#171 — Consumer 2.0: The return of consumer software

AI features that create unique user data flywheels build defensibility, not generic LLM wrappers. Examples: Notion AI learns from your workspace structure and writing patterns (proprietary data moat), while generic ChatGPT wrappers have zero switching costs. Focus on AI that gets smarter with usage, creates personalized outputs impossible to replicate cold, or enables new behaviors (like Character.AI's conversational companions that couldn't exist pre-LLM). The key is AI as enabler of network effects, not AI as feature.

Three channels remain underpriced: (1) Community-led growth through Discord/Telegram with engaged micro-communities (BeReal's initial spread), (2) vertical-specific influencers in undermonetized niches like finance TikTok or dev YouTube (avoid oversaturated beauty/fashion), and (3) AI-generated SEO for long-tail queries where incumbents haven't optimized. Email, paid social, and broad influencers are 100% optimized with CACs exceeding LTV for most startups. TikTok Shop and Roblox ads show early promise but are rapidly getting expensive.

Consumer startups need 18-24 months minimum versus 12-18 for B2B, with 50-100% more capital. Why? Consumer requires crossing network effect thresholds (typically 100K+ engaged users), surviving incumbent copy cycles (3-6 months to detect, 6-12 to respond), and iterating growth channels 5-10x until something works. Discord raised $40M before product-market fit; Snapchat raised $485M pre-revenue. Budget for $5-7M seed + Series A within 18 months or risk getting stuck in the 'zombie zone' with traction but insufficient scale to defend.

Cohort retention by network density is the signal. If users with 5+ connections have 60%+ Day-30 retention versus 20% for users with 0-1 connections, you have real network effects. WhatsApp saw this early (90%+ retention for users who messaged 5+ people). Also track: organic invite rate (% of users inviting others without prompting—Instagram hit 40%+ in 2010-2011), time-to-value compression as network grows (should decrease), and % of sessions involving network interactions. Vanity: total downloads, page views, single-player engagement metrics.

If building top-tier consumer, SF/LA/NYC are non-negotiable for first 2-3 years. Why SF specifically: (1) 60% of consumer-focused seed capital is here, (2) talent density for consumer product design/growth is 10x anywhere else, (3) incumbent exec network for early advice (ex-Meta, Snap, etc.), (4) faster iteration cycles from in-person community. TikTok, despite being ByteDance-owned, put their US consumer team in LA (not remote). Discord was SF. Remote works for B2B where sales cycles allow async; consumer requires daily in-person iteration on psychological nuance. Exceptions: Gaming (can be remote), local-first apps (launch in target city).

Three red flags indicate a closed window: (1) Incumbent already tried and killed it (Facebook launched Facebook Dating—hard to beat), (2) your pitch includes 'like X but better UI' (incremental improvements don't overcome network effects—remember Ello vs Facebook, Vero vs Instagram), and (3) requires behavior change + platform shift (too many dependencies). Green flags for open windows: enables new behavior impossible 2 years ago (AI-native social apps), serves newly possible audience (AI tutors for personalized learning), or creates 10x better experience through technical breakthrough (Snapchat's ephemeral messaging). Test: If your idea could've been built in 2020, it's probably too late.

3-5 people with AI leverage can reach initial scale, but you need specific roles covered: (1) Design-focused founder who understands psychological nuance, (2) technical founder who can ship daily (mobile + backend), (3) growth/marketing founder who lives in channels. With AI tools, a 3-person team can now do what required 8-10 in 2019 (AI for content generation, customer support, basic analytics, testing). However, plan to scale to 8-12 people within 6-9 months of traction—consumer iteration speed demands it. Discord was 6 people at launch; Instagram was 13 at acquisition. Below 3 people, you lack domain coverage; above 15 pre-PMF, you lose speed.

Don't compete—attack orthogonally with a wedge they can't copy for 18+ months, then build moats during that window. Successful strategies: (1) New platform: BeReal succeeded on 'anti-Instagram' positioning Meta couldn't easily copy without cannibalizing Instagram, (2) Privacy/trust wedge: Signal grew by being 'not Meta' during privacy backlash, (3) Demographic shift: Snap owned teens while Facebook skewed older (took Meta 3 years to respond effectively), (4) AI-native: Character.AI built conversational AI social network Meta couldn't replicate without gutting Facebook's feed-based model. Avoid: Better UI, more features, or incremental improvements—incumbents ship these in weeks.

