Tracking Users Across Devices


In this clip, Nitin Rabadia, Marin Software's Director of Audience Marketing, answers a question on how to best track users by device and outlines the two key methods of doing this.

This video is a clip from "Global Mobile Trends - How to reach, influence & convert in-the-moment customers" - watch the webcast on-demand.

Q) How can I link or track users across devices? That's a pretty big question! Is that something that you can answer for us?

A) Yeah... so this is a pretty difficult question - I suppose when it comes to devices, what we're finding is that there's two general ways that you can do this & how you do it will really depend on who you're working with. So, we have what's called probabilstic device matching and we have deterministic device matching - and people listening in may have heard these two phrases before. Probabilistic essentially uses algorithms and data such as IP addresses, user IDs etc. to try and build probabilistic matches for devices. It tries to say, well, if these 4 devices are connecting through the same IP at this point, maybe - or potentially - they're connected to a same user but they overlap that with hundreds of other data points, too, to gain an accuracy score of the likelihood of these devices matching and then allow you to target. This gives you some scale because you're looking pretty much at all devices and all users, however the accuracy isn't quite 100%.

The second way you do that is through deterministic data - and deterministic data essentially involves normally big publishers, who have users who login to multiple devices, so if you think about popular social networks where you might access that through your mobile, potentially through an app, or through some website devices and you're almost always logged in, those companies - again, usually publishers - will hold cross-device data where they can actually target across-devices.

So, what do you do as an advertiser? Do you try and build a probabilistic graph, or do you use the strong deterministic data with some of your publishers? That's really going to depend on where you're advertising and it's going to depend on how many users you can see yourself. So, if you're a medium to large advertiser, typically it might be worth starting to build up your own probabilistic data, or your own deterministic data using your own users, but today, for most advertisers, the best way to do this (the most accessible way), is to use these publishers who are allowing you to target across devices on their particular network and use the tools at your disposal. So, if the social network's got great cross-device data, use it to activate certain audiences but not necessarily all audiences.

So, I'm not sure if that answers the question fully, but it's a tough challenge - some people have cracked it for you, even if it's for a portion of users and you should definitely take advantage of that.