AI Skills Marketplace / Sales Engagement

Metrics Rollup & Anomaly Flagger

The Metrics Rollup & Anomaly Flagger is a free Claude Skill that normalizes sequence performance across platforms with different reply and open-rate definitions, and flags real statistical anomalies in a metric's history using a robust median+MAD modified z-score instead of a naive average a prior spike would distort.

↗ View Source ⬇ Download Full Repo (.zip)
MIT
License
6 functions
Decision branches
MPP unreliable-metric flag + minimum-baseline anomaly guardrail
Escalation rules
Yes
Multi-tool support
Aug 17, 2026
Last updated
26/26 passing
Tests
README.md View raw ↗

What this is (and isn’t)

Pure functions over the metrics data you supply, not a hosted dashboard or a direct integration with any specific sales engagement tool’s analytics API.

Files

FilePurpose
flagger.jsnormalizeReplyRate(), normalizeOpenRate(), rollupMetrics(), detectAnomalies(), median(), medianAbsoluteDeviation() — zero dependencies
test.js26-test suite covering rate normalization, cross-platform rollup, and anomaly detection including hand-verified statistics
SKILL.mdFull skill definition, loadable in Claude Code/Desktop

The two real guardrails

Honest, not invented, open-rate correction: a platform affected by Apple Mail Privacy Protection has opens inflated by prefetched tracking pixels, but nobody knows the true bias without ground truth, so it is flagged unreliable rather than silently “corrected.” And robust anomaly detection: the median+MAD modified z-score method (Iglewicz & Hoaglin, 1993) instead of a mean/stdev z-score, since a naive average lets one real prior spike distort the baseline used to judge every point after it.

FAQ

Is this skill free to use?

Yes. Free to download and run yourself. MV3 charges $175/hr only for implementation help wiring this into your real platforms' actual reporting APIs.

Does this pull data directly from my sales engagement tools?

No. It is a pure function over the raw counts you supply. You still need to pull each platform's reporting data and feed it in yourself.

Why not just average open rate across platforms?

Apple Mail Privacy Protection inflates opens on affected platforms by prefetching tracking pixels regardless of human interaction. Averaging that in with unaffected platforms would produce a number that looks precise but is not comparable. This skill flags it instead of hiding the problem.

Why median+MAD instead of a standard z-score for anomaly detection?

A mean/stdev z-score is distorted by the very spike or crash you are trying to detect, since that outlier pulls the mean and stdev toward itself. Median and MAD (median absolute deviation) are far more resistant to that, which is why this is the standard method (Iglewicz and Hoaglin, 1993) for flagging outliers in data that may already contain them.

Does my metrics data get sent to MV3?

No. All calculations run entirely in your own environment.

JR
Jordan Reeves
ABM & Outbound Pipeline, MV3 Marketing