YC SUS: Gustaf Alströmer and Eric Migicovsky discuss growth tactics
These notes were generated by AI and may contain inaccuracies.
YC SUS: Gustaf Alströmer and Eric Migicovsky discuss growth tactics
Gustaf Alströmer works at YC as a partner for two and a half years. Previously, spent almost five years at Airbnb on growth and product, from growth team to product manager. Worked on growth at other companies before that. Seen really good stuff like Evan B and really hard stuff at other companies. Delivers several lectures to YC core participants during batch on growth. Main person founders turn to for growth questions.
From My Education Club: Building peer-to-peer lending platform. Christophe talked about getting paying customers ASAP before paid growth. How does this work for social networks or platforms with network effects?
Historically, many social products last ten years based on advertising: charge advertisers, not users. Led to idea not to monetize early, just grow. Twitter, Facebook, LinkedIn, Snapchat grew without paid marketing via viral invites, word of mouth. LinkedIn used email and SEO.
For network effects products (product better with more users), critical to get first million or couple million users. LinkedIn early days useless without coworkers to connect.
Fewer companies today start with advertising because extremely hard to compete with Google and Facebook, best advertising platforms. Not like competing with TV or newspapers.
Most companies now charge users. Example: Twitch (post ad-driven), enables creators to charge streamers, earns percentage of transactions.
ROI calculation: Know money made back from each user acquired via paid marketing to scale. If not, can't scale paid growth; everything experimental, time-boxed, limited. No one funds without clear path to make money.
Misconception: Grow with paid, monetize later. Won't work; failure mode for startups today.
From Michelle at Critique Match: Good to focus on growth before monetization model that might hurt growth?
Monetization hurting growth not good model. Align value from product with payment. Paying to unlock features not best; better pay for repeat usage or retention.
Example: The Athletic, YC-funded, app for local news. Subscription required to use app and read news from 150 journalists in US, pay monthly.
Paid growth increasingly important for mobile apps. Back then, looked down on paid growth as product not good enough. Now, Google and Facebook shifted: hard to get free growth. Facebook paid hard, Google first 4-5 listings paid ads. Platforms monetizing usage, so founders lean into sustainable paid growth.
Airbnb: search marketing through Google critical to growth, core for travel companies. Organic on Google/Crooks harder.
Monetization early to optimize, use to grow. Can't hurt growth. Organic possible to certain level, but sustainable paid growth requires making money.
Synthesize: Paid model early gives knowledge of conversion funnel costs. Know money per user to experiment effectively with paid marketing. Without, experiments without viable outcomes.
Example: The Athletic charges $15/month for local sports subscription. Target Toronto Maple Leafs fans on Facebook: "Download Athletic app, get local news from independent journalists for $15/month." Know conversion rate per click, value per click, scale paid marketing.
From Suren at Ivanka: Tools/software contributing most to success as product growth leader? Liebchen lead automation?
Different tools per company/category: consumer, SaaS, hardware.
Basics: event analytics - Mixpanel, Amplitude, Segment, Heap. Track: people coming in, what doing, flowing through product. Core flows: onboarding, signup, payment conversion, returning user metrics. Google Analytics not enough; recommend Amplitude/Mixpanel + Segment.
From day one. Depending on stage: No automated sales tools early. Do things that don't scale (read Paul Graham article). Find customers manually first 50-100, not via tools.
Example: Founder asking to automate email. Sending to other founders? Send from personal account early. Emails responded to look real, from friends/family/coworkers, plain text, higher conversion.
Customer service: No Zendesk/forum early. Receive all customer emails in inbox, reply as founder. Surprises customers positively.
Airbnb example: Joe and Brian shared cell numbers first couple hundred hosts, kept calling for years. Joe never changed number. Learned a lot shaping product.
Early: Baseline analytics + direct customer interaction, no other systems.
From Haystack: Customers have pain but don't know solution, vague feedback. How take feedback, iterate product?
User research: Ask extremely open-ended questions, don't interrupt, no assumptions.