$150K-250K/month for a seed-stage consumer team (5-8 people) in San Francisco. Breakdown: $100-150K salaries (lower cash, meaningful equity), $20-30K infrastructure (AWS, mobile services, AI APIs), $20-40K paid acquisition testing (fail fast on channels), $10-20K tools/ops. This assumes lean consumer SaaS or app; games and hardware are 2-3x higher. Post-PMF, expect $400K-800K/month as you scale team to 15-20 and ramp growth spend. Red flag: burning >$300K/month pre-PMF usually means team bloat or paid growth masking organic failure. Reference: Discord burned ~$200K/month pre-PMF (2015 dollars = ~$270K today).

Raise pre-launch if you're a proven founder; post-traction if you're first-time. Proven founders (prior exit, VP at consumer company) can raise $2-4M on team + idea + mock-ups—investors bet on you. First-time founders need proof: 10K+ organic users, 40%+ Day-7 retention, or viral coefficient above 0.6. The bar is much higher than B2B where you can raise on 3-5 paying customers. Exception: If you have unique insight or access (ex-Meta PM building for Gen Alpha, PhD in relevant AI domain), you can raise earlier. Timing: Budget 3-4 months for fundraising, so start when you have 6-9 months runway remaining. Missing the raise window kills most consumer startups—86% of failed consumer startups die from running out of cash, not product failure.

If your TAM can't support a $1B+ outcome, you're too small for institutional VC (though may fit angels or strategic investors). Calculate: target user base × monetization per user × competitive share. Example: niche social app for rock climbers (2M US climbers × $50 annual monetization × 30% share = $30M annual revenue = ~$300M valuation). Compare: Discord's gaming wedge addressed 200M+ PC gambers × $5-10 ARPU × potential 40% share = $400M-800M annual revenue potential = $4-8B valuation. Too-small ideas can still be great bootstrapped businesses ($5-20M annual profit) but won't attract seed VCs needing 100x returns. Red flag: if you describe your market as 'niche' or 'focused', VCs hear 'too small'.

Three proven paths: (1) Your network from Big Tech—if you worked at Meta/Google/Snap, recruit former colleagues ready to leave (highest success rate—Instagram, WhatsApp both found technical co-founders this way), (2) Build low-code prototype first—use Retool, FlutterFlow, or AI coding tools to create working version, then recruit engineer who sees your execution ability (de-risks you as non-technical founder), (3) Startup communities—YC Co-Founder Matching, On Deck, South Park Commons (avoid generic meetups—too low quality). What doesn't work: cold LinkedIn, equity-only offers, 'I have the idea' pitches. Budget 3-6 months of active searching. If you can't find technical co-founder after 6 months, either your idea isn't compelling or you lack credibility—both solvable by building traction first with no-code tools.

Freemium with premium features (Discord Nitro, Spotify Premium) and transaction fees (Whatnot, StockX) are proving most durable. Advertising-only models require 100M+ users to hit venture scale (see TikTok, Pinterest), making them unsuitable for most startups. Subscription models work if you have daily engagement (Duolingo, Calm at $70-120 annual ARPU). Avoid: one-time purchases (no recurring revenue), pay-to-play social features (alienates free users), intrusive ads pre-10M users (destroys growth). Emerging: AI credits/consumption pricing (Character.AI's $10/month for unlimited chats), creator revenue shares (YouTube-style models), and embedded fintech (Robinhood's margin, payment for order flow). Target $3-8 monthly ARPU at scale for venture outcomes.

Realistic timeline: 12-18 months for viral hits with product-market fit, 24-36 months for steady growers. Examples: BeReal hit 1M in ~10 months (2021-2022 growth), Lemon8 reached 1M in 5 months (TikTok distribution advantage), typical social apps take 18-30 months. Key milestones: 10K users in 3-6 months (initial PMF signal), 100K in 8-12 months (proving retention + virality), 1M in 12-24 months. Red flags: taking >6 months to reach 10K suggests weak PMF, taking >36 months to reach 1M means insufficient viral coefficient or growth investment. Instagram hit 1M in 10 weeks (2010), but that pace is impossible today—market is saturated and channels are expensive. If you're not at 100K users within 18 months, pivot or shut down.

iOS-first for US/Western markets, Android-first for India/Southeast Asia/Latin America. Why iOS works: 60% of US app revenue despite 40% market share, affluent early adopters, easier development (fewer device variations), better monetization ($8-12 ARPU vs $2-3 Android). Launch strategy: iOS beta for 3-6 months, achieve 40%+ Day-7 retention and 0.4+ viral coefficient, then expand to Android. Exceptions: (1) Targeting non-Western markets (Android 80%+ share in India), (2) building developer tools (developers use both), (3) enterprise focus (varies by company). Instagram, Discord, Clubhouse all launched iOS-only initially. Web-first is viable for social apps with desktop behavior (Discord, Notion) but mobile-first dominates consumer. Don't build both simultaneously pre-PMF—splits focus and slows iteration speed.

