Independent campaign measurement
Is your ChatGPT Ads CPA as good as the platform says?
Ad platforms report the conversions they can attribute to themselves. This check compares that headline figure with your independently verified conversions, separates new customers from returning ones and gives you a transparent, arithmetic-only recommendation on what to do next — with no guesswork and no campaign figures leaving your browser unless you choose to share them.
Deterministic arithmetic only — no AI and no campaign figures stored unless you explicitly opt in.
What an independent check looks like
A single fictional campaign, shown end to end. The figures below are fixed for illustration so you can see how each number is derived before entering your own.
Figures entered
- Spend
- £4,800
- Platform conversions
- 160
- Verified conversions
- 103
- New customers
- 61
- Returning customers
- 42
- Target CPA
- £65
Platform CPA
£30.00
Spend ÷ platform-reported conversions
Verified CPA
£46.60
Spend ÷ independently verified conversions
New-customer CPA
£78.69
Spend ÷ new-customer conversions
New-customer share
59.2%
New customers ÷ verified conversions
Platform conversion overstatement
55.3%
Calculated as platform ÷ verified − 1.
Platform CPA understatement
35.6%
Calculated as 1 − platform CPA ÷ verified CPA.
Illustrative recommendation
Repair measurement
Platform-reported conversions are 55.3% higher than independently verified conversions. This exceeds the 50% measurement-repair threshold, so reconcile attribution before making any scaling decision.
Suggested holdout: Yorkshire and North East
A geo holdout is only suggested when there are enough comparable regions and sufficient historical volume to read a result.
Illustrative example — synthetic figures, not a market benchmark.
Why the numbers differ
Attribution is generous
Platforms count conversions they can plausibly claim, including view-through and assisted paths. Your own records usually verify fewer.
New and returning are mixed
A blended CPA hides how much you are paying to acquire genuinely new customers versus re-buying people who would have returned anyway.
A holdout settles it
The only reliable way to measure incremental effect is to hold a comparable region back and compare — when your data supports one.
Run an independent check
Enter your aggregated campaign figures. Everything is calculated in your browser — nothing is sent anywhere unless you choose to contribute at the end.
Campaign calculator
Enter aggregated figures only. Nothing is saved when you generate a result — the calculation runs entirely in your browser.
Regional figures (optional)
3/20 rowsAdd comparable pre-test and current-period figures to screen for a possible geo holdout. Leave rows blank if you are not testing regionally.
Methodology
Every figure is plain arithmetic on the numbers you enter — CPA is spend divided by conversions, overstatement is platform divided by verified minus one, and the recommendation follows fixed, published thresholds. There is no model, no estimate and no hidden weighting.
Privacy
Your campaign figures are calculated entirely in your browser and are only transmitted if you explicitly contribute them. Anonymous page-usage events may be recorded through Umami, but spend, CPA, conversion figures, email addresses and regional data are never sent to analytics. Questions: [email protected].
Limitations
This is a directional check, not an audit. It relies on the accuracy of the figures you enter, uses no external benchmarks and cannot see your underlying data. Treat it as a prompt for a proper test, not a verdict.
ChatGPT is a trademark of OpenAI. Aigeeza is independent and is not affiliated with or endorsed by OpenAI.