Airbnb: Launching new product, show screens on laptop/phone to random people in office lobby. "Tell me what you think this is, what this means." If 10 people, most don't get it, not obvious enough. Users alone, no shoulder guidance.
Used usertesting.com (good at time): Pay random person to do simple task in app, record screen, speak thoughts. Watch 10 videos: designers/engineers shocked people don't get it. Most don't understand products; drop assumptions.
Data overload common. User testing cuts through: If data shows problem, 5 people 10-30 min on pain point. Don't guide, tell goal or speak up. 10 people enough for guidance (per someone, maybe Jakob Nielsen?).
Standardize, use clear language. Avoid clever/smart language in app/marketing; goes wrong. Use UI language: "What happens here? Click this button."
How grow multi-sided marketplace, retain users?
Kickstart new marketplace today: Build supply first, artificially to level where demand happy. Example: Launch Uber/Lyft new city - find 10 drivers (friends), commit drive 4-5 days. Artificially 10 drivers on streets, then get demand.
Example: Peeply (YC?), online marketplace for founders test experiments with target audiences. Get 10 good supply people (friends), volunteer time for sessions. Validate demand first, not strangers.
Quality: Early Airbnb allowed listings without user/house photo - dumb. Require quality: nice car/safe driver for Uber. Curate initial supply: Build profiles for 10 friends.
Validate: If demand good experience, repeat in next 10 days - good sign.
Retain demand: Ask experience, reviews, phone feedback. Hard from churned (don't care), but crucial. Debug marketplace: Why bad experience?
Ask phone number at signup early: Call if wrong/great experience. Emails easy to ignore.
Once demand wants offering, figure pricing: Minimum demand pays for quality supply. Example: Usertesting.com pays $45/test, maybe $20 to reviewer. Not everyone does for $20.
Pricing part of product-market fit journey, experiment/adjust. Not everyone rents home at market price.
From Daniel at Storylow (makes custom photography easy/affordable): Outline pros/cons incentivized growth e.g. Airbnb invite friend $25, Uber subsidized rides.
Spent time on Airbnb referral program, significant growth portion.
Referral programs work well if:
- Products already have word-of-mouth
Products that already have word-of-mouth have people telling their friends about them. They are not private products. They sometimes require a friend because the product is radical enough that a friend can convince someone to use it. Great examples are Uber and Airbnb, new products that people might not trust without a friend recommending them.
Having a friend recommend the product makes the new customer more likely to buy, or the problem gets better if friends are on it. People might not trust the product or idea unless a friend recommends it, which is why referral programs work. The current Uber and Airbnb today are very different from their first month; most people were scared of those ideas initially. Only when they became mainstream did people get comfortable. In early days, if the idea is radical, having a friend convince you it's not scary is a fundamental foundation for a good referral program.
Referrals optimally push people over the edge to try it, pushing the recipient over the edge to use the product. It adds a couple percentages on each side, which matters.
Money is important, but can't just be added to any product; the product must kind of require it.
Products need a regularly active user base. One-time use products make referrals less likely to work because you only have one chance to get someone to tell friends. With Lyft or Uber, every time you use it, you're reminded you could give friends $25, with conversion every time.
For Airbnb, after booking, they placed referral prompts in two spots with 20% click-through rate; half ended up sending invites, which was effective. But people don't come back to Airbnb regularly to be interrupted; you use it once a year or when traveling.
Track percent of weekly active users who see the referral link. Don't put it in settings or menu; people don't go there. Can't interrupt in main workflow. Place it where users are doing something post-interaction, like after booking.
Referrals are paid marketing; you need to make the money back. Many get excited but don't make enough on the backend. If you make $10 per user, most Americans won't act for $5 each side. Need minimum $15 or more to give away, otherwise it won't work well.
Airbnb's program is $55 for tutors in the US to the sender—meaningful numbers, 10-50% of a trip for first-time travel, very impactful.
