Meta Incremental Attribution: What Changes in Ads Manager and Delivery, and When a Shopify Brand Should Test It

Meta's incremental attribution is an ad set attribution model that optimizes delivery toward, and reports only, the conversions Meta's machine learning predicts an ad caused. It's a modeled estimate with no holdout group: nobody is kept from seeing your ads. Expect fewer reported purchases and a higher cost per result than under standard attribution.

For how attribution models split credit across every channel, see our guide to first click vs last click attribution. Incremental attribution is one of Meta's own models, an alternative to the click, view and engage-through windows in our guide to Meta ads attribution settings. This post covers what Meta says the model does, where you can use it, what changes in your numbers and delivery, how it differs from a Conversion Lift test, and when a Shopify brand should test it. We checked every Meta detail against Meta's help center, newsroom and developer docs on October 24, 2026.

What is Meta's incremental attribution?

An attribution model you choose when you create an ad set. Instead of crediting every conversion inside a window, it credits and optimizes for the ones Meta predicts the ad caused.

Meta's incremental attribution article says the model uses machine learning to predict whether a conversion is caused by an ad. Meta lists two aims: focusing campaigns on outcomes more likely to be driven by an ad, and potentially driving more incremental conversions than your existing campaigns.

Its attribution models article sets it beside the standard model:

---

How does Meta calculate incremental conversions?

Meta publishes the outline only: a machine learning model judges each conversion, and the incremental count is a subset of the conversions Meta already counts.

Here is what Meta's help center says about the calculation, on its page for viewing incremental results:

• The incremental results include only the conversions Meta considers incremental. Standard attribution includes all conversions.
• Meta's example: a campaign with 100 total conversions, 70 of which Meta considers incremental, shows 70 under incremental attribution and 100 under standard.
• The figure is designed to show the relative incrementality between Meta campaigns.
• Results aren't available for date ranges before April 1, 2025.

What those pages don't publish: the model's inputs, what it was trained on, how accurate it is for an account like yours, or whether it's checked against your own lift tests. An older Meta metric called Incremental results was described as a statistical model trained and validated on historical lift test data. Meta's incremental attribution pages don't make that claim for the new model, so don't assume it.

Two limits follow from how it works. It only judges conversions Meta received, so a purchase your pixel or server never sent can't be counted as incremental either. And it only judges Meta's own ads: it can't tell you whether Google, email or word of mouth would have closed the sale instead.

---

Where can you use incremental attribution?

In Sales, Engagement and Leads campaigns that optimize for website conversions, set when you build the ad set.

Meta's setup article gives the requirements:

To set it up: in Ads Manager, click Create, choose Sales, Engagement or Leads, and continue to the ad set. Set the conversion location to Website or Website and app, pick a conversion goal, click Show more options, then under Attribution model open the Options menu and select Incremental. Meta notes that its campaign setup flows are changing, so your screens may differ.

"Bid controls" covers three bid strategies, per Meta's cost and bid controls article: cost per result goal, ROAS goal and bid cap. An incremental campaign runs without any of them.

---

What changes in your Ads Manager numbers?

Results drop to the conversions Meta's model counts as caused by your ads, and every metric built on Results moves with them.

Meta says derived metrics such as cost per result are calculated the same way. Extend Meta's own example with spend and a store's order count:

Nothing about the campaign changed between the two columns. Only the counting did. Three things follow for a Shopify brand:

1. Your targets need resetting. A cost per purchase or ROAS target set under standard attribution will look missed under incremental attribution. Set a separate target, or judge the campaign on store numbers instead.
2. Don't compare across models. Meta recommends comparing within one attribution model, and its attribution article says the Campaign Overview table can't compare ad sets on different models.
3. The gap to Shopify widens. Ads Manager claims fewer of your orders than it did. Those orders still happened: Meta is declining to claim the ones it thinks would have happened anyway.

You can see this before switching anything. In Ads Manager, open Columns: Performance, scroll to Compare attribution models, select Incremental and click Apply. Your standard campaigns then show their incremental count beside the standard one. If a campaign's incremental share is far below its neighbors', that's the one to look at first.

---

What changes in delivery and bidding?

Delivery optimizes for the conversions Meta predicts the ad will cause, and you give up bid controls to get it.

