Repeatability
High
The logic is identical for every row: apply category mapping, check quantity threshold, compute per-SKU median and flag deviations. This is structurally the same operation at scale, with no instance-by-instance judgment required.
Ambiguity Tolerance
Medium
The bulk flag and price deviation rules are precisely defined. The category classification is the weak point—'SKUs don't always map cleanly' introduces ambiguity that requires either a human-supplied mapping table or a documented fallback rule for edge cases.
Data & Tool Availability
High
The input is a self-contained CSV export requiring no live API access or external permissions. A code agent with Python (pandas) or a data processing environment can execute this end-to-end with the file and a category mapping reference.
Error Cost
Low
The output is a new CSV with added columns—the original data is untouched and the result is fully reviewable before any downstream use. Miscategorizations are easy to spot-check and correct without irreversible consequences.
Human Judgment Required
Low
All three operations are rule-based and quantitative. The only human input needed is a one-time SKU-to-category mapping for ambiguous SKUs; the agent can flag unresolved ones for human review rather than guessing.