Skip to main content
FHAVI

Free resource

Fifteen checks that tell you whether your tracking is telling the truth

Every ad budget decision rests on events you cannot see. This is the sheet we work through on a store before touching a campaign — run it yourself in one session, with one real order, and you will know whether the numbers you report are the numbers that happened.

Read this first

The event list is not the question

Most merchants check that a pixel is installed, see a green tick in an extension, and stop. Installation is the easy half. The failures that cost money are quieter: an event firing twice, a purchase counted from the browser but not the server, a value in the wrong currency, or two systems that both send the same sale because nothing joins them.

The checks below are ordered the way failures actually surface — browser first because it is visible, then server, then the join between them, which is where the expensive mistakes live.

Do them in one sitting, on production, with a real order you will cancel afterwards. Test data inflates the same reports you are trying to trust.

The sheet

Fifteen checks, in the order they break

01

Browser side

What a normal visitor with no ad blocker should produce.

  1. The base pixel loads on every page and sets a _fbp cookie — not just on the landing page.
  2. The base code runs once. Two copies of the snippet means two PageViews per visit, and everything downstream is doubled.
  3. PageView fires once per navigation, including in-app route changes on a single-page storefront.
  4. ViewContent fires on a product page with a content id that matches your catalogue, not a slug, not a name.
  5. AddToCart carries value, currency and quantity that match the product's real price.
  6. InitiateCheckout fires once when the checkout starts — not on every keystroke, not on cart open.
02

Server side

The half that survives ad blockers, iOS restrictions and cookie loss.

  1. Purchase is sent from your server with the order number as the event id, so a refresh cannot create a second sale.
  2. The server payload carries hashed email and phone. In cash-on-delivery markets the phone number is the only identifier that is reliably real.
  3. Client IP, user agent, _fbp and _fbc are forwarded with the server event — without them match quality collapses.
  4. The access token lives in server environment variables or an encrypted settings row, never in the JavaScript bundle.
03

The join

Where double counting happens, and what it looks like.

  1. Browser and server events for the same action share one event id, so Meta merges them into one conversion instead of two.
  2. In Meta Events Manager, a test purchase shows as a single event with both Browser and Server sources — not two separate events.
  3. GA4 purchase uses a transaction id, so the Measurement Protocol send and any browser send collapse to one transaction.
  4. Values match end to end: order total on the site, value in the event, revenue in GA4, value in Meta. Currency is the store's actual currency in all four places.

Reading the result

Three answers, only one of which is a job

A checklist that ends in a column of ticks has told you nothing useful unless you know which failure you are looking at.
04

All fifteen pass

Stop here.

  1. Your reports and your orders agree. If ROAS still looks wrong, the problem is creative, offer or margin — not measurement.
  2. Re-run the sheet after any theme change, checkout change, domain change or platform migration.
05

Browser passes, server fails

The most common and least visible failure.

  1. Symptom: Meta reports fewer purchases than your order count, and the gap is not explained by cancellations.
  2. Cause is usually a missing token, an expired token, a payload missing user_data, or a forwarder that silently swallows errors.
  3. Fix the forwarder before spending anything. Increasing budget on a broken signal teaches the platform to find the wrong people.
06

Both sides pass, the join fails

The expensive one.

  1. Symptom: revenue looks better than the bank account. Purchases are counted twice — once from the browser, once from the server, with different ids.
  2. This makes bad campaigns look good, which is worse than a gap, because it funds the wrong ones.
  3. The fix is mechanical: one id generated per action, reused by both senders.

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

Ran the sheet and found a gap?

Send us the failing check and the platform you are on. We will tell you what it usually is and whether it is worth fixing yourself.

Prefer email? Write to md@fhavi.com