Diagnose ad performance before changing spend
Build a hypothesis-driven review from enough data to support a decision.
What you’ll produce
A diagnostic memo with observations, possible explanations, and a limited test proposal.
Inputs & connections
- Approved campaign and conversion exports
- Unit economics, attribution window, and test constraints
Begin with approved exports or pasted source material. Live access requires separately configured integrations and permissions; this guide does not install them.
Your starting prompt
Replace the bracketed fields. This is our suggested prompt, not a quote from the source.
Review [ad performance data] for [period] against [business objective]. Separate observed changes from hypotheses. Consider spend, volume, conversion quality, attribution delay, and sample size. List missing evidence before recommending a test. Define one bounded experiment with a control, budget limit, stop condition, and review date. Do not modify campaigns or promise a cost reduction.
How to run it
- 01
Check tracking health and the definition of a valuable conversion.
- 02
Prepare comparable time windows and record external changes.
- 03
Generate competing explanations and choose what can be tested.
- 04
Approve a bounded experiment and review its full business result.
Review before you act
The media owner validates attribution, sample size, spend limits, and conversion quality.
Limitations & success checks
- A lower cost per lead can accompany lower lead quality.
- One creator’s account-level result is not a general prediction.
Measure: Qualified conversion cost and incremental evidence from approved tests.
Sources & attribution
The sources below support the reported example or a tool capability. They do not validate our prompt or prove the outcome in your environment.
May 15, 2026. Reports an advertising-agent experiment. Performance claims describe one account and have not been independently reproduced.
Creator / publisher: Cody Schneider
Editorial review: October 11, 2026. How we label evidence.