What this Skill helps with

Ad Performance Diagnostic turns campaign exports into a structured explanation of where performance changed and what evidence should be checked next. It separates delivery, traffic quality, conversion, tracking, and economics instead of blaming one metric in isolation.

Who this Skill is for

It is intended for paid-media managers, analysts, agencies, and founders investigating a drop in leads, sales, efficiency, or delivery.

What you provide

  • Campaign and ad-level exports for the affected and comparison periods
  • Spend, impressions, clicks, conversions, value, and relevant funnel metrics
  • Targets for CPA, ROAS, CPL, volume, or revenue
  • Notes about launches, budget changes, promotions, tracking, and site changes
  • Breakdowns by audience, placement, device, location, and creative when available

What it produces

  • A metric tree showing where the change occurred
  • Ranked hypotheses with supporting and conflicting evidence
  • Segments contributing most to the movement
  • Data-quality and tracking checks
  • A short investigation and action list

How the Skill works

The workflow compares periods, decomposes the result into volume and rate changes, and checks whether the movement is concentrated in a campaign or segment. It labels conclusions by confidence and keeps missing evidence visible.

Installation and first use

  1. Download and add the Skill to Claude.
  2. State the business target and the change you are investigating.
  3. Attach comparable exports and a change log.
  4. Ask for data-quality checks before recommendations.
  5. Confirm calculations and account context before making changes.

Example workflow and expected output

Provide four weeks of Google Ads data and the prior four weeks after CPA increased. Ask for a diagnostic table covering spend, CPC, conversion rate, CPA, campaign contribution, likely causes, confidence, and the next check for each cause.

Limits and review guidance

The Skill cannot see account settings, landing pages, CRM outcomes, or tracking tools unless you provide that evidence. Correlation does not establish cause. Validate any recommendation in the account and use controlled changes where possible.

Frequently asked questions

Can it diagnose a campaign from screenshots?

Screenshots may help, but exports with consistent date ranges and dimensions produce a more reliable comparison.

Does it need a benchmark?

A target or prior period helps. Without one, the Skill can describe movement but has less basis for judging impact.

Will it make account changes?

No. It produces hypotheses and a review plan.