What it is
A Real-Time Dashboard is a KQL-native dashboard surface for live data. Tiles execute KQL queries against an Eventhouse or KQL Database. Parameters cascade across tiles. Drill-through opens a detail tile or a query workbench. The whole thing refreshes on a clock — typically 10–60 seconds — without page reload jank.
vs Power BI Dashboards
Different jobs.
- Real-Time Dashboards — KQL-native, sub-second, operational. Best for live telemetry and time-series. Auto-refresh that doesn't stutter.
- Power BI Dashboards (and reports) — DAX-native, analytical, narrative. Best for curated metrics, executive views, and historical trend analysis.
Use both, for different audiences.
Building one
- Define your time filter parameter first. Default to "last 1h" and make it user-overridable.
- Author tiles as KQL queries. Time-chart, top-N, scorecard, table — each tile is a query that returns a table.
- Parameterize. User-selectable filters cascade across tiles. Sensor ID, tenant, region.
- Pin to OneLake catalog. Endorse it so others find it instead of building copies.
- Set the auto-refresh. 10s for live monitoring, 60s for "trending right now" views.
Best practices
- Always time-bound queries. A tile that scans all-time data will eventually hit a wall.
- Use materialized views for hot tiles. Cheap to query, always fresh.
- Restrict drill-through to the most useful slices. Too many breaks the operator's focus.
- Pair every Real-Time Dashboard with at least one Activator alert. If a dashboard shows it, someone should hear about it.