Facebook ads for Shopify: setup, tracking and measuring what they sell

Shopify's Facebook & Instagram sales channel does the plumbing. In order:

1. Install the Facebook & Instagram sales channel from the Shopify App Store and connect your Meta business account and Page.
2. Create a Meta pixel through the channel, or add the one you already have. One pixel per store keeps reporting in one place.
3. Choose the customer data-sharing level. Higher levels send more purchase data to Meta, including through the Conversions API, which doesn't depend on the shopper's browser.
4. Let the channel sync your products to a Meta catalog. Dynamic ads and shops read from it.
5. Build campaigns in Meta Ads Manager. Shopify points to Ads Manager for creating ads across Facebook, Instagram and Messenger.

Since August 26, 2025, Shopify says Meta no longer collects sales tax on Facebook and Instagram checkout orders, so check your tax settings if you sell inside Meta's apps.

Shopify's help center lists three ways to feature your products once the channel is set up:

Retargeting reaches people who were already close to buying, so its reported return overstates what it adds. Prospecting reaches people who weren't, so more of what it sells can happen after Meta's window closes or off Shopify.

Shopify suggests waiting 4 to 5 days after a launch before changing anything, and reviewing cost per click, click-through rate, cost per purchase and frequency. The Breakdown menu in Ads Manager splits results by placement and audience.

Keep in mind what the purchase column is. Meta counts a purchase when its pixel or the Conversions API reports one within its attribution window after someone clicked or saw an ad. That's credit under Meta's rules, and it can be too high or too low:

• Too high, when the buyer would have bought anyway: a returning customer who tapped a retargeting ad on the way to checkout.
• Too low, when the buyer purchases somewhere Meta can't see: on Amazon, in a store, or long after the window closed.

The attribution models guide explains why every report splits credit differently, and why none of them can say what the ads caused.

Shopify records each order's shipping address, so you can run a geo holdout test with data you already have:

1. Split the US states into two groups whose orders have moved together over the past few months.
2. Pause Meta ads in one group for several weeks using location targeting. Change nothing else.
3. Compare orders in the two groups during the test with how they compared before it.

lift = (ads-on orders ÷ holdout orders, during) ÷ (ads-on orders ÷ holdout orders, before) − 1

If you also sell on Amazon, count those orders by ship-to state too. Many shoppers who see a Facebook ad buy wherever they already shop. Size the test first with the holdout test calculator, and see how lift is calculated for the full method.

A made-up store spends $5,000 a month on Meta ads. Ads Manager reports $15,000 of purchases, a ROAS of 3.0, much of it from retargeting.

A geo holdout test finds the ads caused about $9,000 of Shopify sales and $2,000 of Amazon sales: $11,000 in all, an incremental ROAS of 2.2. Ads Manager over-credited the store and missed Amazon entirely.

Whether 2.2 is enough depends on your break-even ROAS, which comes from your margin. The contribution margin calculator shows what each order leaves on Shopify and on Amazon after fees and costs.

• Two pixels on one store. Purchases get counted twice or split, and nothing adds up.
• Judging retargeting on reported ROAS. It reaches people who were already close to buying.
• Changing campaigns every day. Shopify suggests waiting 4 to 5 days after a launch.
• Ignoring sales off Shopify. If you also sell on Amazon, some of what your ads sell lands there.

Spillmark runs geo holdout tests for brands that advertise on Meta and sell on Amazon. It plans the test from your own order history and counts Shopify orders by ship-to state as well as Amazon's, so one test shows the lift on each store. It locks the design before the test starts and reports the result with its uncertainty range.