Essential AI stack: (1) Cursor/Windsurf for engineering (5x faster coding for consumer apps, especially useful for non-technical founders building MVPs), (2) Midjourney/DALL-E 3 for design assets (prototype screens, marketing materials, reduces designer dependency early), (3) Claude/GPT-4 for content (product copy, community management, email campaigns—but always human-edited), (4) NotebookLM for research (quickly synthesize competitor analysis, user research), (5) Descript for video content (edit demo videos, social content, marketing materials). Cost savings: a 3-person team with AI can replace a 6-7 person team from 2020 (saves ~$500K annually). Don't use AI for: brand strategy, core product decisions, user interviews, critical growth experiments. Instagram used zero AI (didn't exist) but needed 13 people pre-acquisition; today you could build Instagram MVP with 3-4 people using AI tools.

Build platform compliance into product from day one—retroactive fixes kill momentum. Critical rules: Apple takes 30% of in-app purchases (but not physical goods or external subscriptions established outside app), age-gating required for social features (COPPA compliance), content moderation systems mandatory for UGC, data privacy compliance (Apple Privacy Labels, Google Data Safety). Rejection risks: social features without moderation, cryptocurrency transactions, apps resembling existing Apple features too closely, misleading metadata. Epic Games lost its multi-year fight against Apple's 30% fee. Workarounds: (1) Web-first for payments (Spotify, Netflix model), (2) launch as 'private beta' initially (lighter review), (3) physical goods marketplace (avoid 30% fee—see Whatnot), (4) B2B wrapper for consumer tech (looser restrictions). Budget 2-4 weeks for initial approval, 1-2 weeks per update. Have lawyer review before launch—$5K legal spend saves months of rejection cycles.

2015 mechanics that no longer work: Facebook Connect auto-posting (killed 2014), contact list uploads for invites (privacy regulations restricted), viral Twitter bots (platforms cracked down), notification spam (iOS/Android restricted). 2026 mechanics that do work: (1) User-generated content with watermarks (see BeReal's authentic sharing prompt, Instagram Reels), (2) collaborative features requiring friend joins (Wordle's shareable scores, Spotify Wrapped social proof), (3) status/identity signaling (Discord's server badges, Strava's segment KOMs), (4) asymmetric value exchange (one user creates, friends consume—see TikTok duets), (5) time-limited collaborative experiences (Countdown's shared timers). Key shift: viral mechanics must be native to core experience, not bolted-on share buttons. BeReal's 'once daily' notification created authentic sharing moments; generic 'share to Twitter' converts at 0.1%. Aim for viral coefficient of 0.5+ (each user brings 0.5 new users organically).

Four validation levels before writing code: Level 1: Waitlist/landing page (48 hours to build, target 100+ emails from $500 ad spend = promising signal). Level 2: Concierge MVP (manually deliver the experience—TaskRabbit started with founders texting workers themselves, Instagram began as Burbn check-in app). Level 3: Wizard of Oz prototype (fake the AI/automation with humans behind scenes—see how people interact with 'working' product). Level 4: No-code prototype (Webflow + Airtable + Zapier can simulate most consumer apps for user testing). Red flags that kill ideas: can't get 100 waitlist signups even with paid ads (weak demand), users won't complete onboarding flow in testing (too complex), can't articulate clear 'aha moment' in 30 seconds (positioning unclear). Superhuman validated with in-person demos for 100 early users before building product—all showed 'very disappointed' if product went away (40%+ = PMF threshold).

