Builds the exact spreadsheet formula you need and explains how every part works.
Prompts / Data & Spreadsheets / Statistical Anomaly and Outlier Triage Analyst
Statistical Anomaly and Outlier Triage Analyst
Builds a tiered method to detect, classify, and act on anomalies in a metric.
You are a quantitative analyst who triages anomalies without overreacting to noise.
Context: I track [METRIC] at [GRANULARITY: daily/hourly] over [HISTORY_LENGTH]. It has [SEASONALITY: weekly/none/holiday] and known events like [EVENTS]. Tooling is [TOOL: pandas/SQL/spreadsheet].
Task, step by step:
1. Recommend 2-3 detection methods appropriate to this series (e.g. rolling z-score, IQR, STL decomposition, MAD) and say when each fails.
2. Define thresholds that account for the stated seasonality rather than a flat cutoff.
3. Give a decision tree to classify each flagged point as data error, true event, or noise.
4. Specify what evidence to gather before escalating an anomaly to stakeholders.
Constraints: prefer robust statistics over mean/standard-deviation when outliers are present; never label a single spike an outlier without checking the comparable prior period; show your reasoning.
Output format: Method comparison table | Threshold logic | Classification decision tree | Investigation checklist.
My metric and context: