Fitness coaching SaaS engineered for 12× performance at 200+ paying users

The situation

• The founder had spent 8 months building a coaching platform on Lovable. It had real users (200 paying coaches), real revenue ($4.2K MRR), and a real product. Then growth hit.
• Database queries that took 40ms with 10 users were taking 28 seconds with 200. Real-time session tracking, the core feature, crashed under load. The worst part: Stripe webhooks were failing silently.
• Coaches were completing sessions but not receiving payment notifications. The founder discovered this only when a top coach threatened to leave. The options felt impossible: throw away 8 months of work and rebuild from scratch, or keep shipping on a broken foundation.

What we did

• Properly indexed Postgres schema with connection pooling via PgBouncer
• Supabase Realtime subscriptions replacing broken polling, 98% reduction in server load
• Stripe webhook handler with idempotency keys, signature verification, retry queue
• Redis caching layer for dashboard aggregations (Upstash serverless)
• Sentry error monitoring, 47 previously-silent errors found and fixed
• Full load testing suite, verified to 1,000 concurrent users

The result

• The platform now serves 200+ coaches without breaking a sweat. The founder reported zero churn in the 30 days following launch, compared to 3 cancellations in the week before we started.
• The Stripe fix recovered $1,240 in missed payment notifications from the previous month. Three months after launch, MRR had grown to $11.8K.