YC SUS: Michael Seibel and Eric Migicovsky discuss How to Launch an MVP
These notes were generated by AI and may contain inaccuracies.
Eric is the course facilitator at Startup School. Michael is a partner at Y Combinator helping answer questions. They answer questions about MVPs from the forum, diving in with tailored advice.
Zephyrus is an analytics platform for competitive gamers providing game-specific tools and data to analyze, track, and improve performance. One-liner is clear, especially for those in the space, gets a solid A-.
Consumers Expect More Than Scrappy MVP with Existing Competitors
Consumers should not expect more than a scrappy MVP since startups struggle to build polished products. Find a small population with unsolved problems willing to use something bad—that's a good sign. If users need a polished product even for 10-100 users, competitors are doing well.
Hack: Layer MVP on existing product, e.g., Chrome extension overlaying extra data on software gamers already use for Starcraft 2 analysis. Test hypothesis with least effort: is it important enough for people to care?
Clarify one-liner: "We make software for property managers to automate billing, customer care, and maintenance." Specify efficiency gains, e.g., manage 1 property to 5 or 5 to 15.
Avoid Making Service Too Specific vs. Building Universal Platform
If you have industry knowledge (e.g., former property manager), use insights for general problems. If not, building for first 5 customers educates you—may be expensive, but necessary.
Immerse: Become a property manager, volunteer, or help run a business for a week or two.
Prioritize pain points users are already solving: experimenting with tools, tried competitors. Avoid assuming paper processes need software—some work fine on paper.
Launch MVP for User-Generated Content Platform
Check replay value: reason to return daily/weekly? Find smaller community where it's useful even without everyone, e.g., Facebook at Harvard—one class or grade made it useful with thousands of users.
Focus smaller group with frequent use. Tricks: Generate content yourself—9gag scraped memes, Reddit early days, Sachs Amster. Scrape existing accomplishments from Twitter. CrunchBase built profiles from media reports before people cared.
Aggregate from GitHub, Dribbble—service to generate your page.
Advice for Long Launch Time Products
Careful with MVPs where harm/safety risks. Test smaller versions with off-the-shelf tech: Raspberry Pis, cameras.
Validate demand and demonstrate founder skills. Huge gap between idea and implementation—build to separate from pack.
Embark (self-driving trucks) started with cop car, demoed golf cart with cameras (possibly remote). Comic analogy: skateboard to bicycle to car. Start with autonomous wheelchair or personal mover for retirement communities—safer, cheaper.
Generate sales pre-launch: Letters of intent or pilot agreements (no price needed) show demand to investors.
Sell service first: Run car-sharing in retirement community to prove smartphone requests work—start tomorrow.
Better description: Car mechanic anywhere—refueling, tire changes, annual checkups. Start narrow, not end-to-end.
Descriptions should create mental pictures quickly—don't need to encompass future vision.
Liability and Past Failures
Study every failed startup in YC database—understand unique insight. Price-sensitive customers for car fixes—target cheapest option, not luxury (less demand).
Trust gut on liability with large items people care about; listen to concerns.
Description vague—home gym? Jargon. Website: "Services for your well-being at home"—exercises, meditation, sports.
Launch MVP: Free or Charge?
Charge from day one if real costs (sending people, recruiting, time). Not pure software—avoid freemium.
Providing $1 value for 75¢ loses money; equalizing price drops demand.
Validate unique value: wellness at home (people already buy yoga/dance elsewhere). Charge a lot, even more than alternatives—identifies desperate users. Private yoga at home > studio; if people pay more, you're onto something. Discounts signal only slightly cheaper.
Two one-liners jammed. Study prior local news, interest-based social, dating apps—strong opinion why they suck. Helps build product and answer investors (YC funded many).
Likely personal problem (moved to new city). MVP: Add least new process to workflow—e.g., meetups, city subreddit events—for outsized impact.
Peter Reinhardt (Segment) talk: Most startups dying, convince themselves near product-market fit.
Be skeptic: Does MVP provide value today? Not with 17 future features.
If no, find users in so much pain they'd use it
Ask if the service you could write today provides any value to users. If no, find people in so much pain around this problem that they would use your MVP. Get them to use it and ask again: am I providing value? It's too easy to iterate without providing value—iteration one no value, two no value, three some value. If the problem is severe enough, people will use a really bad product to solve it. Find them.
Best MVP practitioners talk to different groups of users and iterate on the market instead of building more features. Superhuman founder wrote a blog post on iterating to product-market fit. He looked at demographics in first 100 users: founders and salespeople used basic features heavily; product managers dealing with lower email volume got no value and stopped.
Test if people benefit from version one: turn off your product and see complaints. Simulate with outages—check email logs from last week's server downtime to see if anyone cared. This is 'do things that don't scale'—big companies can't, but startups can. Reddit and Twitter did it. In a startup where death is likely, experiment with extreme things.
If people can't live without your website for five minutes, you've got something.
Poshwha—local e-commerce for shopping from home-based small businesses—lacks product-market fit. Should you continue despite thin margins or quit and search for fit?
Hardest question: some founders see immediate failure if not succeeding; others ignore two years of bad facts. Best strategy: time-bound it. Try everything hard for at least six months, then open to pivoting. Easier to be honest about six months of hard work than 'tried all ideas'—ideas are infinite, easy to trick yourself.
During six months:
- Put remaining ideas on whiteboard, pick top one, spend a week working hard.
- Test specific questions: right demographic? Right stores/small businesses?
TaskRabbit founder: early anything went, but realized some wanted IKEA furniture setup. Wish they'd tested landing page 'Get your IKEA furniture set up' sooner—doable in a week using existing framework.
