Routine database maintenance. One table. Took 90 seconds. But it locked the entire system. Every customer, from the smallest startup to the Fortune 500 account, sat frozen for those 90 seconds. Multi-tenancy became personal that day. It's not a technical detail. It's a strategic decision that either multiplies growth or multiplies risk.
Why this matters
Investors evaluate whether companies can "add customers without linearly adding cost and risk." The fundamental question: If your largest customer runs a heavy batch job tonight, what happens to your smallest customer's response times?
Three isolation models
- Shared Database/Schema: Fast and cost-effective, but wide blast radius. Enterprise customers typically reject it for compliance reasons.
- Shared Infrastructure, Separate Schemas/Databases: Balanced approach offering stronger isolation with reasonable operational overhead.
- Full Silo Deployment: Maximum isolation and customisation, but multiplies operational complexity rapidly.
Critical evaluation areas
- Blast Radius Controls: Rate limiting, query timeouts, and resource quotas must be architectural features, not afterthoughts. Noisy neighbour problems are architectural, not operational.
- Data Isolation: Physical boundaries matter more than logical filters. Row-level security, scoped tokens, and encryption keys signal enterprise readiness.
- Operational Maturity: Single-tenant restore capabilities, schema migration procedures, and tenant-independent onboarding processes demonstrate architectural soundness.
Stage-specific expectations
- Seed/Early A: Demonstrate awareness of trade-offs
- Series B/Growth: Evidence of blast radius controls and documented isolation strategy
- Acquisition/Control: Detailed technical validation including query audits and restore testing
Red flags for dealmakers
- Application-only isolation enforcement
- Absence of per-tenant rate limiting
- Shared downtime requirements for all tenants
- Reliance on history rather than technical controls
Multiple concerns trigger pricing adjustments or post-close remediation requirements.
Your turn
What noisy neighbour incident bit you hardest? A runaway query from one customer taking down the whole system? A maintenance window that couldn't account for tenant-specific configurations? Share the scar.
Originally published on the Tech Due Diligence Playbook newsletter on LinkedIn.