Lounge Looks provides online tools to home-based beauty and grooming professionals to grow and manage clientele. Current acquisition: outreach, cold emails, DMs on Instagram—working but not scalable for faster growth.
When switching to scaling channels: only a few exist—paid marketing, different types of marketing, SEO, word-of-mouth, sales, viral growth.
Figure out which fit the product. Lounge Looks fits marketing on demand side due to targeting, has revenue to pay for it.
Challenge with paid: don't reacquire users. For Airbnb, wouldn't pay for repeat Google ad clicks after first time; want app install for future searches.
Acquire first time via paid, then natural entry like app or email where you don't pay. Marketing most likely, referrals second. If earning money, experiment since you know customer value after closing.
Lead time: how long to close a sale. For Airbnb consumer travel side, consumers susceptible to exclusivity, scarcity, urgency messages—sometimes dark patterns messing with minds without real value.
At Airbnb, provided real information: search San Francisco, only 40% listings left—real urgency to book soon, or get bad deals as good ones are taken.
Surface tactics informally, not overpowering: unique find or rare find labels for listings as positive signals.
These work well because users react to them. Prefer they didn't, but reality. Ensure tactics are user-friendly, provide real value, not just psychological manipulation.
For B2B, e.g., SaaS service where employee gets $500 for referring another customer: consider likelihood of knowing others in similar role buying the service.
Payroll software in founder community likely; mainstream less so.
Legal issues: can't give $1000 to pocket after convincing company—bribery. Gusto (payroll) has prominent refer for $500 cash.
Works well for small businesses where owner decides: $500 meaningful to beauty salon owner referring payroll software.
For small businesses, not a bad idea. For larger, complicated.
Smart ways: use Google contacts to suggest who needs it. During customer interviews or onboarding, ask if friends might be interested, who they are.
Roles like recruiting move between companies, good at sharing software/tools.
Two types: educate market on new category, or compete with incumbent.
For B2B: show product with screenshot or video—easier than explaining. Include specific recognizable thing: "a-ha, that's what I can't do myself."
Include social proof: won't try new payroll software, but yes if Dropbox uses it.
Ties to referrals: trusted friend introducing new product.
Consumer: Soylent new; built community forum where users told others to try it.
New consumer products tend to have word-of-mouth if not private (e.g., bathroom products like hair loss not shared).
Hair loss products flipped it by bringing into conversation.
Get brand association: media write-ups on category, even if not specific product. Attach to rising category for social proof.
False question: startups do everything. No black-and-white; tension always there, tightly coupled.
Bad quality kills growth; bugs/tickets cause churn. Retained users drive growth.
Understand what drives growth, retention, churn—what makes people happy/unhappy. Then prioritize.
SEO big channel; people use Google for non-daily questions, search volume changes yearly (e.g., new politicians, "what is Soylent").
Capture keywords for something new. Hard for established like "what to do in Paris"—competitive.
Think of questions people ask, not just keywords.
Two ways: write high-quality content to rank, or automated content (e.g., Airbnb city/listing pages).
Use Google Keyword Planner to check traffic before starting.
Channels change often. Rarely see sustained success to large user base; tactics saturate, platforms charge via ads.
Instagram no links to prevent spam/commercial hijacking.
Scalable: buy ads. TikTok spent a billion on Facebook/Instagram ads pre-2018/2019, then organic discovery.
Arcade: service could take up to a year.
Dropouts indicate dislike.
Two metrics: dropouts, and predictors of dropouts (e.g., stop taking course correlates with dropout).
Measure early signals of bad behavior: 1-star review highly predictive of not returning; no search for another year after first booking.
Find predictive metrics for good/bad behavior.
High Low: platform for anonymous workday high/low points with team leaders.
Company of 100-150 employees wants to implement at MVP stage; no idea how to handle.
