Reliability & continuity
Last updated: 7 August 2026.
Research has deadlines. This page describes how we keep SortedResearch available and how we recover if something goes wrong.
Uptime commitment
- 99.9% monthly uptime commitment for Team and Enterprise plans, with a service-credit ladder (10% / 25% / 50%). See the SLA for the formula, exclusions, and claim process.
- Free and Individual plans receive the same infrastructure on a best-effort basis.
Live transparency
- Status page: public, live health checks of the website, database, AI assistant, and email delivery on every view.
- Machine-readable health endpoint for programmatic monitoring.
- A weekly automated canary exercises the full stack, including AI features, and reports results.
Backups: hourly, encrypted, and vendor-independent
- The full database (schema, data, roles) is exported every hour and stored compressed and encrypted in a private bucket with a different provider than our database vendor. If our database vendor's account itself ever had a problem, the backups would still exist.
- 30-day backup retention.
- A restore-verification script loads backups into a scratch database and confirms integrity.
Recovery targets
| Target | Value | Meaning |
|---|---|---|
| RPO (Recovery Point Objective) | ≤ 1 hour | At most one hour of data would be lost in a worst-case restore |
| RTO (Recovery Time Objective) | ~1–3 hours | Expected time to restore service after a catastrophic failure |
Our disaster recovery runbook is documented and maintained alongside the codebase.
Graceful degradation
The platform is designed so that a single dependency failing narrows the product rather than taking it down:
- If the AI provider is unavailable, studies keep collecting responses and results remain accessible. Only Graham's assistance pauses.
- If the email provider is unavailable, invites degrade to copyable links.
- Participant data collection has no direct dependency on third-party scripts.
Support response targets
First-response targets by plan (from the SLA): Individual ~2 business days, Team ~1 business day, Enterprise ~4 business hours for urgent issues.