Command
Use list_projects({}) to confirm my supplied project host is visible. Use my chosen days value of 7, 30, or 90 for every time-windowed call. Read get_overview({host, days}) and get_agent_analytics({host, section: "conversions", days}). Join only compatible dates and identifiers in my supplied revenue export, documenting its attribution model and missing rows. Separate observed conversions, attributed revenue from that export, and simple co-movement. If identifiers or channel definitions disagree, report the mismatch instead of joining. Draft an evidence table and a measurement plan; claim no causal lift. Return evidence links, coverage limits, and explicit unknowns; never infer unsupported scores.What you need
- A visible project host and a 7, 30, or 90 day window.
- A revenue export with dates, join keys, channel definitions, and attribution rules; any warehouse connection is independent.
- The external reference skill targets another provider; use the Kept prompt above or adapt its tool calls before running it.
Check whether the data can be joined
Visibility and revenue describe different events. Check date coverage, identity keys, and attribution definitions before comparing them. Keep unmatched records visible so a convincing chart cannot hide missing evidence or turn an aggregate association into a customer journey.
Build a defensible measurement note
Deliver a table separating Kept observations from revenue figures supplied by your team. Include the join method, exclusions, and unanswered questions, then propose what additional tracking would make a future attribution study more reliable for commercial decisions.