How These College Students Rebuilt Their App from $2K to $30K MRR
TL;DR: Aayush and Yali spent 5 months building PropGPT first version during NFL season. Launched with influencer marketing driving 20 downloads/day, 5-10 trial conversions. Stuck at $1-2K MRR. Analytics revealed problem: 45% trial conversion but only 13% trial-to-paid. Users loved the IDEA but hated using the app. Product was DIY analytics tool requiring users to input bets and check themselves. Realized users wanted 'right answers to the test' not homework. Shut down all marketing, spent 4 months rebuilding from scratch. New version: presents pre-analyzed bets users can review, not DIY tool. Relaunched April 15 at $1.7K MRR, 15 downloads/day. Went all-in on marketing for NBA playoffs. Conversion jumped to 50%+. Hit $40K MRR peak 2.5 months later with 2K downloads/day. Now: $30K MRR, 40K+ downloads, 3K+ paying customers, 48% trial conversion rate, $3.30 revenue per download. Used PostHog, Superwall, RevenueCat analytics to identify conversion funnel issues. 70th organic video hit 600K views, grew ARR from $8K to $38K in 3 days showing viral power. Tech: React Native, TypeScript, Python ML algorithms, Neon database, RevenueCat, Superwall. Costs: 20¢/conversion Superwall, 1% RevenueCat, $100/mo data APIs, $10/mo database, $20/mo LLM, $10K/mo marketing = ~50% margins.
Key Insights
- Distribution doesn't matter if product is bad - had downloads but couldn't retain users
- Analytics reveal truth: high trial conversion + low paid conversion = users love idea but hate execution
- Extreme product humility required - 'if your product is bad, nobody will buy it'
- Users wanted answers handed to them, not DIY analytics tools - simplified to pre-analyzed picks
- One viral video (600K views) can 5x ARR in 3 days when product-market fit is right
Actionable Takeaways
- Track trial-to-paid conversion separately from download-to-trial - reveals product quality
- Use PostHog, Superwall analytics to see exactly where users drop off in onboarding
- If users convert to trial but don't pay, they like the concept but not the execution
- Shut down marketing when product is wrong - spending on bad product wastes money and misleads data
- Track feature clicks to identify most compelling value prop, then emphasize in marketing