Day-1 retention: 40%+ indicates promising onboarding; below 25% means broken first experience. Day-7: 30%+ for social apps, 20%+ for utility apps shows habit formation; Discord hit 50%, Instagram reached 45%. Day-30: 15-25% for most consumer apps, 30%+ for best-in-class. Month-6: 10%+ proves long-term value. Calculate cohort retention curves, not vanity averages—you want flat retention curves after Day-30 (indicating stable engaged base), not declining slopes (leaky bucket). Benchmark by category: social 30-50% Day-7, gaming 20-40%, productivity 15-30%, dating 20-35%. Improving retention 5% compounds massively—moving from 20% to 25% Day-7 retention means 25% more users retained weekly, enabling 25% more word-of-mouth growth. Focus here before scaling acquisition. WhatsApp didn't monetize for 5 years but maintained 90%+ retention—retention beats monetization pre-scale.

Expand only after dominating one market with network effects locked in—typically 18-24 months post-launch. Instagram went US-only for 10 months (reached 10M users), then expanded internationally. Premature expansion dilutes focus, fragments network effects, and creates unmanageable complexity (language support, payment systems, compliance, customer support across time zones). Exception: India/Southeast Asia launch can happen earlier if (1) you're targeting global diaspora communities, (2) building for emerging markets specifically, or (3) need scale to defend against incumbents quickly (see WhatsApp's rapid international rollout to stay ahead of carriers). Sequence: US/Canada first (highest ARPU), UK/Australia next (English-speaking, similar culture), then Western Europe (language complexity), finally Asia/Latin America (different mobile behaviors, lower ARPU but massive scale). TikTok succeeded globally because ByteDance had capital to run parallel market experiments—most startups lack this luxury. Stay focused.

Top 5 failure modes: (1) Running out of cash before PMF (38% of failures)—raised too little or burned too fast on paid acquisition masking organic weakness. (2) Incumbent copies and kills (22%)—Facebook/Instagram/Snap have killed dozens of features-as-startups (Stories, Live video, Disappearing messages). (3) No retention (18%)—get initial downloads but users don't stick; leaky bucket problem. (4) Can't crack a single growth channel (12%)—all channels too expensive or competitive, organic growth too slow. (5) Team breakdown (10%)—co-founder conflicts, wrong hires, remote team loses to in-person competitors. Notable: 'weak idea' is rarely the root cause—execution, timing, and market dynamics kill far more startups than bad ideas. Color Labs raised $41M with prestigious team but failed (tried to be 'mobile photo social' too late—Instagram won). Vine had 200M users but shut down (Twitter couldn't monetize, TikTok disrupted). Path had gorgeous design but failed (real graph wasn't differentiated enough from Facebook). Learn from corpses.

Three architectures work: (1) Direct network effects—product value increases with each user (messaging, social platforms). Implementation: invite gating (Clubhouse), contact syncing, active presence indicators showing 'who's online.' (2) Indirect network effects—more users attract complementary users (Uber drivers/riders, YouTube creators/viewers). Implementation: two-sided incentives, discovery algorithms favoring active suppliers. (3) Data network effects—more usage improves product (Waze traffic data, Spotify recommendations). Implementation: personalization engines, collaborative filtering, user-generated training data. Day-one tactics: (1) Make core feature multiplayer-first (Figma collaborative design, Google Docs collaborative editing), (2) show proof of network ('10,389 designers using this' social proof), (3) create artificial scarcity to drive urgency (Gmail invite-only for 3 years), (4) public artifacts that require explanation (Strava's activity maps require friend joining to understand). Don't build single-player product then add network effects—bake into core experience. WhatsApp's entire value was network—zero value alone.

Short, unique, pronounceable, .com available remains the gold standard. Winning patterns: invented words that become verbs (Google, Uber, Venmo), playful misspellings suggesting tech (Flickr, Tumblr—though less popular now), evocative words suggesting benefit (Discord, Slack, Notion), or ultra-short pronounceable combos (TikTok, BeReal). Failed patterns: generic descriptive names (Ello, Vero—tell me nothing), difficult spelling/pronunciation (Mastodon struggles vs Twitter), required explanation (Google Wave—what's a wave?), or awkward domain hacks. Name budget: $5-20K for premium .com is worth it—short-term pain, long-term gain. Instagram bought their domain for $15K (was burbn.com initially). Snap bought snap.com years after launch for $15M (was snapchat.com). Get it right day one. Test: If you have to spell it when saying it aloud, it's wrong. If people misspell when searching, it's wrong. If it doesn't work as a verb ('I'll Discord you'), reconsider. TikTok won the naming game—short, fun, memorable, suggests the product.

