Set defensible alerting thresholds for a business metric

I want to set up alerting for a business metric and need help choosing defensible thresholds. Here is the context:

- Metric name and definition: {{METRIC_NAME_AND_DEFINITION}}
- Historical data (paste a sample or describe the range, mean, and typical week-over-week or day-over-day variability): {{HISTORICAL_DATA}}
- Alert channels and who receives them: {{ALERT_RECIPIENTS}}
- Business cost of a false positive (alert fires but nothing is wrong): {{FALSE_POSITIVE_COST}}
- Business cost of a false negative (something is wrong but no alert fires): {{FALSE_NEGATIVE_COST}}

Please do the following:

1. Based on the historical data I provided, estimate the metric's baseline variability (coefficient of variation or typical percentage swing if I gave you raw numbers).
2. Recommend threshold values for three alert levels: Warning, Critical, and Incident. Express each as a percentage change from a rolling baseline (specify the window) and give the absolute value equivalent if possible.
3. Explain the statistical reasoning for each threshold level in plain terms a non-statistician can understand.
4. Estimate the expected false-positive rate per week at each threshold given the historical variability.
5. Suggest one leading indicator I could monitor alongside this metric to give earlier warning before the primary metric crosses a threshold.
6. Note when this approach breaks down: if the metric has strong day-of-week seasonality or a long-term trend, static thresholds will drift and you should re-evaluate them on a cadence—suggest how often.