Skip to main content

Operational Resilience

Operational resilience: designing for the bad week

A system is only as dependable as its behaviour on the day something upstream breaks.

Synexum Labs EditorialDecember 15, 20256 min read

Name the failure modes

For each source, decide what happens when it is late, incomplete or unavailable. Displaying a stale-data state with its timestamp is a design decision; showing yesterday's figure as if it were today's is a defect waiting for an audit.

Degrade deliberately

Partial function is usually better than none. A workflow that can still queue, record and route work while an enrichment service is down keeps the operation moving and preserves the evidence trail for later reconciliation.

Rehearse recovery

Restore procedures that have never been executed are hypotheses. Testing them on a schedule, with the people who would actually run them, is the difference between a recovery plan and a recovery document.