SYSTEM_CONSOLE v2.4.0

Visualization

LAST_UPDATED: 2025-10

Visualization is the interface between the data platform and the business. The core problem it solves is metric consistency: when Finance and Marketing report different revenue numbers from the same source, the issue is almost always logic embedded in separate dashboards rather than in a shared semantic layer.

Key Takeaways

  • 01 Define metrics once in a semantic layer.
  • 02 Version control your metric definitions.
  • 03 Surface data freshness directly in the BI tool.
  • 04 Enforce row/column security at the source where possible.

Checklist

  • KPI definitions documented and approved by business owners.
  • Dashboard freshness indicators present and accurate.
  • Access control groups mapped to domain products.
  • Performance tuning (caching/aggregates) configured.

Semantic layer principles

The semantic layer acts as a translator between complex database structures and business terms. It prevents "logic leakage" where different dashboards calculate the same KPI (e.g., Gross Margin) in different ways.

Centralized Metrics

Define 'Total Revenue' once in the semantic model. If the logic changes, update it in one place and all dashboards reflect it.

Simplified Access

Users see 'Customer Name' and 'Total Spend' without needing to know about the 5-way join happening in the background.

Performance

Caching Enable query caching for high-traffic executive dashboards.
Aggregates Use pre-computed aggregate tables in the Gold layer for "big picture" queries.
Partition Strategies Ensure BI queries always use partition filters (e.g., date ranges) to minimize warehouse costs.

Data freshness surfaced in BI

Trust is destroyed when users make decisions based on stale data. Every dashboard should explicitly show when its underlying data was last refreshed.

Dashboard Status
Data Freshness: 4m ago
Source: sales.orders_gold
GCP mapping
Looker (Semantic Layer & BI), BigQuery BI Engine (Performance), Looker Studio (Ad-hoc Visualization).

Failure modes

  • ! Metric Fragmentation: Marketing and Finance report different "Revenue" numbers because they use different SQL logic in their respective tools.
  • ! Dashboard Bloat: Hundreds of unused dashboards make it impossible for users to find the "Single Source of Truth."
  • ! Cost Spikes: Ungoverned ad-hoc queries on massive raw tables causing unexpected cloud bills.