In the framework of app install campaigns in the days of iOS 14, mobile attribution is one of the biggest issues marketers need to face.
But let us take one thing at a time.
What is attribution? Basically, mobile attribution (also known as multi-touch attribution) is how marketers understand the journey you take to arrive in their app and what you do once you’ve landed there. In order to carry out an install and post-install event, a user goes through clicking on various advertisements and eventually arrives at a conversion (i.e. sale). For instance, they google something, click on an advertisement for a product they have googled, they create an account with their email and then eventually purchase.
The multi-touch system is very articulated and needs a complex attribution technology which is able to identify the points of contact with the product that were made before the purchase and which media sources or ad network generated them.
There are two types of attribution: deterministic attribution and probabilistic attribution. Deterministic attribution involves no randomness and is possible in the cases in which the IDFA is still available. With deterministic attribution you know for sure which specific app install campaign drove a particular conversion. When an app install is deterministically attributed to a campaign, there’s a 100% probability that a specific campaign drove an app install.
When there are no more cookies and now that, with iOS 14, Apple introduced the ATT prompt which allows people to hide their IDFA, the vast majority of iOS users choose not to be tracked. Talking about app install campaigns, within this framework, it is necessary to estimate with a good accuracy which have been the contact points between a user and a product.
In particular, what we need to estimate is which one is the last of these contact points. And this is exactly where probabilistic attribution is used: in apps as an alternative to IDFA and SKAdNetwork, on websites as an alternative to cookies. Probabilistic attribution is a type of mobile attribution that relies on probabilities and not on IDFA, therefore, it is fully aligned with Apple’s new changes that came with the introduction of the ATT prompt in iOS 14. It is based on patterns and uses machine learning in order to identify probable conversions, so, in the context of an app install campaign, there is a set of probabilities that one or more campaigns drove an install.
The advantages of probabilistic attribution are:
- It is fast;
- It shows trends;
- It allows you to act quickly when setting up a campaign.
The disadvantages of probabilistic attribution are:
- It is estimated data;
- It is not precise;
- It has potential risks of not protecting the user’s privacy.
Now, let’s look at a very simplistic example of how probabilistic attribution works. First thing is collecting deterministic data across as many marketing touch points as possible (this includes impressions and clicks on digital ads, exposure to broadcast radio and television spots, podcast and OTT ads). Next, using attribution, build a probability map of these touch points leading to conversions, or customer acquisition events. While not 100% accurate, these probabilities can be used when deterministic tracking is not available.
Probabilistic Attribution And Fingerprinting
Before probabilistic attribution, a similar but less privacy-friendly technology was used, fingerprinting. It used more information than probabilistic, it was more accurate, but it tracked the users uniquely and without telling the users that they were being tracked. It relies on a “fingerprint” of a particular device which is basically made of information like mobile or computer hardware, operating system, IP address, web browser etc. There are two main kinds of fingerprinting: device fingerprinting and browser fingerprinting. In device fingerprinting information is collected by the apps a user installs on their mobile device; in the browser is collected by the browser utilized by the user. It is important to notice that with iOS 14 fingerprinting has a very limited space.
Briefly, things are changing fast, and companies need to keep up with the times. In the context of app install campaigns, in the iOS 14-ruled world where deterministic attribution is going to be significantly impacted, probabilistic attribution is key.