem is waiting to be addressed. No personal connection to the problem is required; this recipe just requires the pattern-recognition to look at an incumbent sector and recognize when something large is broken. Jared also adds a bonus recipe: find a co-founder who already has an idea. Startup School's co-founder matching tool lists many founders actively looking for a technical or business partner — if you have neither a co-founder nor an idea yet, this can solve both gaps simultaneously.
Personal Reflection
Of all the frameworks in this talk, the one that hit me hardest was [KQ-01] — founder market fit. I've sat across from founders who were genuinely passionate and technically strong, but working on ideas in domains they'd never touched. The passion was real, but when I asked "why you?", the answer was always some version of "because I think this is a big market." That's not founder market fit — that's market research. The two are not the same thing.
The tarpit idea pattern ([CM-2]) also rang very true. I've been on the receiving end of pitches for "a better way to coordinate group plans" and "an app to connect people with shared interests" more times than I can count. What strikes me is not that the founders were unsophisticated — often they were quite sharp — it's that they had never asked "why has no one solved this in 20 years?" That single question would have saved them months.
Looking back at my own startup thinking before I found this framework: I was guilty of [CM-1] and [CM-3]. I had ideas that started with a technology I was excited about and worked backward to find a problem. I had ideas I got attached to without stress-testing them. The 10-question checklist above would have been a useful cold shower.
The insight that surprised me most was [GI-2] — boring ideas have a higher hit rate. I'd always assumed the grind of building a boring business would feel worse than building a glamorous one. But Jared's point holds: six months into any startup, the daily reality looks largely the same regardless of how exciting the original concept sounded. The excitement of the idea doesn't persist into execution; the quality of the business does. That reframe genuinely changed how I evaluate ideas — including my own.
Closing
The framework Jared presents is not a checklist that guarantees success — it is a toolkit for thinking more clearly at the stage where most founder energy is wasted: picking what to build. The key insight I took away is that the two extreme mistakes (jumping in blindly and waiting forever for the perfect idea) are mirrors of each other, and most of the common mistakes — SISP, tarpit ideas, and the rest — are just elaborate versions of those two traps.
The three counterintuitive signs of a good idea ([GI-1] hard to start, [GI-2] boring, [GI-3] has competitors) stuck with me most. They explain something I've observed firsthand: the ideas that sound exciting and obvious at a cocktail party are usually the most dangerous ones to pursue, while the ideas that make people say "that sounds like a lot of work" or "why would anyone build that?" are often the ones worth building.
If you have a startup idea, the most useful thing you can do right now is work through the 10 key questions above honestly — especially [KQ-01] founder market fit. If your answers reveal gaps, that's not a reason to abandon the idea; it's a map for what you need to validate before you commit. And if you want a sharper evaluation, copy this article into a markdown file and ask an LLM to apply the framework to your idea. The labels (CM, KQ, GI, R) give it enough structure to generate pointed feedback.
Finally, if you found this useful, watch the original YC Startup School video — "How to Get and Evaluate Startup Ideas | Startup School" — for Jared Friedman's full delivery, including the company stories and live examples that don't fully translate into text.
Appendix: Example Startups Mentioned in the Talk
The following companies are cited in Jared Friedman's talk as concrete examples of the frameworks above.
- PlanGrid — Founder market fit example: Tracy (construction background) and Ralph (developer) were obviously the right team to build an iPad app for viewing construction blueprints.
- Coinbase — Small but rapidly growing market example: when founded in 2012, the Bitcoin trading market was tiny but clearly had potential to become a billion-dollar market.
- Brex — Acute problem example: before Brex, startups literally could not get a corporate credit card from any bank, making the alternative to Brex's solution "literally nothing."
- Checkr — Recently possible idea example: DoorDash, Instacart, and Uber's rise created sudden demand for high-volume background checks that existing slow incumbents couldn't serve well.
- DoorDash — Used twice: as context for Checkr (a major customer driving new background check demand), and as a "things you personally wish existed" recipe example — founders were Stanford undergrads who wanted food delivered to their dorm.
- Instacart — Context for Checkr: one of the gig-economy delivery services whose rapid hiring created the new use case for background check APIs.
- Uber — Context for Checkr: one of the gig-economy platforms that drove demand for the new API-based background check use case.
- Rappi — Proxy example: DoorDash's success in the US served as proof that food delivery in Latin America (Rappi's market) would work.
- Flexport — Proxy for Nuvocargo: Flexport's success in freight logistics was cited as evidence that a similar model focused on US–Mexico imports could work.
- Fivetran — Good idea space example: after two failed pivots within the data analysis space, they eventually stumbled into a product companies actually wanted, showing that a fertile idea space lets you drift into the right idea.
- Stripe — Schlep blindness example: thousands of developers saw the problem with credit card integrations but were scared off by hard steps like dealing with banks, leaving a $100B opportunity on the table.
- Gusto — Boring space example: payroll software was obviously broken but ignored by founders because it seemed dull, giving Gusto a clear path with little competition.
- Dropbox — Existing competitors example: Dropbox was the ~20th cloud storage company, yet succeeded because Drew identified the specific insight all competitors missed: syncing files automatically via OS integration instead of manual uploads.
- Replit — Organically interesting example: founder Amjad built it as something he personally found interesting, not as a deliberate startup idea.
- Rezi — Team expertise example: founders with real estate and debt-financing backgrounds searched only within that idea space and quickly found a strong idea (Opendoor for rental apartments) with perfect founder market fit.
- Opendoor — Comparison point for Rezi: Rezi's idea was described as "Opendoor for rental apartments," with Opendoor's model serving as the known reference.
- VetCove — Personally encountered problem example: founders' father was a vet who ordered supplies via 1-800 numbers; they were uniquely positioned to see the gap that thousands of vets lived with but no programmer had noticed.
- Gathertown — Recent world changes example: founders pivoted when COVID made it obvious that new remote-hangout behaviors had created fresh startup opportunities.
- Nuvocargo — New variant on a recent success example: founders analytically modeled it as "Flexport for Latin America" (US–Mexico imports) and picked it based on market size and proxy evidence even without deep domain expertise.
- A2B — Talking to people example: young founders with no domain expertise picked the trucking industry as a fertile idea space, then systematically interviewed people to find the fuel-card problem.