The key idea
AI-generated commentary must distinguish measured changes from hypotheses about why they happened.

Our method
Define metrics, periods and source ownership. Validate exports, flag missing values and calculate figures through reliable analysis tools. Ask for observations first, possible explanations second and checks needed third. Keep the calculations available for review.
How to assess progress
Reconcile totals, denominators and date ranges. A correlation is not a causal finding. Measure correction effort and decision usefulness as well as drafting time.
Decision
Choose the operational question the report must answer: where to investigate a change, which segment needs attention or whether measurement is trustworthy. Define the unit of analysis first. A campaign, a customer and a session cannot be substituted for one another without changing the question.
Exercise
Take two comparable reporting periods and reconcile their totals with the source. Decompose one important change by a meaningful dimension, such as channel or product group. Ask the model for three observations using the validated figures, then challenge each with an alternative explanation and the additional data needed to distinguish it.
Deliverable
Deliver a short investigation ledger: observed change, absolute volume, affected segment, plausible explanations, missing check, owner and next review date. Attach the calculation sheet so another analyst can reproduce the figures. Label recent or incomplete periods clearly rather than silently comparing them with mature data.
Pitfall
An overall improvement can hide deterioration within individual segments when the traffic mix changes. Inspect both totals and segment results. Do not recommend reallocating effort solely from a blended average, especially when conversion definitions or tracking coverage changed during the period.
Demos
A synthetic CSV covers four campaigns across two periods. The question: does improving overall return hide a problem?
View detailsWant this workflow built for your team?
We help marketing teams save time, deliver more and improve business results with AI solutions, workflows, tools and practical training.
Discuss my AI workflow