Introducing the Insights Engine: Your Mac, Explained

The Problem with Raw Data

Traditional system monitors are designed for developers and sysadmins who can interpret raw metrics. They show you CPU percentages, memory composition breakdowns, and disk I/O rates — assuming you know what to do with that information.

But most Mac users don't need to know that their CPU hit 92% during a compile. They need to know why their laptop is running hot, whether it's normal, and what they can do about it.

That's the gap the Insights Engine fills. It transforms 30 days of historical data into actionable intelligence, surfacing patterns you'd never spot by staring at real-time gauges.

How It Works

The Insights Engine is entirely rule-based. There's no machine learning, no cloud processing, no data leaving your Mac. It queries the same SQLite database that powers the Historical Analytics view and applies a series of detection algorithms to identify meaningful patterns.

Every insight includes:

• A severity level — Warning (requires attention), Noteworthy (good to know), or Informational (FYI)
• A module tag — CPU, Memory, Disk, Thermal, Battery, Network, or GPU
• Plain-English explanation — What happened, why it matters, and what you can do
• Timestamp — When the pattern was detected

Insights refresh when you open the Insights view. Right-click any insight to dismiss it, snooze it (1 hour, 6 hours, or until tomorrow), or copy the details. Warning-level insights can also fire a macOS notification — throttled so the same warning can't spam you every few minutes while the condition persists.

What's New in 1.2.0

The Insights Engine picked up a lot in the 1.2.0 release. These are the changes most users will notice first:

• Insights History view. A new sidebar entry under Analytics shows every insight your Mac has produced, with 24-hour / 7-day / 30-day range, category, and severity filters. Insights had been persisted for 30 days all along — now you can actually browse them.
• Quick Actions. Each insight detail popover now has bordered buttons for the right follow-up: Open Activity Monitor, Storage Settings, Energy Settings, Disk Utility, Network Settings, Display Settings. One click takes you straight to the pane instead of hunting through System Settings.
• Snooze. Sometimes a rule fires in the middle of a task you know is going to produce high usage — a video export, a big compile, a backup window. Snoozing suppresses that specific rule for 1 hour, 6 hours, or until tomorrow, without disabling it permanently.
• Per-rule enable/disable. Settings → Insights now lists every registered rule with a toggle. Don't care about the night-time CPU rule because you run scheduled backups? Turn it off.
• Process-aware messages. Memory and CPU insights now name the actual apps. Instead of "Close unused apps to reduce memory pressure," you get "Top memory users right now: Google Chrome (3.2 GB), Xcode (1.1 GB), Slack (800 MB)." The detail popover shows the full top-5 breakdown.
• SMART disk health rule (Direct only) surfaces a warning the moment any drive's SMART status flips from Verified to Failing — ahead of the drive actually failing, while the data is still readable.
• Bluetooth reconnection churn rule (Direct only) flags devices that repeatedly connect and disconnect within an hour and names the worst offender. Good for catching a failing AirPod battery or radio interference before it ruins a call.

On the reliability side:

• Hysteresis on the CPU and Thermal "high usage / high temp" rules. Metrics oscillating around the threshold (79 → 81 → 79 → 81) no longer flip the warning on and off — rules require a 3-sample sustained streak before firing.
• Network anomaly floor. The rule now requires a baseline above 100 KB/s before comparing today's throughput. At tiny baselines a single browsing spike produced misleadingly huge ratios.
• Battery drain anomaly uses a median across per-interval rates instead of a global mean, so one short hot-GPU interval can't trigger a false alarm.
• Dismissals persist correctly. Previously a dismissal could lapse at midnight because the insight id contained the date. Dismissals now persist by rule id with a 24-hour TTL.
• Notification throttle. The same warning no longer fires a duplicate macOS notification every refresh — 24 h per (rule, severity).

What the Insights Engine Detects

CPU Insights

The CPU analyzer looks for patterns that indicate performance issues or unusual workload:

Memory Insights

Memory analysis tracks pressure events, compression ratios, and usage trends:

Disk Insights

Disk insights focus on SSD health and capacity planning:

Thermal Insights

Thermal analysis helps you understand heat patterns and fan effectiveness:

Battery Insights

Battery analysis monitors health, cycles, and drain patterns:

Network Insights

Network analysis detects unusual traffic patterns:

Designed for Clarity

We spent considerable time crafting the language for each insight. Technical accuracy matters, but so does accessibility. An insight about thermal throttling needs to explain what throttling is, not assume you already know.

Each insight also includes context. "CPU hit 90%" is less useful than "CPU hit 90% during your afternoon compile, which triggered thermal throttling for 3 minutes." The context transforms a data point into an understanding.

Getting Started

The Insights Engine is available in all versions of MacPulse (both Direct and Mac App Store). Some insights — like thermal throttling detection and fan curve effectiveness — require SMC access and are only available in the Direct version.

To view your insights:

1. Open MacPulse
2. Navigate to the Insights view in the sidebar
3. Review insights sorted by severity (warnings first)
4. Right-click any insight to dismiss it

Insights refresh automatically when you open the view. The more historical data MacPulse collects, the more accurate the pattern detection becomes — give it a few days to build up meaningful trends.