Meta Ads Attribution Models: The Complete Guide 2026

Without CAPI, an account loses between 25 and 30% of its conversion data before any analysis even begins. And that’s before we talk about windows, models, or third-party tools. Attribution in Meta Ads isn’t a secondary configuration matter — it’s the filter that determines which budget decisions look correct and which don’t. If your knowledge of attribution predates 2022, this guide was written for you. The Facebook Attribution tool was retired in August 2021. The 28-day windows disappeared after iOS 14.5. Meta launched incremental attribution in 2025, redefined what counts as a “click” in March 2026, and introduced Predictive Budget Allocation. All of these changes directly affect how you interpret Meta Ads ROAS and what decisions you make with that data.

Key takeaways

Without CAPI, you lose between 25 and 30% of your conversion data (wetracked.io, 2026)

The default attribution window in 2026 is 7-day click + 1-day view; the 7-day view window was removed in January 2026

75% of iOS users opted out of tracking, making CAPI a technical requirement, not an option (AdAmigo.ai, 2026)

Since March 2026, only link clicks count for click-through attribution

Incremental attribution reveals the true prospecting ROAS is 1.90x, not 8x (Cassandra App, 2025)

How Meta Ads assigns conversions
Attribution windows in 2026
The iOS 14.5 impact on attribution
The end of Facebook Attribution
Click-through vs. engage-through: the 2026 changes
Incremental attribution: Meta’s new model
Recommended practical setup
Meta Ads vs. third-party attribution tools
Frequently asked questions
Sources

How Meta Ads assigns conversions

Summary: Meta Ads attribution combines a time window, action type, and credit model to decide which conversions are assigned to which ads. The window choice doesn’t just affect reporting — it shapes how the algorithm optimises delivery. Without complete conversion data (CAPI + Pixel), any attribution setup starts from an incomplete foundation. Source: Meta Business Help Center, 2026.

Meta Ads analytics dashboard showing conversion paths and campaign performance metrics

Attribution in Meta Ads directly determines which campaigns look profitable and which don’t. Understanding how it works is the prerequisite for any budget decision based on real data. According to the Meta Business Help Center, Meta’s attribution system relies on three elements working together.

To assign conversions, Meta combines three pieces:

The attribution window: the number of days between a user seeing or clicking an ad and completing the desired action.
The attribution action: whether the conversion is counted after a click or after a view.
The attribution model: how credit is distributed across different touchpoints.

Why does the attribution window matter so much? It’s not just a reporting preference. It determines how Meta’s algorithm chooses who sees your ads. A 1-day window makes the algorithm optimise for fast converters; a 7-day window lets it capture longer decision cycles — which makes a real difference in B2B or high-ticket products.

Understanding these models is also the prerequisite for correctly interpreting Meta Ads ROAS and avoiding decisions based on inflated figures.

For any attribution setup to be reliable, you first need complete data reaching Meta. The tool that makes this possible is the Conversions API (CAPI): without it, a meaningful share of your conversions simply doesn’t exist in Meta’s system. The rest of this guide assumes you have CAPI implemented or are actively considering it.

Attribution windows in 2026

The attribution windows available in Meta Ads Manager have changed significantly since 2021. On January 12, 2026, Meta permanently removed the 7-day view and 28-day view windows from the Insights API. For advertisers using the 7-day view window, this caused a 15–30% drop in attributed conversions overnight, according to Dataslayer. Those relying on the 28-day window lost between 30 and 40% of their attributed conversions.

Window	Description	Status
1 day after click	Conversion within 24h of clicking the ad link	Active
7 days after click	Conversion within 7 days of clicking the ad link (default)	Active
1 day after view	Conversion within 24h of viewing the ad without clicking	Active
1-day video engagement	For videos: conversion after watching ≥5 seconds, within 24h	Active
7-day view	Conversion within 7 days of a view	Removed January 2026
28-day click / 28-day view	Long post-click and post-view windows	Removed 2021

Default window in 2026: 7 days after click + 1 day after view.

Why the window affects delivery, not just reporting

A 1-day window makes the algorithm optimise for fast converters — people who decide within hours. A 7-day window lets it capture longer decision cycles.

For most businesses, the 7-day click + 1-day view setup remains the right balance. For ecommerce with impulse-buy products, 1-day click gives cleaner, less noisy data.

