3.3.5. Advanced Analytics
💡 First Principle: A single slow laptop is a ticket; the same slowdown on 300 laptops of one model after one driver update is a pattern, and patterns are invisible one ticket at a time. Advanced Analytics looks across the fleet for those patterns — and then lets you drill into a single device's history or query it live — so you fix causes instead of symptoms.
Foundations. Advanced Analytics is built on Endpoint analytics, so devices must meet Endpoint analytics prerequisites and Endpoint analytics must be configured. Its reports and single-device views support Windows devices that are Intune-managed or co-managed and Microsoft Entra joined or hybrid joined; device query for multiple devices, covered below, also reads collected inventory from other platforms. Once licensed it's enabled automatically, though features can take up to 48 hours to appear. It's included in Microsoft 365 E3 since mid-2026.
What it adds. To the Endpoint analytics reports (Reports > Endpoint analytics): Resource performance, Battery health, Anomalies and Device scopes. To single-device views: battery health, resource performance, Device timeline (which replaces the Application reliability tab) and Device query. Beyond those, it adds device query for multiple devices and the audit-only STIG baseline for Windows 11. Device query is covered in 3.4.3.
Anomaly detection watches for application hangs, application crashes and stop error restarts, and grades each anomaly by severity. For medium- and high-severity anomalies it builds device correlation groups, grouping affected devices by shared attributes — app version, driver update, OS version and device model — and showing both devices affected now and devices at risk because they share those attributes. The prevalence rate is the percentage of devices in a correlation group that are affected. Detection uses statistical models, including a threshold-based heuristic whose thresholds you can't customize, and Microsoft describes the goal as correlating deployment objects and configuration changes to suggest root causes.
Device timeline (Devices > Windows > device > User Experience > Device Timeline) shows a device's app crashes, unresponsive apps, boots, sign-ins and detected anomalies in order.
Battery health scores each device as a weighted average of a battery capacity score and a battery runtime score, each 0–100. Capacity below 60% counts as most impacted, 60–80% as moderately impacted, and an estimated runtime under 3 hours as most impacted. Resource performance scores CPU and RAM spike time over 14 days, where a spike means CPU above 50% or RAM above 75%; it covers Windows 365 Cloud PCs as well, and its insights can suggest upgrading a Cloud PC's configuration.
Custom device scopes filter the Endpoint analytics reports by a single scope tag; you can save up to 100 with 20 active, and a scope needs at least 10 devices.
"Risk-based policy recommendations." The outline uses this phrase, but it isn't the name of a button in Intune — it describes letting analytics and threat signals drive policy decisions instead of static rules that treat every device the same. Microsoft's training gives examples:
- Feed device and threat signals into Conditional Access, so moderate risk prompts for MFA and high risk is blocked.
- Pause an update ring and run remediations when crashes spike after an update.
- Replace degraded batteries before they fail, based on battery health.
- Temporarily exclude a device model hit by an anomaly from strict enforcement while it's fixed.
- Write custom compliance scripts (PowerShell on Windows, Bash on macOS) that check the conditions analytics revealed.
The built-in recommendations you'll actually see live inside the Battery health and Resource performance reports — devices with low capacity or runtime, high CPU or RAM spikes, or Cloud PCs that need a bigger configuration.
⚠️ Exam Trap: Advanced Analytics doesn't replace Endpoint analytics — it extends it. If Endpoint analytics isn't configured, or devices don't meet its prerequisites, the Advanced Analytics reports have nothing to show, whatever license the tenant holds.
Reflection Question: After a driver update, crash reports rise on one laptop model. Which Advanced Analytics view tells you the model is the common factor, what does "at risk" mean in that view, and what risk-based action could you take while waiting for a fix?