Comparison

Purpose

Comparison produces canonical structured evidence comparing baseline replay behavior against candidate simulation behavior.

Use Comparison when customer systems need deterministic, recordable evidence about whether candidate policy setup materially changes runtime outcomes.

When To Use It

Use Comparison when:

A valid replay_reference is available.
Candidate policy setup is available.
Customer systems need structured evidence for review, approval, release, experiment, rollback, or audit workflows.
Customer correlation metadata should be returned unchanged for customer-owned traceability.

Request Boundary

Comparison requires:

replay_reference.
candidate_policy_setup.

Comparison may include optional customer_correlation_metadata, such as:

experiment_id.
corpus_id.
customer_case_id.
Customer-defined trace or release identifiers.

FoxCommand returns supported customer correlation metadata unchanged and does not interpret it as governance lifecycle state.

Response Evidence

A successful Comparison response may include:

Runtime response API version.
status.
schema_version, currently comparison.v1.
comparison_id.
replay_reference.
baseline_replay.
candidate_simulation.
comparison_result.
customer_correlation_metadata.
Bounded errors.

Replay Record remains the atomic evidence object. Comparison evidence references the baseline replay through baseline_replay.evidence_object: "replay_record" and baseline_replay.replay_reference.

Deterministic Evidence

Successful Comparison responses include a deterministic comparison_id.

Customer correlation metadata does not determine the comparison identifier. The identifier is based on deterministic runtime comparison material, candidate policy setup, baseline outcome, simulated outcome, and comparison result.

Customer Responsibilities

Customers should:

Store comparison_id, schema_version, comparison_result, replay_reference, and customer correlation metadata.
Preserve the relationship between comparison evidence and customer-owned governance, policy, corpus, experiment, case, review, approval, release, or rollback records.
Decide how structured evidence affects customer lifecycle decisions.

Boundary

Comparison is not batch execution, experiment management, corpus storage, aggregate reporting, dashboarding, CI/CD integration, promotion workflow, or rollback orchestration.

Next Step

Use Governance RCA when a material difference needs bounded attribution.