When people hear I rebuilt my Saju SaaS product four times in six months, they usually laugh. Some think I'm bad at planning. The truth is messier and, I think, more useful to share: I was right about the problem, completely wrong about the solution, and learned that rebuilding isn't always a failure—sometimes it's the fastest path to product-market fit.
Why I Built It: The Problem Was Real
I grew up in Seoul with a grandmother who consulted her Saju reading before every major life decision. My mom did the same. When I moved to San Francisco in 2024, I found Korean diaspora communities—in Discord servers, on Reddit, in WeChat groups—desperately searching for authentic Saju readings online. Most paid $50 to $200 for a 30-minute consultation with a practitioner. It was expensive, inconvenient, and the quality varied wildly depending on who you found.
The core insight was undeniable: digitize Saju interpretation for global Korean communities. The market existed. The demand was real. But I didn't understand what "digitize Saju" actually meant, and that confusion drove every subsequent rebuild. I'd identified the right problem—access to quality Saju interpretations at scale—but I had no idea what the solution should look like.
Rebuild One: The Astrology App Clone (Month 1)
My first instinct was to treat Saju like Western astrology. I copied the structure of popular apps like Costar and Co-Star—slick mobile interfaces that make astrology feel personalized, interactive, and shareable. I built a React Native app with a Firebase backend, integrated GPT-4 for reading generation, and bundled the real Saju calculation logic (the heavenly stems and earthly branches) into the experience.
The launch was fast: 340 signups in the first week. 12% weekly active rate. It looked genuinely promising until I spent time observing actual user behavior. The reality was brutal: 89% of my users were non-Korean. They had zero cultural context for Saju. They downloaded it, read something mystical-sounding, and never returned. The actual Korean users—just 3% of my cohort—were frustrated because the readings lacked any real depth or specificity. They wanted legitimate interpretation, not gamified astrology.
I'd optimized for user acquisition instead of user retention. I'd built something that went viral for the wrong audience entirely.
Rebuild Two: The Consultant Marketplace (Month 2-3)
The obvious pivot: connect users with actual Saju practitioners. I recruited five experienced practitioners from Seoul, built a booking and payment system, and launched as a two-sided marketplace with a 30% commission split.
This version gained traction faster. Month two showed 240 interested practitioners and 450 active users. Transaction volume: $3,200 in bookings, generating $960 in gross profit. For a few weeks, this felt like the right direction.
But the friction became unsustainable. Practitioners wanted to vet every client before accepting appointments. Users wanted instant answers, not calendar syncing delays. Payment processing required compliance in multiple regions (US, South Korea, Canada). Most critically, I was drowning in customer support—15 hours a week—mostly explaining who these practitioners were and justifying consultation fees. The marketplace model created a support burden that killed velocity.
The deeper problem: I was building infrastructure for a marketplace that didn't actually exist yet. Neither supply nor demand was strong enough. I was applying venture-scale thinking to a niche product, and it showed immediately.
Rebuild Three: The Structured Reading Product (Month 3-4)
I stripped everything down. No marketplace. No practitioners. Just a clean, structured Saju reading platform.
Users input birth date, time, and location. A backend service calculates their exact Saju chart (the four pillars: year, month, day, hour). Rules-based logic generates detailed interpretations across seven categories: career trajectory, relationships, health, fortune, core personality, challenges, and lucky elements. I added a premium tier at $4.99/month for unlimited readings, personalized monthly updates, and a yearly forecast report.
Numbers: 620 new users in month one. 28% converted to premium. $850 MRR. Session length was 6 minutes on average—solid engagement compared to 90 seconds in the app clone. But something was still missing: users didn't return. The reading felt thorough but final. They'd read it, get their answer, and disappear. The product was useful exactly once. Not useful repeatedly.
Rebuild Four: Building for Longevity (Month 5-6)
The key insight came in June: Saju readings aren't a one-time transaction. For Korean users especially, they're a practice—something you return to, reference, update as your year unfolds. Users wanted to come back periodically, compare their reading with partners or family members, understand timing and cycles.
I rebuilt around that reality. Daily Fortune: a 2-3 minute reading tied to the lunar calendar, giving users a reason to return every morning. Compatibility Readings: compare your chart with up to five people—spouse, family, business partners. Yearly Planning: a Saju-based roadmap for the next 12 months. A moderated Community feed where users share readings and insights.
I raised pricing to $6.99/month for premium. I launched a one-time $19.99 "Full Analysis" report—a beautiful 30-page PDF with historical context, deeper interpretation, and personal recommendations.
Results: Monthly active rate jumped to 32% (from 12%). Premium conversion: 41%. MRR: $3,400. The Full Analysis report saw 18% attach rate among paying users, generating $2,100 in first-month revenue. Churn dropped from 18% to 8%. Return users increased 4x.
What Four Rebuilds Taught Me
First: Don't optimize prematurely. I could have spent two months adding features, refining the marketplace, hiring support staff. Instead, three weeks of "wrong" product taught me more than three months of careful execution would have.
Second: Iterate alongside users. After rebuild one, I conducted 20 user interviews. After rebuild two, 30 more. The pattern only became clear through repeated observation. By rebuild four, I knew exactly what users needed because I'd watched them fail to find it in three previous iterations.
Third: Niche products grow differently. I was applying B2C consumer app playbooks to a niche product. The math is different. 1,000 deeply engaged users beats 50,000 churned ones. Niche products succeed through depth and retention, not breadth.
Fourth: Your first product hypothesis will be wrong. The sooner you accept this, the faster you ship. I rebuilt four times not because I failed—I succeeded at learning. The only cost was time and velocity.
The four rebuilds took six months and $2,400 in infrastructure costs. No hiring, no office, no runway burn. If you're building in a niche right now, ship something small, watch it fail, rebuild. Don't wait for perfect—perfect doesn't exist until your users help you define it. I'm keeping a public build log at Saju App documenting what's working, what isn't, and why. Come share what you're learning in your own niche.
Top comments (0)