• The optimization target changes. Standard attribution optimizes for conversions inside your windows; incremental attribution optimizes for incremental conversions. Meta doesn't publish how that shifts who sees your ads. Our expectation, not Meta's statement: the model has less reason to bid for shoppers who were already on their way to buying, such as people with a full cart or recent customers.
• No cost or ROAS target. Without bid controls, the campaign can't hold a cost per result or a ROAS goal you set. If your account depends on those to stay profitable, that's a real cost of testing.
• A new campaign, so new learning. You can't switch a running campaign over, so the test starts from a fresh campaign.
• It learns only from purchases Meta receives. If your setup misses orders, the model has fewer to learn from, the same as standard attribution. Our guide to why Meta isn't tracking Shopify purchases covers the fixes.

Is incremental attribution the same as a Conversion Lift test?

No. Incremental attribution predicts which conversions an ad caused. A Conversion Lift test measures it by keeping a random group of people from seeing your ads.

Meta draws the same line. Its March 3, 2026 announcement calls incrementality experiments such as Conversion Lift the best way to learn what an ad caused, and the gold standard in measurement, while noting that fully adopting them takes time.

Both count only conversions Meta receives. A geo test that measures your own Shopify orders by region doesn't depend on that. Our incrementality testing guide covers how to design, size and read either kind of test.

---

When should a Shopify brand test incremental attribution?

When Meta's reported results and your store's numbers disagree enough to change your budget, and you can run one campaign without bid controls.

Good reasons to test it:

• Meta's ROAS in Ads Manager looks much better than your store's revenue and MER suggest.
• A large share of your Meta spend reaches people already in your funnel: retargeting, recent visitors and existing customers.
• The Compare attribution models column shows a big gap between standard and incremental results on your largest campaigns.
• You want Meta to bring in more first-time buyers, and you can measure that with new-customer CPA and ROAS in Ads Manager.

Reasons to wait:

• Your campaigns rely on a cost per result or ROAS goal to stay profitable.
• Your setup is missing purchases. Fix tracking first, or both models learn from the same partial data.
• You need proof that Meta causes sales. That takes a lift test.

How to run a fair test

1. Look before you build. Add the Incremental comparison column to your current campaigns for the last 30 days and note each campaign's incremental share.
2. Build a new campaign on the incremental model. Copy the objective, creative and budget of a standard campaign you keep running beside it.
3. Run full weeks. Give it at least two full weeks before judging, so weekday and weekend buying are both in the data.
4. Judge on numbers outside Meta's model. Compare Shopify revenue, new-customer CPA and MER for the test period, plus your click-based attribution. Don't compare the two campaigns' Results columns; they count different things.
5. Confirm big decisions with a lift test. The two campaigns can reach the same people, so neither is a clean control. If the answer moves real budget, test it with a holdout.

---

Where Upstack fits

Upstack doesn't run incrementality tests or marketing mix models. Upstack Attribution reports first click, last click and any click against the revenue in your Shopify account, rather than summed from what each platform claims. During an incremental attribution test, that gives you the store-side view: whether Shopify revenue and new customers moved while Ads Manager's Results dropped. It doesn't tell you what Meta caused. For that, run a lift test.

Upstack Signal sends purchases server-side to Meta with the customer matched, which matters for both models because each one judges only the purchases Meta receives.

---

Frequently asked questions

What is incremental attribution on Meta?

It's an attribution model in Meta Ads Manager that uses machine learning to predict whether each conversion was caused by an ad. Ad sets that use it optimize delivery for those incremental conversions, and their Results count only those conversions.

How do you turn on incremental attribution in Meta Ads Manager?

Create a Sales, Engagement or Leads campaign. In the ad set, choose Website or Website and app as the conversion location, pick a conversion goal, click Show more options, and select Incremental under Attribution model. You can't use bid controls with it or change the model after publishing.

Why are my results lower with incremental attribution?

Because it counts only the conversions Meta considers incremental, while standard attribution counts every conversion in its windows. In Meta's example, a campaign with 100 conversions shows 70 under incremental attribution, and cost per result rises to match.

Is Meta's incremental attribution the same as a lift test?

No. Incremental attribution is a model's prediction with no holdout group. A Conversion Lift test randomly keeps a control group from seeing your ads and measures the difference. Meta itself calls lift experiments the gold standard in measurement.

Can you see incremental results without switching campaigns?

Yes. In Ads Manager, open Columns: Performance, choose Compare attribution models and select Incremental. Your standard campaigns then show modeled incremental results, for dates from April 1, 2025.

See how Upstack Attribution reports Meta against the revenue in your Shopify account, by first click, last click and any click.