About this prompt
CAC Spike Diagnosis
When an acquisition metric jumps, the reflex is to pause something, and the pause usually destroys the data that would have explained the jump. This prompt makes the model behave like an analyst: it ranks three possible causes by how well the timing and magnitude of the evidence fit, lists what supports and contradicts each, names the one data cut to pull next and what result would confirm the top cause, adds a cause you probably have not considered, and ends with the action to avoid this week.
How to use. Give it the metric movement with dates, what changed inside the account during the window (creatives, budgets, bids, audiences, landing pages), what changed outside (competitor promos, seasonality, platform outages, tracking changes) and the channel mix. The more precise the change log, the better the ranking. Use the output as the body of your Monday anomaly note.
What to change. The prompt is written for CAC but works for CTR, CVR, ROAS, CPL or trial-to-paid — just change {{metric_movement}}. For product metrics (activation, churn), replace "performance-marketing analyst" with "growth analyst" and the structure holds.
Not a fit. Anomalies with no change log — the prompt will tell you it has insufficient data, which is the correct answer but not a useful one.
