DETECT hunt Tools

MeasureTune, Retire

Check coverage and alert load

Detection counts overstate coverage. A detection program is working when its live detections cover the ATT&CK technique set and their alerts fit your triage hours. This page measures both, and both change when you move the levers. For your own techniques one by one, use Map your ATT&CK coverage.

InteractiveFramework v0.136.0, updated November 17, 2026.

Tips

  • Open your own metrics.json to see your program in place of the sample.
  • Raise the minimum robustness to see which coverage survives.
  • Set your cap on analyst minutes per true positive to see each detection's precision floor.
  • Run the retirement check on one detection at a time. It tests all five retirement triggers.

The method behind it Metrics

The kit's python tools/metrics.py writes it to site-data/metrics.json. Open it or drop it on this page. It stays in this browser.
On this page

Coverage

Coverage is counted per ATT&CK tactic. Each detection counts as its Summiting the Pyramid level divided by 5. Seven detections matching a tool's filename score below four matching the behavior: renaming a binary defeats the first group. Raise the minimum level to see what survives.

Robustness-weightedDetection count

Alert budget

A detection's run cost is its alerts times the minutes to triage one. Set your numbers and the page works out what fits, what the floor is, and what has to give.

The levers open on the sample program: 30 detections at 1.5 alerts a day each, 14 minutes a triage, three analysts with two hours. That program is over budget, so the levers start from one that has to give.

The cap is the framework default of 120 minutes.

Retirement check

Check one detection at a time. Enter what it did over the last two 30-day windows and the last 180 days, and the page runs all five retirement triggers against it, using the floor your cap sets. The page has only 180 days of input, so it cannot run the 365-day review trigger; a detection with no alerts in 180 days is flagged as a candidate for it.

    Program snapshot

    The sample program's measured numbers, with its 30-day precision checked against the floor your levers just set.

    Misses by cause

    Each incident or exercise that no live detection caught first, by the miss code its detection review recorded. The longest bar is where next quarter's fixes go.

    Definitions

    The framework defines numbers, each owned by one stage or by the program as a whole, and tags each as a KPI or a KRI. The repository produces the build-side ones from score.yml and the file layout; the rest come from what analysts record about each alert.

    Chart data: metrics.json written by tools/metrics.py. Definitions: framework.json v0.136.0, CC BY-NC 4.0. ATT&CK pin Enterprise 20.1. Level mix per tactic is spread from each tactic's weighted total, so the bars match metrics.json exactly.