Test hypothesis like 'people want baby stuff short-term' with landing page on existing site for one demographic/use case.
When to quit: Peter Thiel—convince the critic in the company. Identify skeptic's sign that proves something's there; if not hit in time, move. Don't trust optimist—assume pessimist role, set objective measure even skeptic would accept.
Segment example: Peter didn't believe in segments, co-founder did. Test: post on Hacker News. Expected no upvotes; topped front page—'even as skeptic, must acknowledge demand.'
Crystal at Sitefull: after surveys/interviews, MVP live—top methods for feedback?
Startups scale feedback effort too fast. Effective: call users personally, invite to office, host dinner, WhatsApp group. Airbnb: customer service was co-founder Joe's cellphone.
Want so much information flow from customers that you do things that don't scale.
Don't think Intercom/Zendesk yet—you don't know questions to ask. Need high-bandwidth, broad feedback. Talk openly: 'How did you find us? Experience so far?' Write it down, weekly debrief with co-founders.
During office hours, discuss user backgrounds (where they work, etc.) to pull insights for MVP tweaks/experiments like targeting singles if three mentioned it.
Spencer at Carnuba: first online back-office/field service management for independent car care pros (mechanics, detailers). End goal: marketplace connecting car owners to local independents.
MVP: back-office SaaS to build supply side. Pitch end goal or now?
Structure pitch:
- What you're working on now.
- Traction.
- Business model.
- Vision.
Easier to sell vision with context from first three. Without, confusing—'Why not build marketplace today?'
Evaluates founders: smart? Solved pain? Short-term goal gives metrics; pure vision lacks differentiation.
Barter Service MVP with White-Label
Autumn at jump.io: barter service connecting startup entrepreneurs.
Pros/cons white-label for quick MVP? To test demand: great, but Google Sheet could too—go simplest. To make scalable software company raising money: build your own.
Pricing for Fleetcode (India)
Fleetcode: business assistant for transport/logistics—tracks vehicle profit/loss.
Price expensively for high value bar, verify pain. More often underprice than overprice—lean over to force issue. Indian founders: spend 1-2 sentences on market context for SF investors.
Debfood E-Learning
Debfood: e-learning for software dev (text + YouTube, 1K monthly users). Charge subscriptions now or build features?
Depends: lifestyle business needing money? Charge now. Venture-scale free product? Maybe not. Ask users what they want/frustrations. Check engagement.
Be careful asking non-customers for product insights—they'll give bad ones.
Co-Founder Conflict
Deal with conflicts on direction: have respectful conversation early—avoids blowups. Time-bound skeptic's test for direction; run experiment.
YC advice: co-founders room, plan 2-week sprint. All ideas on whiteboard (no deleting), then prioritize. Avoid developers preemptively shooting down based on build difficulty—trade-off benefit vs. cost.
Talking to First B2B Customers
Elias: communicate want to sell + beta test?
Be honest. Seek excited early customers for close founder work, quick feedback/changes—many pay more. Spin positive: 'Work extremely closely, awake 24/7 to make success.' Avoid pretending enterprise—disappoints all.
Same for investors: just ask unclear questions—shows sophistication.
Incentives for Housewives Platform
Oscar: Blue—housewives resell supplier products via WhatsApp (Meesho-like).
Give Uber-style incentives? Uber began with UberBlack in SF (expensive, no subsidies)—taxi shortage pain so great, no need. Incentives came later for expansion.
Don't copy myths. Solve hilarious pain first, charge high. SF Uber: limos/stretch for wealthy; felt like king.
Ashly: new tech service—patent before MVP?
Patents not critical for rapid-growth tech startups. Priority: find what people want, sell, iterate/grow. Non-startup world overemphasizes patents (heard 'got screwed' stories).
Talked to NYC YC founder pre-launch patent worry: benefits mostly post-product-market fit. Projecting future problems prematurely.
Founders project problems they will have in the future. There are a multitude of problems that you may have. Keep focused on the problems you actually have: no users, no product, and no revenue. Solve those first. You will run into patent stuff later.
Rahmi asks: We have a project that solves a very critical and wide pain point but it totally depends on YouTube. If YouTube decides to block it, we will be out of business. Is it a good risk?
This risk could render the business worthless. This is something that has changed over the last 15 years. When entering the startup world, systems were open: Google felt like a very open system for finding websites (didn't have competing products), messaging (SMS) felt open, email was open (most emails sent were received, even promotional ones).
Now, almost all systems are closed. These companies exploited open systems to get big and understand that if their system remains open (YouTube, Facebook), they will be exploited by the next person. You're not only competing with other players in your market; you're competing against the platform you're on. If it's a significant revenue opportunity for the platform, they have every incentive to buy you or kill it. They won't buy you for a lot of money because they have leverage over you.
Ask: Is your dependence on YouTube today or forever? If forever, be very cautious about building this into a business. Consider making it a lifestyle business: monetize quickly so if it goes away, you've made money. Bring in subscription revenue that hardcore YouTubers must have, willing to pay $50 or $100 a month.
The only other way to get away with this is like the eBay-PayPal case: your product becomes so mainstream on YouTube so quickly that if they tried to shut it down, there would be a mass revolt. That's hard.
Example: Remember Vine? Twitter shut down Vine and replaced it with Periscope. Vine was popular with all the cool kids in San Francisco, and they still did it.
Thank you for excellent questions. Back next Wednesday at 11 a.m. Pacific time to answer questions about next week's lectures.
Keep Y Combinator in your library
Save the videos and channels worth coming back to, and find them again in one place.