The iOS 14.5 impact on attribution

Summary: iOS 14.5 wiped out 75% of the tracking signal on Apple devices. Without CAPI, an account loses 25–30% of conversion data. The technical solution is Pixel + CAPI, which raises the match rate from 60–70% to 85–95% and Event Match Quality from 3.5–5.0 to 7.5–9.0. Sources: AdAmigo.ai (2026), wetracked.io (2026).

Data server with active connections representing CAPI server-side tracking versus browser-based pixel

In April 2021, Apple launched iOS 14.5 with the App Tracking Transparency (ATT) framework. The effect was structural: 75% of iOS users opted out of tracking, according to AdAmigo.ai. The IDFA (iOS device identifier) dropped below 6% availability, eliminating the primary cross-app attribution mechanism.

The practical consequences:

28-day windows disappeared; Meta moved to a practical maximum of 7 days.
Reported conversions became partly modelled: Meta uses statistical estimation for conversions it can’t directly measure.
Advertisers without CAPI lose 25–30% of events before the data ever reaches Meta.

CAPI and Event Match Quality

The correct technical response is the Meta Conversions API (CAPI). Unlike the Meta Pixel, which operates from the user’s browser, CAPI sends data directly from your server to Meta. No device dependency. No reliance on browser tracking consent.

In accounts I’ve migrated from Pixel-only to Pixel + CAPI, recovered events consistently land between 15 and 25%, with some high-mobile-traffic ecommerce cases reaching 30%. The match rate rises from 60–70% (Pixel only) to 85–95% (Pixel + CAPI), according to wetracked.io.

But the metric I always check first in an audit is Event Match Quality (EMQ): a 0–10 score Meta assigns to each event based on the quality of user identifiers you send. With Pixel only, accounts typically see EMQ of 3.5–5.0. With Pixel + CAPI properly set up, EMQ rises to 7.5–9.0, according to adsuploader.com. The improvement in algorithm optimisation quality shows within 7–10 days.

Sources: wetracked.io (2026), adsuploader.com (2026)

No CAPI means every attribution conversation starts the same way: what you see in Ads Manager isn’t what happened. It’s an estimate.

Does it make sense to analyse attribution windows before fixing this? No. CAPI is the first technical fix you should tackle before trying to interpret your campaign attribution.

The end of Facebook Attribution

The Facebook Attribution tool was permanently retired in August 2021, according to the Meta Business Help Center. It was the standalone panel inside Business Manager that let you compare multi-touch models (last click, linear, positional, data-driven) and analyse cross-channel conversion paths. Meta had stopped onboarding new advertisers from May 1, 2021.

The reason was explicit: iOS 14.5 restrictions had made reliable cross-publisher tracking — the foundation the tool relied on — impossible.

What replaced Facebook Attribution:

In 2026, conversion measurement in Meta rests on three pillars:

Meta Pixel: browser-based tracking, subject to iOS limitations.
Conversions API (CAPI): server-side tracking, independent of the device or browser.
Aggregated Event Measurement (AEM): a privacy system that models conversions on iOS 14.5+ using aggregated, anonymised data.

No native Meta tool currently allows you to compare multi-touch attribution models across channels. For that analysis, you need third-party tools — covered below.

Click-through vs. engage-through: the 2026 changes

Summary: Since March 2026, only link clicks count for click-through attribution. Social interactions (likes, shares, saves) are reclassified under the new “engage-through” category, with a 1-day conversion window. 2025 and 2026 data are not directly comparable because of this change. Sources: Search Engine Land, Jon Loomer Digital (2026).

In March 2026, Meta published a significant restructuring of its attribution model under the headline “Simplifying ads measurement for a social-first world”. The changes directly affect how conversions are counted, according to Search Engine Land.

1. Click-through attribution now only counts link clicks

Before 2026, any interaction with an ad — a like, share, save, or comment — could count as a “click” for attribution purposes. This artificially inflated attributed conversions. From March 2026, only clicks on the ad link count for click-through attribution.

2. “Engage-through” replaces “engaged-view”

A new category called engage-through captures:

Social interactions (likes, shares, saves, comments)
Non-link clicks
For videos: watching at least 5 seconds (threshold dropped from 10 to 5 seconds)

If someone takes any of these actions and converts within 1 day, the conversion is recorded as engage-through, not click-through.

Impact on your data and year-over-year comparisons

This change causes a visible drop in reported click-through conversions. Your campaign isn’t performing worse. Meta simply no longer attributes conversions to interactions that weren’t real clicks.

