Business Intelligence and Dashboards

The Business Intelligence pillar of the MATIH Platform provides a complete dashboard and reporting system that integrates deeply with the conversational analytics layer and the data governance framework. Unlike standalone BI tools, MATIH's BI capabilities are built on top of the semantic layer, ensuring that metrics are consistent across every dashboard, every query, and every conversational response.

The bi-service (Java, port 8084) manages the complete lifecycle of dashboards and reports:

Widget Types

The BI Workbench supports a rich set of widget types, each configurable through a visual editor:

Cross-Filtering

Dashboards support cross-filtering: clicking a data point in one widget filters all other widgets in the same dashboard. This interaction model enables exploratory analysis without writing queries:

Click "Electronics" in Category Bar Chart
  -> Revenue Line Chart filters to Electronics only
  -> Regional Map highlights Electronics revenue by region
  -> KPI Card updates to show Electronics-specific metrics
  -> Table filters to Electronics rows

Cross-filtering is implemented via a shared filter context in the BI Workbench frontend (Zustand store) that propagates filter changes to all connected widgets.

The semantic-layer service (Java, port 8086) provides a business-friendly abstraction over raw data schemas. It is the foundation for consistent metric definitions across the entire platform.

Metrics Definitions

A semantic metric defines how a business concept maps to a SQL computation:

Metric: "Revenue"
  SQL: SUM(order_total) WHERE order_status = 'completed'
  Granularity: daily, weekly, monthly, quarterly, yearly
  Dimensions: product_category, region, customer_segment, channel
  Filters: Exclude returns, exclude internal orders
  Owner: Finance team
  Data Quality Threshold: 95% completeness required

Metrics defined in the semantic layer are:

• Reusable -- The same metric definition powers dashboards, conversational queries, and API responses
• Consistent -- All consumers see the same computation, eliminating "my numbers don't match" conversations
• Governed -- Metric definitions require approval to change, with full version history
• Quality-aware -- Each metric references data quality scores, and consumers are warned when quality drops below thresholds

Dimension Hierarchies

Dimension hierarchies define drill-down paths for interactive exploration:

Row-Level and Column-Level Security

The semantic layer enforces access policies that apply regardless of how data is accessed:

The render-service (Node.js, port 8098) handles server-side rendering of visualizations and documents:

Scheduled Reports

The bi-service can schedule automated report generation and distribution:

1. Schedule definition -- Cron expression (e.g., "every Monday at 8:00 AM") configured per dashboard
2. Render trigger -- bi-service calls render-service to generate PDF/PNG of the dashboard
3. Distribution -- notification-service delivers the rendered report via email, Slack, or Microsoft Teams
4. Recipient management -- Per-dashboard recipient lists with tenant-scoped access control

A unique capability of MATIH is the seamless integration between dashboards and conversational analytics:

This integration means that dashboards are not static artifacts. They are living documents that can be explored, questioned, and enriched through conversation.

Deep Dive References

• BI Service Architecture -- Complete bi-service implementation details
• Frontend: BI Workbench -- React component architecture, state management, and widget system
• Semantic Layer -- Detailed semantic layer configuration and MDL (Metric Definition Language) reference
• Render Service -- Server-side rendering pipeline and PDF generation