Real-Time Analytics Platform · View 09 of 21 · Data

Analytics Data Model

The tables the hot store actually holds, their keys and cardinality — including the dead-letter and schema-contract tables that most models omit.

Editable source SVG draw.io All views

Modelling decisions

Raw payload kept as a dynamic column so a schema change never blocks ingestion
Both event_time and ingest_time are stored; every window is computed on event_time
Dimensions are slowly changing type 2, joined on the surrogate key valid at event_time

Pre-aggregation

mv_metric_1m is the query target for standard dashboards; raw scans are the exception
Top-N views are derived from the 1-minute view rather than recomputed from raw
A dashboard query that misses every view is a design signal, and is alerted on

Assumptions

Cardinality of tenant × metric × dimension stays below roughly 2 million active series
User identifiers arrive already hashed; the platform stores no clear personal identifier
Ad-hoc dimensions are attributes of the enriched table, not new materialized views

◀ Storage Zones, Tiering and Retention All views Schema Contract Governance ▶