To track actual purchases from Meta Ads, your ordering site needs to send a Purchase event to Meta every time an order is confirmed, with the order value, currency and a unique order ID. It should be sent from both the browser (Meta pixel) and your server (Conversions API), with the same event ID so Meta counts it once. Then you verify it with a real test order and reconcile Meta's count against your back office every week, because the two numbers will never match exactly.

Food delivery makes this harder than standard ecommerce. Orders come through the website, the app, phone calls and aggregators; checkout is often a third-party widget; and some orders are cancelled or changed after payment. This article covers the setup we use, the test we run before spending, and how to read the gap between Meta-reported purchases and the orders your kitchen actually cooked.

What "tracking purchases" actually means

There are three different numbers, and it helps to keep them apart:

  • Meta-reported purchases. Purchase events that Meta attributes to your ads, inside its attribution window. This is what campaigns optimize toward.
  • Site-recorded orders. Every order your website or ordering system logged, from any source. GA4 ecommerce data usually sits here.
  • Back-office or POS orders. Orders that were paid, prepared and delivered. This is the number that pays the bills.

The goal is not to make the three equal. It is to make the first one accurate enough that Meta optimizes on real orders, and to know the ratio between the first and the third so you can judge spend correctly.

Step 1: audit what is firing today

Before building anything, check what already exists. In most food delivery sites we inherit, at least one of these is true:

  • The pixel is installed, but Purchase fires on page load of the thank-you page, so a refresh counts as a second order.
  • Purchase fires without a value, or with a hardcoded value, so Meta cannot optimize for order size.
  • The checkout is on a different domain or inside a widget, and nothing fires at all after payment.
  • GA4 has a purchase event, but the Meta pixel does not, or the other way round.
  • Several plugins each fire their own Purchase, so every order counts two or three times.

Use Meta Events Manager to see which events arrive, from which source (browser or server), and with which parameters. Check GA4 ecommerce events at the same time. You want one clear purchase event in each tool, firing once per confirmed order.

Step 2: send Purchase from browser and server, deduplicated

The browser pixel alone misses orders: ad blockers, browser privacy settings and people closing the tab before the confirmation page loads all reduce what it sees. The Conversions API sends the same event from your server, which is more reliable because it does not depend on the customer's browser.

Sending both is the recommended setup, but only if Meta can tell they describe the same order. That is what the event_id is for. Give the browser event and the server event the same event_id (the order ID works well) and the same event name, and Meta deduplicates them. Without a shared event_id, every order is counted twice.

The Purchase event should carry at least:

  • value and currency, taken from the actual order total
  • event_id, unique per order and identical across browser and server
  • content_ids, matching the product IDs in your catalog, if you run catalog ads
  • customer information parameters (such as hashed email or phone) sent through the Conversions API, which help Meta match the event to a person

Fire Purchase only when the order is confirmed by your ordering system, not when someone reaches a page.

Step 3: verify with a live test order

Do not trust any setup until a real order has gone through it. We place a live test order with a real payment, then check:

  1. Events Manager shows one Purchase, not two, with browser and server events deduplicated.
  2. The value and currency match the order total.
  3. The content_ids match items in the catalog.
  4. GA4 recorded the same order with the same transaction ID.
  5. The order appears in the back office, and the test can be identified and excluded from reporting.

Events Manager's test events tool lets you watch events arrive in real time while you place the order. If anything is off, fix it and test again. A tracking problem found after a month of spend costs a month of optimization on wrong data.

Step 4: build the catalog from your menu

For food delivery, a Meta catalog turns your menu into ads that can show individual dishes and sets. It is also what lets Meta connect a Purchase to the specific items ordered, as long as the content_ids in the event match the IDs in the catalog.

Keep the feed tied to your real menu: correct names, prices, photos and availability. Items that are out of stock or discontinued should come out of the feed, or the ads will sell food you cannot cook.

Step 5: optimize campaigns for Purchase

Once Purchase is reliable, set campaigns to optimize for it. Campaigns optimizing for traffic or engagement will find people who click and scroll; campaigns optimizing for Purchase look for people who order. Purchase optimization needs a steady flow of events to learn from, which is one reason the tracking work above comes first.

What this looked like for a sushi delivery brand

We worked with a sushi delivery brand in Ukraine that had an online menu but weak tracking and no purchase-optimized campaigns. We audited the Meta pixel and GA4 ecommerce events, verified Purchase with a live test order, built a Meta catalog from a 101-item product feed, and launched purchase-optimized campaigns at about $10 per day.

