Google uses modeling to estimate online conversions that can’t be observed directly. Modeling allows for accurate conversion attribution without identifying users (for example, due to user privacy, technical limitations, or when users move between devices). Including modeled conversions allows Google to offer more accurate reporting, optimize advertising campaigns, and improve automated bidding.
How modeled conversions work
Google’s models look for trends between conversions that were directly observed and those that weren’t. For example, if conversions attributed on one browser are similar to unattributed conversions from another browser, the machine learning model will predict overall attribution. Based on this prediction, reported conversions are then updated to include both modeled and observed conversions.
Google's conversion modeling approach
Check for accuracy and communicate changes
Holdback validation (a machine learning best practice) maintains the accuracy of Google’s models. A portion of observed conversions (validation data) are held back and split. Then, validation data that was run through the model is compared with validation data that wasn’t. The validation results are used to check for inaccuracy and to further tune the model. Google will communicate modeling changes that might have a large impact on your data.
Maintain rigorous reporting
Modeled conversions are only included when there is high confidence of quality. If there isn’t enough traffic to inform the model, then modeled conversions will not be attributed to ad interactions (or, in the case of Google Analytics, are attributed to the "Direct" channel). This approach allows Google to recover loss of observability while also preventing over-prediction.
Customized for your business
Google’s more general modeling algorithm is separately applied to your data to reflect your unique business and customer behavior.
Don’t identify individual users
Google doesn’t allow fingerprints or other attempts to identify individual users. Instead, Google aggregates data (such as historical conversion rates, device type, time of day, geo, etc.) to predict the likelihood of conversions from a specific ad interaction.