Skip to main content
FHAVI

Free resource

Getting Google Ads to count the sales you actually made

Google Ads reporting fails in a different direction from Meta's: it tends to count more, because a conversion action set to count every conversion will happily count the same order twice. This is the checklist for making the number defensible.

Group 1

Conversion actions

01

Decide these deliberately

  1. One primary action for a purchase, marked primary so it drives bidding — everything else secondary.
  2. Count set to “one” per purchase, not “every”, unless repeat purchases within a session are genuinely separate orders.
  3. Conversion window set to something defensible: 30 days is common, 90 mostly inflates.
  4. View-through conversion window kept short or off, since view-through is the easiest number to over-claim.
  5. Micro-conversions — add to cart, checkout start — marked secondary so they inform optimisation without bidding on them.
  6. No duplicate conversion actions for the same event, which is the most common cause of doubled counts.

Group 2

Values and currency

02

Checks

  1. Each transaction sends a dynamic value equal to the order total, not a fixed average.
  2. The currency in the conversion action matches the currency the account bills in and the store charges in.
  3. Shipping and tax are included or excluded consistently, and the choice is written down.
  4. The purchase tag fires on the confirmation page only, not on the cart or the payment redirect return.
  5. For cash-on-delivery markets, decide whether placed orders or delivered orders are the conversion — and be aware the two give very different optimisation.

Group 3

Identity and consent

03

Checks

  1. Enhanced conversions are enabled for web, sending hashed first-party data rather than relying on cookies alone.
  2. Email and phone are normalised before hashing — lowercase, trimmed, country code on phone.
  3. Consent Mode is implemented if you take EU or UK traffic, with the four signals wired to the banner.
  4. Server-side sends respect the same consent state as the browser.
  5. No conversion is recorded for an order that was cancelled before payment cleared.

Group 4

Verifying against reality

04

The reconciliation nobody does

  1. Pick a week and compare three numbers: orders in the store, conversions in Google Ads, and conversions attributed to paid search.
  2. A small gap is normal; a large one is a counting configuration, and a negative gap means duplication.
  3. Check the conversion action's own report for “all conversions” versus “conversions” — the two columns are not the same number.
  4. Confirm the attribution model in use, and that a change of model was recorded with a date, because a model change looks like a performance change.
  5. Keep the weekly comparison. It is the only way to tell a tracking regression from a market shift.
05

What Google cannot see

  1. Orders declined at delivery or returned after dispatch — in cash-on-delivery markets that is a large share of reported conversions.
  2. Sales where the user blocked storage and no modelled conversion applies.
  3. Cross-device journeys without any identifier, which are increasingly common.
  4. Which is why the delivered-order count, not the reported conversion count, is the number to judge spend against.

More resources

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

Ads reporting more sales than you can find?

That is almost always a counting setting or a duplicated action. Send the conversion action list and we will point at the culprit.

Prefer email? Write to md@fhavi.com