Entityresolution Diagnostics

AWS Entity Resolution Diagnostics

When to use

Any AWS Entity Resolution investigation — schema mapping, matching workflows, ID mapping, provider services, output configuration, or matching algorithm issues.

Investigation workflow

Step 1 — Collect and triage

aws entityresolution list-schema-mappings
aws entityresolution list-matching-workflows
aws entityresolution list-id-mapping-workflows
aws entityresolution list-provider-services

Step 2 — Domain deep dive

aws entityresolution get-schema-mapping --schema-name <name>
aws entityresolution get-matching-workflow --workflow-name <name>
aws entityresolution get-id-mapping-workflow --workflow-name <name>
aws entityresolution list-matching-jobs --workflow-name <name>

Step 3 — Detailed investigation

aws entityresolution get-matching-job --workflow-name <name> --job-id <id>
aws entityresolution get-id-mapping-job --workflow-name <name> --job-id <id>
aws cloudtrail lookup-events --lookup-attributes AttributeKey=EventSource,AttributeValue=entityresolution.amazonaws.com --max-results 20

Read references/guardrails.md before concluding on any Entity Resolution issue.

Tool quick reference

Gotchas: AWS Entity Resolution

• Schema mappings define how input data columns map to Entity Resolution attributes. They reference Glue Data Catalog tables.
• Matching workflows use either rule-based or ML-based matching. Rule-based requires explicit match rules; ML-based uses AWS-trained models.
• ID mapping workflows map identities across datasets using provider services (e.g., LiveRamp, TransUnion).
• Provider services are third-party integrations that require separate subscriptions and configuration.
• Output goes to S3. The IAM role must have write access to the output location.
• Entity Resolution does NOT modify source data. It produces match results as output files.

Anti-hallucination rules

1. Always cite specific workflow names, job IDs, or API responses as evidence.
2. Entity Resolution is NOT Glue ETL. They are separate services.
3. Schema mappings reference Glue Data Catalog, NOT raw S3 files directly.
4. ML matching is NOT customizable — it uses AWS pre-trained models.
5. Provider services require third-party subscriptions.
6. Spend no more than 2 minutes on any single hypothesis. Pivot if inconclusive.

Runbooks