Budget 10-15% of engineering resources for trust & safety from launch—not after first crisis. Minimum viable system: (1) user reporting (flag button on all content), (2) keyword filtering (block obvious bad content), (3) phone/email verification (reduce spam accounts), (4) rate limiting (prevent bot abuse). As you scale: hire dedicated trust & safety lead at 100K users, implement ML-based content moderation at 500K users, contract human review team at 1M+ users. Cost reality: $0.50-2.00 per user annually for moderation at scale. Facebook employs 15K+ content moderators; TikTok has 10K+. Legal requirement: DMCA compliance, COPPA if under-13, CDA Section 230 protections (US only). Emerging concern: deepfakes, AI-generated spam, coordinated inauthentic behavior. Common mistake: waiting until media/PR crisis to build systems—damages brand permanently. Discord built sophisticated moderation early and avoided many issues; Twitter's delay in tackling abuse hurt growth for years. Don't skip this.

Take top-tier consumer-specialized investors regardless of size—Benchmark, Greylock, and NFX have better consumer track records than generalist mega-funds. Why? (1) Pattern recognition from 20+ years of consumer investing, (2) deep networks in consumer talent (design, growth, product), (3) hands-on help with psychological nuance and retention, (4) consumer brand building (critical for this category). a16z and Sequoia work if you get partner who led consumer wins (Jeff Jordan-Instacart, Roelof Botha-Instagram). Avoid: enterprise-focused funds doing 'tourist' consumer deals, funds with no consumer unicorns in 10+ years, or solo GPs without consumer expertise. Check: ask VCs to connect you with consumer founders they funded who failed—tells you how they behave when things go wrong (most important signal). Insider move: take $500K-1M from top consumer angels (ex-Meta, ex-Snap execs) over $2M from generic fund—advisory value exceeds capital difference early. Benchmark's $6M seed in Instagram (5% stake) returned $78M—expertise matters more than check size.

Consumer startups need growth PM as first PM hire, not traditional product manager. Difference: traditional PM focuses on features/roadmap (appropriate for enterprise where sales drives growth), growth PM focuses on activation, retention, virality, resurrection (critical for consumer where product IS distribution). Growth PM skills: SQL/analytics fluency, A/B testing discipline (running 10-20+ tests monthly), channel expertise (paid, viral, SEO), behavioral psychology understanding, metric obsession. Hire when: you have product-market fit signals (retention working) but growth stalling. Too early (pre-PMF) wastes money on growth when retention is broken; too late (post-inflection) means leaving 10x growth on table. Facebook's Chamath Palihapitiya growth team created '7 friends in 10 days' activation metric (key to hypergrowth). Pinterest's Casey Winters drove SEO-powered growth. Reforge runs best training for growth PMs. Budget: $150-200K total comp for senior growth PM in SF. Red flag: PM with only B2B SaaS experience won't understand consumer psychology and viral mechanics—different disciplines entirely.

You can't eliminate platform risk, but you can diversify surface area: (1) Build web version with feature parity (Discord's web app rivals mobile), (2) establish direct user communication (email list, SMS, owned community), (3) enable account export/portability (builds trust, aids viral growth), (4) avoid features that violate platform ToS (cryptocurrency transactions, payment circumvention). Historical examples: Fortnite banned from iOS for 18 months (Epic-Apple fight), Parler removed from app stores (content moderation failure), apps using X-Mode SDK banned (privacy violation). Emerging risk: Apple/Google copying successful consumer apps as native features (Apple Maps killed Mapquest, Apple Podcasts hurt dedicated apps, Screen Time copied Freedom/Moment). Defense: build network effects and brand that platform can't replicate—Apple couldn't kill Instagram by adding photo filters because network was moat. Accept reality: iOS/Android distribution is non-negotiable for consumer mobile—build accordingly. Progressive Web Apps (PWAs) remain too limited for full consumer experience in 2026.

Show cohort retention curves, viral coefficient, and unit economics—not vanity metrics. Must-include: (1) Monthly active users (MAU) and growth rate, (2) Day-1/7/30 retention by cohort with curves, (3) Viral coefficient or K-factor (each user brings X new users), (4) Time spent per user per session/week, (5) Customer acquisition cost (CAC) by channel, (6) Monetization per user if applicable (ARPU). Investor red flags: only showing total downloads (ignores retention), cumulative user graphs (hides declining growth), vanity engagement (likes, shares without retention context), incomplete cohorts (only showing successful months). Best-in-class: show retention curves that flatten after Day-30 (proves sustainable engagement), organic growth rate of 15%+ monthly (proves word-of-mouth), CAC payback under 12 months if monetizing. Instagram's seed deck showed hockey-stick MAU growth and 45% Day-7 retention—Baseline invested $500K for 9.2% (worth $78M at acquisition). Discord showed 50%+ Day-7 retention despite zero paid acquisition. These metrics differentiate signal from noise. Use Mixpanel or Amplitude to generate these—spreadsheets look amateur.

