Medallion · Diagram Design

Five-tier e-commerce analytics storage

A quality ladder for clickstream events and order records, from immutable landing data to dashboard-ready metrics and cold retention.

Five-tier e-commerce analytics medallion Clickstream events and order records move from raw storage through anonymized, staging, and aggregated tiers before lifecycle archiving. MASK IDS CLEAN + JOIN AGGREGATE LIFECYCLE Raw raw-commerce Tool
NiFi ingest
 Format
JSON · Parquet
 Writer
Data Engineer
 E-COMMERCE EXAMPLE clickstream events order records Anonymized anon-commerce Tool
Trino SQL
 Format
Iceberg · partitioned
 Writer
Data Engineer
 E-COMMERCE EXAMPLE session pseudonyms masked customer IDs Staging staging-commerce Tool
Trino · notebooks
 Format
Iceberg · cleaned
 Writer
Data Scientist
 E-COMMERCE EXAMPLE joined order facts validated event rows Aggregated analytics-marts Tool
Trino INSERT
 Format
Iceberg · metrics
 Writer
Data Scientist
 E-COMMERCE EXAMPLE daily conversion rate revenue by channel Archive archive-commerce Tool
Object storage lifecycle
 Format
Cold tier · immutable
 Writer
Data Engineer
 E-COMMERCE EXAMPLE retained clickstream closed order snapshots SQL PATH Trino INSERT INTO … SELECT filter · join · aggregate — repeatable transforms NOTEBOOK PATH Python notebooks on object storage explore · validate · model — interactive analysis

THE HEADLINE

Aggregated is the query surface

Dashboards read stable metrics instead of recomputing joins across raw commerce records.

Quality rises by tier

  • Identifiers are masked first
  • Events and orders are validated
  • Metrics are versioned for reuse

Two write modes

Trino handles repeatable promotion jobs; notebooks support exploration before logic is operationalized.