An audit trail is a high-consequence architectural concern when failures could silently affect money, compliance, data integrity, access, or trust. It does not make every surrounding feature equally rigorous: rigor should follow consequence, reversibility, detectability, and blast radius. section 3.5
Treat audit-trail decisions with stronger rigor when:
- Failure can remain hidden or silently corrupt data, calculations, permissions, or business outcomes. section 3.5
- Records support financial reconciliation, regulatory compliance, access control, or investigation of consequential changes. section 3.5
- The decision becomes difficult to reverse after users, data, integrations, or operational processes depend on it. section 3.5
Table 3.5 Decision rigor rubric
|
Factor
|
Lower rigor
may be acceptable when...
|
Higher
rigor is justified when...
|
|
Reversibility
|
The
decision can be changed, rolled back, or replaced without major disruption.
|
The
decision is hard to reverse once data, users, integrations, or operational
processes depend on it.
|
|
Detectability
|
Failure is
visible quickly through tests, monitoring, user feedback, or operational
signals.
|
Failure can
remain hidden, appear later, or silently corrupt data, calculations,
permissions, or business outcomes.
|
|
Blast
radius
|
Failure
affects a small internal group, a temporary workflow, or a non-critical
capability.
|
Failure
affects customers, money, safety, compliance, sensitive data, or core
operations.
|
|
Cost of
failure
|
The
consequence is inconvenience, rework, delay, or local operational pain.
|
The
consequence is legal exposure, financial loss, security breach, regulatory
failure, physical harm, or major reputational damage.
|
|
Recovery
|
The team
can recover through rollback, retry, correction, or a known manual process.
|
Recovery
requires coordinated incident response, data repair, customer communication,
regulatory reporting, or complex restoration.
|
Compare audit-trail rigor using reversibility, detectability, blast radius, and cost of failure rather than applying one standard to the entire system. section 3.5




