Skip to main content
FHAVI

Free resource

Four numbers that separate a loss from a duplicate

When Meta disagrees with your order count, there are only two possible directions: Meta saw fewer sales than happened, or more. They have entirely different causes, and the fix for one makes the other worse. Establish the direction before you touch anything.

Start with the arithmetic

Write these four down before opening a support thread

Pull the same date range from both systems: orders in your store, purchases in Meta, sessions in Meta, and add-to-carts in Meta. Four numbers, one week, no filters. The ratios between them are the diagnosis.

Use a completed week, not today. Same-day numbers are still settling and will make a normal system look broken.

What you seeWhat it usually meansWhere to look first
Meta purchases ≈ ordersTracking is broadly working. Any gap is ordinary loss, mostly blockers and iOS.Improve match quality; do not restructure.
Meta purchases well below ordersLoss — events not reaching Meta, or reaching it without identifiers.Server forwarder, then browser pixel blocking.
Meta purchases above ordersDuplication — browser and server both counting, with different event ids.Event id pairing, then retry logic.
Purchases fine, value wrongPayload problem — currency, order total, or shipping counted twice.The value and currency fields in the server payload.

If it is duplication

Finding the second copy

01

Checks in order

  1. Open one recent purchase in Events Manager and look at how many separate events exist for the same order — one with two sources is correct, two events are not.
  2. Check whether the browser and server events share an event id. If the browser generates one and the server generates its own, that is the whole bug.
  3. Check whether the server retries on failure while generating a new id per attempt. A timeout plus a retry then becomes two purchases.
  4. Check whether Purchase fires both on the confirmation page reload and from the server on order creation. Two triggers, one order.
  5. Check whether a second pixel, container or plugin is also sending purchases — a duplicate tag is common after a theme migration.
02

What a correct setup looks like

  1. One event from the browser, one from the server, same event name, same event id, merged within 48 hours.
  2. The event id derives from the order number, so it is stable across retries and reloads.
  3. Deduplication is measured, not assumed: one test order, checked in Test Events with both sources visible.

If it is a loss

Finding the missing half

03

Checks in order

  1. Send a test event from the server and confirm it appears in Test Events. If it does not, the forwarder or the token is the problem.
  2. Check the server's response from Meta — not the fact that your code ran. A rejected event and an accepted one look identical in your own logs unless you record the response.
  3. Load the storefront with an ad blocker disabled and confirm the browser pixel fires. Then repeat with a blocker and see what disappears.
  4. Confirm user data is present: hashed email, hashed phone, client IP, user agent, fbp and fbc. Events without identifiers are received and then matched to nobody.
  5. Check that events are not being held by consent management for visitors who never grant consent — that is a real loss, and it is legitimate.
04

What good looks like

  1. Match quality per event is visible and improving, not merely present.
  2. Purchase carries an order id, a value, a currency and at least one hashed identifier.
  3. A dropped event produces a visible, readable error rather than silence.

After the worksheet

The judgment call

A 5–15% gap between orders and reported purchases is normal in a market with ad blockers and privacy restrictions. A 40% gap is a fault. The line between them is not a rule you can look up — it is the ratio your own store had before something changed.

That is why the fourth number matters: your baseline. Without it, every fluctuation looks like a crisis and every real regression looks like noise.

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

Want it diagnosed rather than guessed?

Send the four numbers and the platform you are on. We will tell you which failure you have, and whether the tracking audit is worth $149.

Prefer email? Write to md@fhavi.com