One important note: 2025 and 2026 data are not directly comparable. If you’re comparing year-over-year performance, you’ll see drops that don’t reflect a real deterioration. Before cutting budget or pausing campaigns based on a year-over-year decline, check whether the drop coincides with the March 2026 transition period, according to Jon Loomer Digital.

Meta also announced partnerships with Northbeam and Triple Whale to incorporate Meta view-through data into their external attribution models, narrowing the historical gap between Meta’s data and third-party platforms.

Incremental attribution: Meta’s new model

Summary: Meta’s incremental attribution, available since April 2025, uses randomised holdout groups to measure conversions caused by the ad, not just those that followed it. An analysis of 792 marketing mix models reveals the true incremental ROAS is 1.90x (prospecting) and 3.64x (retargeting), compared to the 8x Meta reports in its platform. Source: Cassandra App (2025).

Comparative analysis chart showing prospecting versus retargeting campaigns measuring the true impact of Meta ads

In April 2025, Meta launched incremental attribution as an optional model inside Ads Manager, according to Logical Position. It’s the most honest tool Meta has given advertisers in years — precisely because it can show results that are less flattering than standard attribution.

Unlike last-click attribution, which assigns all credit to the final touchpoint, incremental attribution uses randomised lift tests: Meta creates a holdout group that doesn’t see the ad and compares their conversions with the exposed group. The difference is the true incremental conversion.

I’ve seen retargeting campaigns showing an 8x ROAS in standard attribution drop to 2x in incremental. That gap doesn’t mean retargeting is worthless — it means most of those conversions would have happened anyway. Ignoring that number leads to overspending on retargeting and underinvesting in prospecting.

The 2025 aggregate data confirms this pattern: Meta reports an average ROAS of 8x, but incremental measurement reveals 1.90x for prospecting and 3.64x for retargeting, according to Cassandra App’s analysis of 792 marketing mix models. Retargeting is incrementally more efficient than prospecting, but both sit far below the figure in the dashboard.

Source: Cassandra App (2025) — analysis of 792 Marketing Mix Models, 59 prospecting and 72 retargeting models

Why does this matter in practice? Last-click attribution overstates retargeting’s impact and understates prospecting’s. If you run incremental attribution and see low incrementality in retargeting, that doesn’t mean it’s useless — it means you can’t hold it to the same ROAS target as prospecting.

Recommended practical setup

The right attribution setup isn’t universal — it depends on your purchase cycle and product type. These are the recommendations by business profile, based on Jetfuel Agency’s configuration guide.

For most advertisers

Stick with the default: 7 days after click + 1 day after view. It’s the balance Meta recommends for most objectives.

If you sell with short decision cycles (eCommerce)

Consider using 1 day after click for more conservative data. It helps you evaluate true ROAS without the noise of conversions that would have happened regardless. For Advantage+, note that the conversion threshold dropped from 50 to 25 per week in 2026 — making it easier to exit the learning phase on lower budgets. For a full breakdown of how Meta campaigns are structured for ecommerce, see the Meta Ads for ecommerce guide.

If you have long purchase cycles (B2B, professional services)

The 7-day window may not capture the full journey. Supplement with a third-party attribution tool that can follow the path beyond what Meta can see.

Always implement CAPI and monitor EMQ

Without the Conversions API, any attribution analysis is incomplete. CAPI is the prerequisite for any reliable setup. Once it’s running, keep an eye on Event Match Quality in Events Manager: it should be above 6, ideally close to 8.

Predictive Budget Allocation: how it affects attribution

In 2026, Meta introduced Predictive Budget Allocation: the system dynamically shifts spend across ad sets based on real-time conversion probability, rather than distributing budget evenly. Attribution windows now affect not just reporting, but how the algorithm redistributes budget in each auction cycle. A shorter window speeds up the system’s response; a longer one provides more historical signal for its decisions.

Evaluate retargeting with incrementality

Retargeting campaigns tend to show low incrementality: most of those conversions would have happened regardless. That doesn’t mean retargeting is worthless — it means you need to evaluate it with different metrics than prospecting.

Meta Ads vs. third-party attribution tools

Meta, like every advertising platform, grades its own homework. This creates a permanent discrepancy with Google Analytics, GA4, and third-party tools. The March 2026 changes (click-through = link clicks only) were explicitly designed to narrow this gap and align Meta’s data with how GA4 counts sessions and conversions.

