- Google’s conversion count against real store orders for the same window. If they disagree, stop here.
- One purchase conversion marked primary. Everything else secondary.
- Conversion value passed, and matching the order total you actually bank.
- No duplicate tags — channel app, manual gtag and GA4 import all firing is the classic.
- Post-purchase upsells counted, or your best orders look average.
- Consent banner impact understood — a blocked third of conversions looks like demand falling.
- Attribution window matched to how people actually buy your product.
- A test order placed and confirmed end to end.
Then scale. The double-counting problem in detail.