ML attribution

ML attribution is a method of probabilistic attribution of app installs in ad campaigns targeted for iOS 14.5+.

Previously, for campaigns targeted on iOS below 14.5, the IDFA method was used. Starting with iOS 14.5, Apple has changed its privacy policy and how ad traffic is attributed (for more information on campaigns targeting iOS 14.5+, see this article). Due to these changes, an alternative way of attributing ad traffic was required.

ML attribution is a new way to get conversion data for iOS 14.5+ campaigns. This method works on aggregated and anonymized device data and can be as accurate as 90%.

Thanks to ML attribution technology, the purchase of mobile advertising for the iOS audience in myTarget when working through myTracker allows you to get campaign analytics with the lowest possible delays in receiving data, conduct the most accurate post-install campaign analytics in the context of the usual metrics: purchases, subscriptions, and others custom events.

Before creating an ML attribution campaign, you need to complete the following steps:

1. Integrate the myTracker SDK into the application and then set up the myTarget integration in the myTracker interface. Instructions
2. Connect the application to the myTarget account. Instructions
3. Get your myTracker tracking link. This link will need to be used when creating a campaign. You can add additional parameters to the link, they are described in the article

Then create your campaign as usual. When creating a campaign, open the "Detailed settings for mobile devices" section and enable the "Use ML-attribution" option.

Pay attention to the peculiarities of probabilistic ML attribution:

• Using is possible only with myTracker mobile tracker.
• Possible delay with simulation and data display up to 1 hour.
• In the early days of the campaign, there may be reduced accuracy due to the small amount of accumulated data.
• It will take 2 to 4 weeks to mark up the audience and show full statistics in myTracker, during this period the attribution accuracy may be reduced.
• It is not possible to set up retargeting or create Look-alike audiences. This is because the system does not use any identifiers or precise data about the device.