Attribution Models

Modeled Conversions

Statistical estimates of conversions that happened but couldn't be directly measured due to privacy restrictions.

Modeled Conversions use machine learning to estimate conversions that tracking can't observe. When user blocks tracking (ad blocker, iOS ATT opt-out, cookie rejection), platforms can't directly measure conversion. Instead, they model it: 'Based on similar users who we CAN track, we estimate this user converted'. Google and Meta both use modeling. How it works: Platform identifies untrackable user → Finds similar trackable users (same demographics, behavior) → Applies their conversion rate → Adds modeled conversion to reporting. Accuracy: Varies, typically 60-80% accurate. Controversy: Some advertisers distrust modeled data ('fake conversions'), others accept it as best available estimate. Reality: Without modeling, attribution loss would be 40-60%. With modeling, it's 15-25%. Modeled conversions are estimates, not ground truth, but better than nothing.

Related Terms

Frequently Asked Questions

What are Modeled Conversions?

Modeled Conversions are statistical estimates of conversions that occurred but could not be directly measured due to privacy restrictions, such as ad blockers, iOS App Tracking Transparency (ATT) opt-outs, and cookie consent rejections. These estimates are generated using machine learning algorithms that analyze the behavior of trackable users to predict the conversion behavior of unobserved users. This process is crucial for bridging the gap in attribution data, which is often referred to as attribution loss. By using modeling, platforms like Google and Meta can provide a more complete, though not perfectly accurate, view of campaign performance, helping advertisers make better optimization decisions in a privacy-first world.

How do marketing platforms use Modeled Conversions to improve ad performance?

Marketing platforms use Modeled Conversions to ensure their automated bidding and optimization algorithms have a more complete dataset to work with. When a platform's tracking pixel cannot fire due to user privacy settings, the platform's machine learning model steps in. It identifies the untrackable user and finds similar users who were successfully tracked, applying their conversion rate to estimate the unobserved conversion. This process significantly reduces attribution loss, which could otherwise be as high as 40-60%. By incorporating these estimates, the platform's algorithms can more accurately assess the true return on ad spend (ROAS) for different campaigns, leading to more effective budget allocation and better overall campaign performance.

Why are Modeled Conversions important for modern attribution and what is the controversy surrounding them?

Modeled Conversions are important because they are a necessary response to the widespread signal loss caused by privacy regulations and browser restrictions, which make direct, user-level tracking increasingly difficult. Without them, advertisers would face massive attribution gaps, leading to poor optimization and budget misallocation. The controversy stems from the fact that modeled data is an estimate, not a ground truth measurement. Some advertisers view them as 'fake conversions' and distrust the data, especially when it inflates platform-reported ROAS. However, the reality is that while they are not 100% accurate (typically 60-80%), they provide a far better basis for decision-making than relying solely on the significantly incomplete data from direct observation alone.

Want accurate attribution without the complexity?

Causality Engine automates attribution reconciliation and provides real-time insights for Shopify brands.

Join Waitlist →