Stop Refreshing: Analytics That React in Seconds, Not Hours
This masterclass will explore how Microsoft Fabric Real-Time Intelligence can be leveraged to ingest live telemetry, query data in motion, detect anomalies automatically and trigger action the moment a condition is met — building, testing and scaling streaming pipelines that turn real-time data into real-time decisions.
Most analytics tell you what happened. Some decisions expire before the nightly refresh has finished running — a sensor drifting out of range, a queue building faster than the team can clear it, a pattern that looks wrong right now rather than tomorrow morning.
This session builds a complete streaming pipeline from scratch: ingesting live telemetry, querying it while it is still arriving, detecting anomalies automatically, and triggering an action the moment a condition is met. It also covers the part most streaming demonstrations quietly skip — live data arrives late, out of order, and sometimes twice. We will deliberately break the pipeline, watch it report a confidently wrong number, and then fix it properly.
The techniques are demonstrated on Microsoft Fabric Real-Time Intelligence, but the underlying concepts — windowing, watermarks, and anomaly detection on time series —transfer directly to any streaming platform.
Learning outcomes
By attending this session, participants will discover:
1. From Batch to Data in Motion
- What actually changes when data is unbounded and never "finished"
- Ingesting and querying events while they are still arriving
- Building a dashboard that updates without a refresh button
2. The Problem Most Streaming Demos Skip
- Why late and out-of-order events produce confidently wrong answers
- Tumbling, hopping and session windows in practice
- Watermarks, and deciding how long to wait for stragglers
3. Detecting What Actually Matters
- Separating seasonality, trend and genuine outliers in time series data
- Applying built-in anomaly detection to a live stream
- Tuning sensitivity so real problems surface without causing alert fatigue
4. From Detection to Action
- Triggering alerts and downstream processes automatically
- Designing thresholds that teams will actually trust
- A latency test for judging when real time is genuinely justified
Target Audience
This session is aimed at data engineers, analytics engineers, BI developers and data architects who work primarily in batch and are evaluating streaming, along with technical leads responsible for monitoring or operational analytics. Familiarity with SQL is assumed; no prior experience with streaming or anomaly detection is required.
Attendees will leave with:
- A working mental model for windowing, watermarks and late-arriving data
- A practical route to anomaly detection on streaming data
- An understanding of how to turn detection into automated action
- A test for deciding when real time is worth its cost — and when it is not
This masterclass is part of our ‘Member Masterclass’ series delivered by our member organisations on a monthly basis. If you or your organisation would like to run your own masterclass for the benefit of our community, please get in touch.