Even so, Meta’s native attribution is a walled garden: it can’t see what happens in other channels (Google, email, organic). For a complete picture, the most widely used tools in 2026 are:

Tool	Methodology	Ideal profile
Triple Whale	Click-based + proprietary pixel	Shopify eCommerce $5M–$50M
Northbeam	Probabilistic ML, multi-touch credit	Complex multi-channel, from ~$1,000/month
Measured	Lift tests (true causality)	Enterprise, rigorous optimisation

When is a third-party tool worth the investment? When your channel mix is complex enough that Meta can’t see the full picture alone. Meta measures within its own ecosystem. External tools aggregate data from every channel to give a complete view of ROI. Before investing in third-party tools, ensure your Meta setup is solid — the Meta learning phase guide explains how the algorithm needs stable conversion data to exit learning and deliver reliable results.

Architecture note: To understand why Meta measures the way it does, the article on Andromeda Meta Ads explains the AI architecture behind the attribution system and how the Andromeda algorithm integrates conversion signals into delivery optimisation.

Want to review your Meta Ads attribution setup? I work directly in your account as a freelance Meta Ads consultant — no intermediaries, measurable results.

If you want to know whether your current windows accurately reflect your campaigns’ true impact, book a free audit.

Frequently asked questions

What’s the recommended attribution window for ecommerce in 2026?

For ecommerce with short purchase cycles, the recommendation is 1 day after click as the primary window. It gives more conservative data and reduces the noise from conversions that would have happened without the ad. The default 7-day click + 1-day view window is right for higher-ticket products or longer decision cycles, according to Jetfuel Agency’s guide.

Why don’t my Meta conversions match Google Analytics?

The gap between Meta and GA4 is structural. Meta counts conversions within its attribution windows (including view-through), while GA4 only records sessions with a click. The March 2026 changes, which limit click-through attribution to real link clicks, partially close this gap, according to Search Engine Land. Even so, both tools are measuring different realities and will never fully align.

What is Event Match Quality and how does it affect attribution?

Event Match Quality (EMQ) is a 0–10 score Meta assigns to your events based on the quality of user identifiers you send. A low EMQ (3–5) means Meta can’t reliably identify who a conversion belongs to, which hurts both attribution accuracy and algorithm optimisation. With Pixel only, typical EMQ is 3.5–5.0; with Pixel + CAPI properly set up, it rises to 7.5–9.0. You can monitor it in Events Manager; it updates every 48 hours, according to adsuploader.com.

What is incremental attribution and how do I activate it?

Meta’s incremental attribution uses randomised control groups to identify conversions directly caused by the ad, not just those that occurred after seeing it. It’s been available since April 2025 as an optional model in Ads Manager: go to campaign settings and select “Incremental attribution” under measurement options. Initial results showed improvements of more than 20% in true measurement efficiency (Logical Position, 2025).

How does incremental attribution differ between prospecting and retargeting?

The difference is significant. Prospecting campaigns tend to show higher incrementality because they reach users with no prior brand exposure who wouldn’t have converted without the ad. Retargeting campaigns show lower incrementality because they reach people who already had purchase intent. 2025 data shows an incremental ROAS of 1.90x for prospecting and 3.64x for retargeting, versus the 8x Meta reports for both (Cassandra App, 2025).

How do I compare 2025 data with 2026 when attribution windows changed?

Carefully. The removal of the 7-day view window (January 2026) and the click-through changes (March 2026) make 2025 and 2026 conversion numbers not directly comparable. Before concluding a campaign is underperforming in 2026, check whether the drop coincides with either of those changes. The cleanest way to establish a valid baseline is to use incremental attribution as your reference, since it measures causal conversions regardless of definition changes.

What exactly changed in attribution in March 2026?

In March 2026, Meta limited click-through attribution to real link clicks on the ad. Previously, any social interaction (like, share, save) could count as a “click” for attribution. It also created the “engage-through” category for conversions from social interactions or video views of at least 5 seconds. The result is a reduction in reported click-through conversions without any real change in actual performance, according to Jon Loomer Digital.

Does CAPI fully resolve post-iOS 14.5 attribution issues?

CAPI significantly improves event coverage, recovering 15–30% of lost events, but it’s not a complete solution. Even with CAPI in place, iOS conversions are still partly modelled through Aggregated Event Measurement (AEM). CAPI removes the browser dependency, but doesn’t eliminate ATT’s impact on Meta’s ability to track individual iOS users, according to the Meta Business Help Center.