Skip to main content
FHAVI

Free resource

Three numbers for the same week, and which one you decide with

Every platform claims credit for the same order. Meta says 60, Google says 45, GA4 says 38, the store says 71. None of them is lying — they are answering different questions with different windows. This sheet shows the four side by side so you can pick one for decisions and stop rebuilding the same argument every Monday.

The sheet

Four columns, one week

01

What to pull

  1. Store truth: orders and net revenue by creation date, excluding cancelled and fully refunded orders.
  2. GA4: purchases and purchase revenue, session-scoped, with the attribution model stated.
  3. Meta: purchases and purchase value at the 7-day-click, 1-day-view setting you actually optimise on.
  4. Google Ads: conversions and conversion value, with the counting setting (one per click, or every conversion) stated.
  5. Spend per channel for the same seven days.
  6. Derived: reported ROAS per platform, and blended ROAS computed from store revenue over total spend.

Why they differ

The four reasons, in order of size

02

Know which one you are looking at

  1. Attribution windows: 7-day click and 1-day view on Meta against session-scoped in GA4 against last-click in Google. A purchase eleven days after the click is counted by one and not the other.
  2. Cross-device and privacy: a click on a phone, an order on a laptop, with no shared identifier. Nobody counts it consistently.
  3. Deduplication failures: the platform counts two purchases for one order because browser and server fired with different ids. This one is a bug and is fixable.
  4. Unowned orders: direct traffic, email, a customer who came back from memory. Platforms do not claim these and the store does.

Decision

Which number to run the business on

03

The rule that ends the argument

  1. Use blended ROAS — store revenue divided by all ad spend — for budget decisions, because it is the only figure nobody can inflate.
  2. Use platform-reported figures for comparison between channels, with the caveat that each is generous to itself.
  3. Investigate a gap only when it changes shape — a step change, or one platform diverging from the others. A stable gap is arithmetic, not a problem.
  4. Re-run the same week's sheet after any tracking change. If the numbers move without the business moving, the change was the cause.
  5. Write the model and windows you report on into the sheet header. Half of all attribution disputes are two people using different settings.

Keep going

The rest of the library

Every checklist here is free and ungated. The research pages carry the reasoning behind them.

All resources · Research library

Spending on numbers you cannot reconcile?

Send one week of exports from all four sources. We will build the comparison once and tell you which gap is real.

Prefer email? Write to md@fhavi.com