Multi-device attribution is, without a doubt, the greatest struggle of the digital marketing industry. Digital marketers want to know what paths customers take before making an online purchase, but with the wide variety of devices customers are now using, these paths are often hard to track down. For instance, if a customer starts exploring a product from her smartphone, but ends up later purchasing that product from her desktop PC, our current attribution models would only credit the PC with driving this purchase. In reality, the smartphone should have gotten most of the credit, as that was the device which introduced the product to her in the first place.
According to a 2014 Facebook study, 53% of people who have 2 devices switch between them to complete tasks, while 77% of people who have 3 devices do the same. Among all the devices covered in the Facebook study, 22% ended their tasks with a tablet, and 58% with a laptop. Although consumers are using all of these devices, our current attribution models make it hard to figure out the impact of each device in driving purchases.
So what is the solution to this conundrum? How can we track users across all of their devices?
We discovered multiple opportunities to solve this puzzle of how to follow a user across multiple device environments. Here’s what we recommend to crack the code of multi-device attribution:
Householding: Connect multiple devices based on the IP range network. You can consider users who belong to the same IP network, based on their usage behaviour, as a single user.
Deterministic Approach: Connect multiple devices based on a user id. The accuracy for this approach is around 95%! This is when users are asked to sign into social media accounts before they can access services (and one of the reasons why Facebook is better off in multi device attribution compared to Google). In this approach, your users are identifying themselves for you, as long as they are logged into these accounts on each device.
Probabilistic Model: This approach is purely based on predictive algorithms. Many programs (including Google programs) analyze multiple elements such as cookies, browsing patterns and timestamps to make predictions about users who have explored purchases through multiple devices.
While we aren’t yet able to completely track our multi-device users, using these strategies can help us start to understand them, and intelligently navigate our multi-device world.