Your entire platform runs in one Supabase region in Tokyo, on Vercel, by a tiny team. That's concentration risk stacked on concentration risk. If any layer fails, my board loses its decision platform. Why is that acceptable?

The full executive answer

It's acceptable because of what the product is and what stage we're at — and I'll give you the recovery numbers rather than adjectives. First, blast-radius framing: Vouli IQ is a decision-support and diagnostics platform, not a transaction system in your operational path. If we're dark for four hours, your board is inconvenienced, not halted — no payments stop, no operations freeze. That materially changes the availability economics versus, say, your core banking vendor, and it's why we've prioritised confidentiality and integrity engineering over multi-region availability engineering so far.

Second, the actual continuity posture: the database layer has automated backups with point-in-time recovery on AWS infrastructure; application deploys are immutable and content-addressed, so any bad release rolls back in minutes; the codebase, configuration, and infrastructure definitions are fully version-controlled, meaning the platform can be re-provisioned from scratch onto a fresh cell — the EU and India region cells are already coded and config-ready precisely because the architecture treats a cell as reproducible, not hand-crafted. A public status page gives you outage transparency without waiting for an account manager. On provider failure: Supabase-on-AWS and Vercel are among the most operationally mature platforms available, and inheriting their SRE teams is — at our stage — a stronger availability control than anything we could self-host.

The honest gaps with dates: there is no live multi-region failover today — a full Tokyo-region AWS outage means downtime for its duration, and I won't quote a fantasy RTO for that scenario; cross-region warm standby is roadmapped with the first enterprise contract that requires a formal RTO/RPO commitment, because provisioning it has real cost that should follow real demand. Documented and rehearsed disaster-recovery runbooks (as opposed to recoverable-by-design architecture) are part of this year's SOC 2 readiness scope. And key-person risk is real — mitigated by the extensively self-documenting codebase and CI safety net, resolved by hiring, not by promises.