Creative: test → promotion → scale

Business result

Scaled winners per unit of search spend and margin at a fixed outcome horizon

Next step

Understand the cause

Working output

SQL / notebook

Unit of analysis

Creative ID × launch attempt × branch × setup

Decision owner

To agree

Hypotheses and diagnostic checks

• State the leading explanation and at least one alternative before calculating.
• Define a check that can answer: What did this attempt produce, versus the creative's later production history?
• Define a check that can answer: Did candidates receive comparable exposure and equally completed outcome windows?
• Define a check that can answer: Is later creative value counted only once across repeat launches?
• Record the observation that would disprove each explanation.

Acceptance checks

• Each promotion is linked to an identified attempt.
• Cohorts have the same completed horizon; later value is not duplicated across ad sets.

Deliverables

• A diagnosis with competing explanations and the next check
• Read-only queries or notebook, parameters, results and a reproducible analysis dataset.
• A dated finding with source period, cutoff, methods, limitations and next action.