Incremental Attribution on Meta: What It Measures and Where It Falls Short

In April 2025, Meta launched a new feature in Ads Manager: Incremental Attribution. It is an AI-driven measurement model that identifies conversions directly caused by ads, filtering out organic sales that would have happened without any ad exposure. For DTC brands trying to understand true Meta performance, it is the most significant measurement update Meta has released in years.

But Meta's Incremental Attribution has a structural ceiling. It only measures what Meta can see. For omnichannel brands selling across Amazon, TikTok Shop, and retail, that ceiling matters.

What Meta's Incremental Attribution actually is

Standard attribution models on Meta — last-click, view-through, data-driven — assign credit to ads that appeared in a customer's journey before a conversion. They measure correlation, not causation. A customer who was going to repurchase anyway and happened to see a retargeting ad still gets credited as an attributed conversion.

Meta's Incremental Attribution takes a different approach. It uses holdout testing, powered by Meta's machine learning infrastructure, to estimate how many conversions would have happened without ad exposure. The feature runs within Ads Manager and does not require a separate lift study to be set up: Meta handles the test design automatically.

The result is a more accurate read on which conversions your ads actually caused, and which would have occurred organically. Incremental conversions meaning, in this context, is precise: purchases that required the ad to happen.

How Meta's Incremental Attribution works

The methodology is based on holdout testing. A portion of the target audience is withheld from seeing the ad (the control group), while the rest is exposed to it (the test group). Meta compares conversion rates between the two groups and uses that difference to calculate incremental lift.

This process is powered by Meta's Andromeda algorithm, which applies machine learning across Lift study data to optimize ad delivery toward users most likely to generate incremental conversions, rather than users most likely to convert regardless. The model is designed to reduce inflated performance figures and improve campaign quality over time.

How to enable it: optimization and measurement are separate

Meta's Incremental Attribution operates in two distinct modes. Both are available in Ads Manager.

Optimization mode adjusts how Meta delivers your ads, targeting users where incremental impact is most likely. To enable it: create a campaign with an objective like Sales or Product Catalog Sales, click "Show more options" under the performance goal section, and select Incremental Attribution. Meta's model takes over from there.

Measurement mode adds incremental attribution data to your reporting columns without changing how your ads are delivered. To access it: click Columns → Compare Attribution Settings → Incremental Attribution. This lets you see incremental conversions alongside your standard attribution windows without running a full lift test.

How incremental attribution compares to standard Meta attribution windows

One useful way to understand the gap between standard attribution and incrementality is to compare them directly. Published analysis of millions in April 2025 Meta spend found the following:

• 7-day click, 1-day view (7DC1DV): 37.7% of attributed purchases were not incremental

• 7-day click (7DC): 12% of attributed purchases were not incremental

• 1-day click (1DC): 9.4% of incremental purchases were not attributed at all

The 7DC window is a reasonably close proxy for incrementality, because a customer who clicks an ad and purchases within seven days is more likely to have been genuinely influenced by the ad. The 1-day view window overcounts the most, capturing purchases from customers who simply saw an ad in passing before buying something they had already decided to buy.

These numbers will vary by brand, audience mix, and product type. But directionally, they illustrate why standard attribution figures and true incremental figures diverge.

The structural limitation: Meta's tools only see Meta-attributed sales

Meta's Incremental Attribution is a meaningful improvement over standard attribution, but it shares one constraint with all within-platform measurement: it can only observe conversions that touch Meta's tracking infrastructure.

That means it captures purchases on Shopify or your DTC site (via the Meta Pixel or Conversions API). It does not observe purchases that happen on Amazon after a customer sees a Meta ad. It does not capture sales at retail. It does not track the lift in branded search that Meta campaigns often generate. These omissions are not a flaw in Meta's methodology; they are a data access constraint. Meta can only measure what it can see.

For brands where DTC is the only meaningful sales channel, this has limited impact. For omnichannel DTC brands, it can be significant.

Why the halo effect is invisible to Meta's measurement

When a consumer sees a Meta ad and searches for your brand on Amazon a few days later, that purchase is classified as Amazon organic. It does not appear in Meta Ads Manager. It does not appear in Conversion Lift results. It does not appear in Meta's Incremental Attribution reporting. The causal chain from ad to purchase is real; it is simply outside the scope of what any Meta tool can measure.

Third-party geo incrementality testing runs across all channels simultaneously. Ads are paused or held dark in control geographies, while test geographies continue to run normally. Revenue is compared across Shopify, Amazon, retail, and all other connected platforms. The difference is total incremental revenue, regardless of where the order was placed.

The Graza team ran this type of cross-platform test and found that Meta's true causal impact was 2.7× higher than platform-reported data. The gap was not overcounting within Meta — it was the halo effect on other channels that Meta's tools structurally cannot capture. After scaling Meta spend roughly 60%, the brand saw 106% more Meta sales, a 9% POAS improvement, and a 20% MER improvement.

That is the difference between incrementality-adjusted attribution and within-platform measurement: scope. Meta's Incremental Attribution is causal but narrow. Cross-platform geo lift tests are causal and complete.

Should you use Meta's Incremental Attribution, third-party testing, or both?

Meta's Incremental Attribution is worth enabling, particularly the measurement column. It is free, runs automatically, and gives a materially more accurate read on Meta performance than standard attribution windows. For brands where DTC revenue is the primary goal, it provides a solid causal baseline.

The case for third-party cross-platform testing grows with the share of off-platform revenue. If Amazon, TikTok Shop, or retail account for a meaningful portion of your business, Meta's tools will undercount Meta's contribution to those channels. That undercount leads to under-investment in campaigns that are generating cross-channel demand you can't see in Ads Manager.

The most complete measurement stack runs third-party incrementality testing as the primary causal layer, then uses those results to calibrate day-to-day attribution. Meta's own tools play a useful complementary role for in-platform optimization and rapid campaign-level reads.

Frequently asked questions

What is incrementality in Meta?

Incrementality in Meta refers to the measurement of conversions that were directly caused by Meta ads, as opposed to conversions that would have occurred organically without any ad exposure. Meta's Incremental Attribution feature, launched in April 2025, uses holdout testing and machine learning to estimate this figure automatically within Ads Manager. It is available both as an optimization setting that adjusts ad delivery and as a measurement column in reporting.

What is the difference between standard and incremental attribution?

Standard attribution on Meta (last-click, view-through, data-driven) assigns credit to ads that appeared in a customer's journey before a conversion. It measures correlation: the ad was present, and a purchase followed. Incremental attribution measures causation: whether the purchase would have happened without the ad. A customer who was going to repurchase anyway and clicked a retargeting ad is an attributed conversion but not an incremental one. Analysis of Meta spend suggests the 7-day click window overcounts by roughly 12% on average, and the 7-day click plus 1-day view window by as much as 37.7%.

What is incremental attribution?

Incremental attribution is a measurement approach that identifies the revenue or conversions directly caused by a specific ad or campaign, above what would have occurred without it. It requires a controlled comparison — typically a holdout group that does not receive ads — to isolate the causal contribution. Unlike standard attribution models that assign credit based on ad touchpoints, incremental attribution establishes whether the ad was actually necessary for the conversion to happen.

What does incremental conversions mean?

Incremental conversions are purchases or actions that occurred specifically because of an ad, above the number that would have happened organically. In Meta's Ads Manager, incremental conversions are calculated by comparing conversion rates between users who were exposed to an ad (test group) and those who were not (control group). A high number of total attributed conversions with low incremental conversions indicates that a large share of those purchases would have happened regardless of ad exposure.