Facebook ads for ecommerce: setup, campaigns and what they really sell

The steps below are from Meta's Business Help Center as of October 2026. They work for any store platform:

1. Set up a business portfolio that you have full control of. Commerce Manager needs one to hold your catalog.
2. Install the Meta Pixel on your site. It logs events, such as adding to cart or buying, which Ads Manager uses to choose who sees your ads and to report results.
3. Turn on the Conversions API. It sends events from your server, store platform, app or CRM. Meta says this data is less affected than the pixel by browser loading errors, connectivity issues and ad blockers, and suggests running both. Many integrations don't need a developer.
4. Create a catalog in Commerce Manager and choose Online products. Add products from a partner platform, a data feed (a spreadsheet or XML file, uploaded once or on an hourly, daily or weekly schedule) or by hand. Connect your pixel so catalog ads can pick products for each person.
5. Create a campaign in Ads Manager with the sales objective, so delivery aims at purchases on your site.

Meta recommends one catalog where possible. On Shopify, the sales channel handles most of steps 2 to 4: see Facebook ads for Shopify. Meta's terms bar placing your pixel on sites you don't own, so it can't see sales on Amazon. Facebook ads for Amazon covers that setup.

Meta's current names for the two campaign types built around online sales:

For Advantage+ sales campaigns, Meta's advice is to connect your own customer data (the pixel and Conversions API), supply a wide variety of creative, and build audience segments so reports break results down by new, engaged and existing customers.

Catalog ads pick products using pixel or app events and activity on Meta. Meta lists three uses: retargeting people who viewed products and didn't buy, reaching people who haven't visited your site, and cross-selling or upselling. Retargeting reaches people who were already on their way to buying, so its reported results overstate what it adds.

Meta's A/B test compares two versions of an ad strategy, changing one variable such as the image, text, audience or placement. Meta shows each version to a separate part of your audience, makes sure nobody sees both, and recommends the same budget for each. It judges the winner on cost per result or cost per conversion lift.

Meta advises against testing by turning ad sets on and off by hand, because the audiences can overlap and the results aren't reliable. Change one thing per test so you know what made the difference.

An A/B test tells you which of two versions did better. It doesn't tell you whether either one sells enough to pay for itself. That's a separate question, covered below.

Meta says a new or significantly edited ad set goes through a learning phase while delivery works out who to show it to. It usually ends after about 50 results in the week after the last significant edit. Shops ads need at least 17 website purchases and 5 purchases through Meta. During learning, results are less stable and cost per result is usually higher.

That number sets a floor on budget. If a made-up store pays about $25 per purchase and optimizes for purchases, one ad set needs roughly this much a week to leave learning:

50 purchases × $25 = $1,250 a week

If an ad set can't get enough results, Ads Manager marks it "Learning limited". Meta's suggestions: wait to edit until learning ends, avoid frequent budget changes, combine similar ad sets instead of running many small ones, and set a budget that can reach enough results.

It depends on two numbers for your store: how many sales the ads cause, and how much each sale leaves after product costs, fees and shipping. The purchase column in Ads Manager gives you neither. Are Facebook ads worth it walks through break-even ROAS and holdouts for Amazon brands in more detail.

Meta counts a purchase on your site for an ad if it happens within a set number of days after someone viewed or clicked it, using the ad set's attribution setting. That's credit, not cause. Some of those buyers would have bought anyway. Others buy on Amazon or in a shop, where the pixel can't see. A study of large ad experiments at Facebook found that methods without a control group often missed the true effect, sometimes by a lot (Gordon and others, 2019). The attribution models guide explains why.

profit from ads = sales the ads caused × contribution margin − ad spend
worth it when incremental ROAS > break-even ROAS = 1 ÷ contribution margin

What is ROAS covers the difference between reported and incremental ROAS, and the break-even ROAS calculator finds your threshold from your own costs.

The direct way to find the sales your ads cause is a holdout group: places or people who don't see the ads, compared with those who do. For a store, a geo holdout test is the practical version. You pause Meta ads in a group of US states for a few weeks and compare orders there with the rest of the country, against how the two groups compared before.

Count orders from every place you sell. Your store and Amazon both record each order's ship-to state, so one test can show the lift on each. Size the test first with the holdout test calculator, and see how lift is calculated for the method. What is incrementality explains the idea from the start.

A made-up store spends $6,000 a month on Meta ads. Ads Manager credits them with $18,000 of purchases, a ROAS of 3.0. Each sale leaves a 40% contribution margin, so break-even ROAS is 2.5. On the report, the ads look profitable.

A geo holdout test finds the ads caused about $10,000 of store sales and $3,000 of Amazon sales: $13,000 in all, an incremental ROAS of 2.17. That's below break-even. The ads make $5,200 of margin on $6,000 of spend, a loss of $800 a month.

The answer for this store isn't "stop". It's to find which campaigns cause the sales and test again. Another store with the same report could find the opposite. The incremental ROAS calculator turns a test result into these numbers.

Spillmark runs geo holdout tests for brands that advertise on Meta and sell on Amazon, and it counts Shopify orders by ship-to state too, so one test shows the lift on each channel. It plans the test from your own order history and tells you before you start if it's too small to see a realistic effect. It locks the design before the test starts, keeps every test on record, and takes off Amazon fees and your product costs to show the profit.