Repeatability
High
The structure is identical each time: ingest spend/performance data, compute ROI metrics by dimension, rank and flag outliers. This can be templated and run on a schedule with minimal variation in logic.
Ambiguity Tolerance
Medium
Core metrics like ROAS and CPA are well-defined, but 'underperforming' and 'highest ROI' require threshold decisions (e.g., ROAS below 1.5x? bottom quartile?) that the user hasn't specified. An agent can surface options, but someone needs to set the cutoffs.
Data & Tool Availability
High
18 months of structured daily data across defined dimensions is exactly what agents handle well. Assuming the data is exported or accessible via API, there are no significant access barriers — this is a clean, bounded dataset.
Error Cost
Medium
A misidentified 'underperformer' could lead to pausing a campaign that was actually recovering, wasting prior spend and losing momentum. Errors are reversible but not costless — budget decisions made on bad analysis have real financial consequences.
Human Judgment Required
Medium
Statistical analysis is fully automatable, but interpreting why a campaign underperforms (seasonality, creative fatigue, audience saturation, external events) and deciding whether to pause vs. restructure benefits from business context the agent may lack.