When people hear "185 database tables," they assume complexity for complexity's sake. But every table exists because a business requirement demanded it.
Here's how I designed the Nexural schema — the decisions that worked, the ones I'd change, and the patterns that scale.
Phase-Based Schema Design
I didn't design 185 tables on day one. The schema grew across 7 phases, each adding a domain:
| Phase | Domain | Tables | Key Decision |
|---|---|---|---|
| 1 | Auth & Users | 12 | Supabase Auth + custom profiles |
| 2 | Subscriptions | 8 | Stripe webhook-driven state machine |
| 3 | Trading | 35 | Instruments, positions, signals, watchlists |
| 4 | Community | 25 | Discord sync, moderation logs, reputation |
| 5 | Analytics | 30 | Metrics, reports, telemetry events |
| 6 | Research | 40 | Strategies, indicators, backtest results |
| 7 | Operations | 35 | Alerts, newsletters, audit logs |
Each phase had its own migration batch. I never modified tables from a previous phase during a new phase's development. This kept deployments safe.
The Three Rules I Followed
Rule 1: Normalize Everything Except Hot Paths
The canonical data is always normalized. \
