Attribution Modeling | AppsFlyer

Attribution Modeling

Using one of several methods to pinpoint the origins of a non-organic install


What is attribution modeling?


Attribution modeling is the structure through which advertisers are able to attribute and measure the value of different marketing efforts on and across different channels, which later informs budget spend and overall mobile marketing decisions. Because attribution modeling is based on assigning value to a predetermined advertising action taken by users within a given time frame, advertisers can more clearly identify which channels are best performing according to the company’s goals.

In short, attribution modeling is the map both advertisers and ad networks follow in order to measure the user journey and generate revenue from user ad interactions, respectively.  

Because a user is likely to interact with more than one ad on more than one channel, there are several types of attribution modeling:


  • Last touch
  • Multi touch
  • View-through


Last touch attribution, which is currently the attribution modeling standard, occurs when an install or re-engagement is matched to the last interaction, or touch, in the user journey within the attribution window. (An attribution window, also known as a lookback window, is the allowable period of time from the last touch, either a click or view, to the install.) In this case, the ad network which produced the last touch gets the credit and payment. For example, if the cost per install of the user journey below is $2, the media source which last “touched” the user, Network C, would get this full amount.

attribution modeling

Multi-touch attribution, also known as fractional attribution, on the other hand, is a method for identifying various touch points along a user journey that lead to a conversion, whether it’s an install, purchase, or other designated in-app activity. Multi-touch attribution can be single-channel (across one device), omni-channel across different devices (e.g., mobile, desktop, and TV), or even offline. In multi-touch attribution, Networks A and B in the example below would receive credit (and payment) as an “assisted” install, while Network C would be credited with the install.

attribution modeling

Multi-touch or fractional attribution alludes to weighted credit given to media sources that indirectly contributed to the conversion. Since fractional attribution remains largely unused as an active form of attribution modeling, it is considered as a future alternative to current attribution modelling methods due to the granularity of its attribution analysis and crediting process. Ultimately, the idea is that all media sources that touched users along the way (before the last touch) will receive some fraction of the payment.

Note that it is the attribution ecosystem which ultimately dictates which attribution modeling method will be used for measurement and payment. While multi-touch attribution is the broadest model for crediting multiple sources, there are numerous attribution models that follow these guidelines, such as U-shaped/position-based attribution and W-shaped attribution, among others.   


Why is attribution modeling important?


Simply put, attribution modeling allows advertisers to determine how to attribute and measure the performance and value of their media sources for their marketing efforts. Without working within a certain attribution model (in other words, not performing attribution at all), advertisers are unable to see the full picture of their user acquisition and revenue-generating efforts.

Attribution gives granular insights on the who, what, when, where, why, and how of specific media sources, user and ad traffic, quality (in retention and LTV) of active users, and long-term ROAS/ROI, among other metrics. Referring back to multi-touch attribution modeling, the various touch points of a user journey before install, or “assists,” reveal more of the actual story about how and why a user converted and play a major role in future budget allocation decisions. Ultimately, with a wider perspective on the user journey, app marketers can budget more effectively and realistically for each media source that contributed to an install.

However, not only does attribution modeling support marketing efforts on the advertiser side, but it also ensures accurate and fair install crediting and payment on the network side. Having an unbiased, 3rd party attribution provider is key for establishing trustworthy transaction-based attribution reporting system and giving both credit and responsibility to networks where due.

In short, attribution modeling is the structure in which attribution for mobile installs can take place, leaving the mobile ecosystem, for both advertisers and media sources alike, balanced and dynamic in the long run.

« Back to Glossary Index