The distribution of advertising click-through rates differs greatly from last month.
Detect drift that a NULL-rate check alone would miss.
Which quality approach is most appropriate?
Compare distribution statistics, category ratios, and percentiles with a reference period while considering seasonality.
Detailed explanation
Removing the baseline prevents drift measurement.
Removing the baseline prevents drift measurement.
A baseline comparison can reveal value changes even when the schema is unchanged.
A baseline comparison can reveal value changes even when the schema is unchanged.
AWS API events do not represent the business-value distribution.
AWS API events do not represent the business-value distribution.
Columns and types can stay the same while values change meaning or distribution.
Columns and types can stay the same while values change meaning or distribution.
Check official documentation
Compare the following points with the official documentation.
AWS Glue Data Quality profiling、CloudWatch anomaly detection、data driftの公式説明を確認する。
What to confirm
スキーマドリフトと値分布ドリフトの違いを説明できる。
Key points
- Data drift
- Distribution
- Percentile
- Reference period
Notes
- Environment: AWS公式DEA-C01試験ガイドと各サービスの公式ドキュメントの確認
- Confirm current AWS service behavior in the official DEA-C01 documentation.
Foundation review
Seasonality
Normal periodic variation that should not be confused with an incident.
Response
Correlate drift with releases, tracking tags, regions, and upstream changes.