Turn campaign data into a report with defensible calculations

A marketing report should make its calculations inspectable. Ask the assistant to validate the data before recommending changes. This workflow analyzes a supplied table and does not edit ads or budgets.

Prepare the input

Provide the period, column definitions, currency, attribution window, and known exclusions. Include raw counts alongside rates where possible. Clarify whether conversions are leads, purchases, or another event.

Run the workflow

Ask for data checks first: missing values, duplicates, inconsistent units, and totals. Then compute only the metrics supported by the table. Require the formula and denominator for each rate.

Prompt to copy

Data: [table]. Definitions and period: [details].
Validate the data, then report supported metrics with formulas.
Separate observations, possible explanations, and proposed tests.
Do not infer revenue if it is absent or causation from correlation.
Flag small samples and incomparable periods.
Do not change campaigns, bids, or budgets.

Review a worked example

Illustrative data has 100 clicks and 5 leads. Lead conversion is 5/100, or 5%. It is not a purchase conversion rate. If spend is provided but revenue is not, a return-on-ad-spend claim is unsupported. Check a few calculations manually and inspect the attribution assumptions.

Compare period lengths and conversion delays before declaring a trend. Recent clicks may not have had time to convert under the stated window. Keep arithmetic errors separate from attribution uncertainty. If the data cannot answer the business question, the useful output is the missing data request, not a confident recommendation.

Use it again

Keep metric definitions stable across reports and explain changes. Review recommendations as hypotheses to test. Explore Bot for personal analysis or contact Golub about controlled reporting connections.