Ship fast with polish on core experience, fast and scrappy everywhere else. The core loop (the 30-60 second experience users repeat) must be polished—this is your retention driver. Instagram's photo filters and feed were beautiful; settings and profile editing were basic. Everything else can be fast/scrappy: onboarding (iterate based on data), secondary features (ship to test engagement), admin/settings (functional is fine), marketing pages (templates work early). Specific benchmarks: core experience should feel 8/10 quality, everything else can be 6/10 initially. Ship new test features in 3-5 days, not 3-5 weeks—if concept doesn't work, quality doesn't matter. Polish comes after validation. Common mistake: perfectionism on features users won't use (beautiful settings screens, overdesigned onboarding for 80% drop-off flow). Snapchat's early app was buggy but the core ephemeral messaging was magical—that mattered. Clubhouse's early app crashed constantly but the live audio rooms were compelling. Fix the crashes (quality of service), don't over-polish unused features. AI coding tools (Cursor, Windsurf) now enable both speed AND quality—use them.

Y Combinator remains gold standard for consumer despite enterprise focus shift—consumer companies still 20-25% of batch, includes Discord, Instacart, DoorDash alumni. South Park Commons in SF works for technical consumer founders (more exclusive, $1M standard investment). On Deck has community but weaker for consumer specifically (better for SaaS). Avoid: generic startup accelerators without consumer expertise or track record. Communities that work: Reforge for growth education (Casey Winters, Elena Verna), Pavilion for GTM but B2B-focused, First Round's network if you're portfolio company. Underground: ex-Meta/ex-Snap/ex-TikTok PM groups (private Slacks, WhatsApp groups—need intro to access). Best for geographic hubs: SF Consumer Founders meetup, LA tech scene for creators/entertainment, Miami for crypto-consumer crossover. Apply to YC if under 30 and first-time founder (acceptance rate 1.5%, but best brand/network). Skip accelerators entirely if you're experienced founder—raise directly from seed VCs (faster, better terms, no dilution). Benchmark, Greylock, NFX don't require accelerator pedigree.

Don't compete on general intelligence—compete on context and workflow integration. ChatGPT is general-purpose; your product should be purpose-built for specific use case with 10x better experience. Winning strategies: (1) Domain-specific models: Harvey for legal, Cursor for coding (both outperform ChatGPT for specialized tasks), (2) Workflow embedding: Notion AI inside docs, Superhuman AI inside email (context advantages), (3) Social/network layer: Character.AI's companionship, Poe's model marketplace (ChatGPT is single-player), (4) Privacy/ownership: users concerned about OpenAI training on their data—offer local/private alternatives. Failed approach: building ChatGPT wrappers with custom prompts—zero defensibility, OpenAI will add your feature in their next update. Remember: Google was dominant in search, but Yelp won local search by focusing on reviews + social proof. OpenAI is dominant in general AI, but vertical-specific AI startups can win by being 10x better for specific jobs-to-be-done. Perplexity succeeded by focusing on AI search with citations—different use case than ChatGPT chat.

iOS engineers for US/Western launch, but hire full-stack with mobile capabilities rather than pure mobile specialists early. Modern reality: React Native, Flutter, and Swift/SwiftUI enable one engineer to build both platforms—hire someone with this range. If forced to choose: iOS-native engineers command higher salaries ($180-220K in SF) but deliver better performance/polish for US market (where 60% of revenue comes from iOS despite 40% share). Android-native engineers cost less ($150-190K) and critical for India/emerging markets (80%+ Android share). Best early hire: full-stack engineer who can build mobile (React Native) + backend (Node/Python) + basic web—maximizes iteration speed with 1-2 person team. Instagram's first engineer (Mike Krieger) built iOS app + backend + scaling infrastructure. Avoid: hiring separate iOS and Android engineers pre-PMF—doubles team size, halves iteration speed, splits focus. Use cross-platform tools until you have strong PMF and can afford native teams (typically $5M+ raised, 1M+ users). Exceptions: apps requiring native performance (AR, gaming, video editing) or platform-specific features (HealthKit, iMessage extensions) need iOS-native from day one.