In the first three weeks the campaigns recorded 29 purchases and $610 in revenue on $172 of spend, a ROAS of 3.56 and about $5.90 per purchase. These are Meta-reported figures. Campaigns the brand ran before we started are excluded.

We state "Meta-reported" deliberately. Those 29 purchases are orders Meta attributed to the ads within its attribution window. The back office will show a different number, for reasons covered in the next section.

Why Meta and your POS will not match

Expect a gap between Meta-reported purchases and back-office orders. Common reasons:

  • Attribution windows. Meta credits a purchase made within a set period after someone clicked or viewed an ad. Some of those people would have ordered anyway.
  • Other channels. A person sees the ad, then orders by phone or through an aggregator. The order is real, but Meta never sees it.
  • Cancellations and changes. An order can be refunded or edited after the Purchase event fired.
  • Duplicates or misses. Imperfect deduplication inflates the count; blocked browsers without a server event deflate it.
  • Test orders. If they are not excluded, they show up in every count.

We cover the broader reconciliation method in attribution after iOS. In practice: compare weekly, track the ratio, and investigate when it moves sharply.

A worked calculation: what the gap does to your margin

The following is illustrative arithmetic, not a client result.

Say Meta reports 30 purchases on $180 of spend, so $6 per purchase, with an average order of $21. Suppose that after food, packaging and delivery you keep $7 per order before advertising. At $6 per purchase, the ads look profitable on the first order.

Now suppose that when you match them against the back office, only 80% of those reported purchases turn out to be paid, delivered orders. That is 24 orders, and the real cost per order becomes $180 / 24 = $7.50. You are now slightly losing money on the first order, and the decision about whether to scale depends on how often those customers reorder.

The same campaign can be profitable or not depending on a ratio most businesses never measure. That is why we reconcile before we scale.

Checklist: tracking Meta Ads purchases for food delivery

  • One Purchase event per confirmed order, not per page load.
  • Value and currency taken from the real order total.
  • Pixel and Conversions API both active, sharing the same event_id.
  • Customer information parameters sent server-side where you have consent to do so.
  • A live test order verified in Events Manager, GA4 and the back office.
  • Test orders identifiable and excluded from reports.
  • Catalog built from the live menu, with content_ids matching the event.
  • Campaigns optimizing for Purchase, not traffic.
  • Weekly comparison of Meta-reported purchases with back-office orders.
  • Phone and aggregator orders tracked separately, so you know what Meta cannot see.

How we work on food delivery accounts

We start with measurement. Before touching budget we audit the pixel, the Conversions API setup and GA4, place a live test order, and fix whatever is missing. Then we build the catalog, launch purchase-optimized campaigns at a modest daily budget, and report Meta-reported numbers next to back-office numbers so you see both.

Results depend on your menu, prices, delivery area and how fast orders arrive, so we do not promise a ROAS. We do promise that the numbers you see are labelled honestly. If the ordering site itself needs work, our web development team handles the checkout and event setup alongside the Meta Ads work.

To find out whether we are a fit, apply here. We reply in writing; the plan for your account is part of the paid Growth Sprint. If your business generates leads rather than orders, the equivalent problem is covered in why solar leads are cheap but not converting.

Frequently Asked Questions

Is the Meta pixel enough to track food delivery orders?

Usually not. The pixel depends on the customer's browser, which can block it or close before it fires. Adding the Conversions API sends the same Purchase from your server, and a shared event_id stops Meta from counting the order twice.

Why does Meta show more purchases than my POS?

Meta counts purchases within its attribution window, which can include people who would have ordered anyway, and may count cancelled orders or duplicates if deduplication is incomplete. Your POS counts what was actually paid and delivered. Track the ratio between the two over time.

Why does Meta show fewer purchases than my POS?

Orders placed by phone, through aggregators, or after the ad on a different device often never reach Meta. Missing server-side events and blocked browsers also reduce the count. Fewer reported purchases does not always mean the ads are underperforming.

Do I need a catalog for food delivery ads?

It is not required, but it helps. A catalog lets ads show specific dishes and lets Meta connect purchases to items, provided the content_ids in your Purchase event match the catalog IDs.

How much should I spend to test purchase-optimized campaigns?

Start with a budget your business is comfortable losing while the campaigns learn, and judge it after the tracking is verified, not before. A modest daily budget is enough to confirm that real orders are recorded correctly.

Intelligent Syndicate Research

Written by the operators who run the accounts. No ghostwriters, no invented personas.