BI News

Real-time analytics in Fabric: Eventstream, KQL, and the BI layer

By Adrien Yvin ·

fabricrealtimeeventhousekql

Batch BI still runs the business. But ops teams now ask for minutes, not nightly refresh. Fabric’s real-time path connects Eventstream, Eventhouse, and your existing Power BI layer.

The streaming stack in Fabric

Typical flow:

  1. Eventstream ingests IoT, app events, or CDC streams
  2. Eventhouse (KQL database) stores high-velocity tables optimized for time-series queries
  3. Semantic model or DirectQuery exposes curated views to Power BI
  4. Optional Real-Time dashboard for sub-second ops tiles without full report chrome

You do not need every event in a Import model. Most teams aggregate in KQL, then surface hourly or 5-minute grains to BI.

Power BI vs Real-Time dashboards

Need Surface
Exec KPIs with drill-down Power BI report on aggregated KQL export or DirectQuery
NOC wall, line status, alerts Real-Time dashboard
Combined batch + live Hybrid: Import history + live tile from KQL

Governance at velocity

  • Retention policies on raw Eventhouse tables (7 days hot, archive to OneLake)
  • PII scrubbing in Eventstream before persist
  • Separate workspace for streaming experiments vs certified BI models
  • Document which measures are live vs delayed on report headers

BI developer tips

  • Push complex shaping to KQL materialized views, not DAX on billion-row imports
  • Test query cost on capacity during peak event rates
  • Name streams and databases like warehouse schemas (ops.machine_events, not test123)

Pilot scope

One line of business, one event type, one dashboard with three metrics. Run parallel to batch report for two weeks. Compare incident response time before expanding.

Exploring Eventstream to Power BI for your plant or app telemetry? Book a 30-minute call.