If demand, ask for price to see what they're willing to pay. If $50/month per
If they're willing to pay $50 a month per employee or $20 a month per employee, that's a great first customer. If not, they might not actually use it or care enough about what you're building. The idea sounds new and interesting, but people might say 'that sounds interesting, I want to try it' without committing to paying $50 a month. Figure out their value, which might take time, but people are excited and trying even an MVP. The danger is taking this as positive feedback and building without interacting with customers. Go back and ask if they'd be a paid customer for the early MVP. If yes, you have your first customer. If no, identify the delta between your MVP and what they'd pay for.
Luggage sharing is novel in India with a huge untapped market. Personal use case: checked out of Airbnb in Harlem, heading to Penn Station for train to DC, didn't want to carry bag for a couple hours. Hotels like Everybody do this, but Airbnbs checkout at 11 a.m., leaving you to carry luggage the rest of the day—clear need, but no great product exists.
Build an app: users find it via Google, install and keep on phone. Next time, open app, see map with storage spots by typing address. Build trust: experienced storing luggage at restaurant side felt unsafe, like someone could grab it (had computer inside). Offer guidance to storage locations and guarantees for safety.
Don't try to put it in front of people every checkout—hard for small startup to partner with Everybody unless you have thousands of locations. Integrate with hotel bookings, rental cars, Airbnb. Talk directly to Airbnb hosts: include pre-formatted line in checkout emails with address, directions, adding luggage storage option.
Start in big cities like New York, London with lots of travelers needing luggage solutions, especially walkable cities (not LA where you drive). Focus on airports, train stations where people go next—that's where they'll drop off/pick up. Innovation: open app, push button, someone picks up luggage within 10 minutes (piece together with on-demand like Uber), drive to pickup spot. Would pay $15-20 bucks to avoid dragging luggage all day—doesn't exist yet.
In Japan, drop luggage at any 7-Eleven, sent via network to hotel—cool system. Once people discover this easiest way, they'll do it forever. More innovations: send luggage back to home country. Existing companies are small, use websites not apps, cheap, use hotels for revenue, no great plans. Opportunity to build: people will pay $20-30 for a day.
Metro Push sends push notifications from public service broadcasters. Offering one year free before ad-free subscription—don't need a year free. Longest sensible: 2 weeks or month to onboard, try product, see value. Example: assign Atlantic, read four articles on Toronto FC, realize 'this is great app'—that's enough, no need for more free months. Start charging then.
Ad revenue vs. subscribers: 10-20X difference, make more from fewer subscribers. Ads difficult on generic platform—CPM low, earnings zero to 50 cents/month max (likely couple cents to 10 cents/user/month with high engagement). Can't acquire customers, service, support, buy content for that—ad networks don't build sustainable companies. Investors won't fund if serious about ads only.
Users comfortable with micro-transactions, subscriptions. Substack (YC company ~1 year ago) enables email newsletters you charge for—revolutionary then, but people pay for reliable high-quality content from good writers. Aligns business model with users, solves societal problem—not in conflict like Facebook or newspapers.
Win as B2B by solving existing need with consumer-grade growth skills competitors won't use (traditional: case studies, old-school marketing, big events). They skip invite flows, targeted Facebook, great referral programs built into product.
Grow inside company: connect with Googler, get emails from co-workers via domain (e.g., mycamera.com look-alike audiences).
Intercompany virality trickier. Examples: Notion, Airtable, Rally (collaboration docs). Viral docs like Burning Man packing list (10,000s users). Replicate viral spreadsheets on Airtable (e.g., wedding packing list). Make embeddable: embed Airtable on website, drives traffic.
Slack: add employee from another company to your Slack room without their own Slack—great method.
For virality, think about collaborative content vs. just invites. Build into product: Google Docs—@ someone, they comment regardless of account—viral. Use address book/Google Apps for inside-company growth: share to contacts, give accounts.
Make every document easily copyable. YC company Be Found(ed?) creates airslash links (e.g., airslash/something)—physically viral as people share.
Lots of actionable advice. Back next week for another livestream.
Keep Y Combinator in your library
Save the videos and channels worth coming back to, and find them again in one place.
