By 2030, Americans will own as many as 15 connected devices. While we're not fully there yet, the introduction of new devices every year might expedite that trend. In 2016, Adobe's coined the phrase "Devices don't buy products, people do" which is still very relevant in 2020 and will be in the future. So how does Adobe establish the link between multiple devices to one person? In this post, I'll provide an overview of the various ways by which Adobe is able to link multiple devices to a person or household.
Concept of a Device Graph
According to this article, "A device graph, also known as “identity management,” is a map that links an individual to all the devices they use, which could be a person’s computer at work, laptop at home, tablet and smartphone". I'd like to add other smart devices to this list but the most common use cases for cross device analytics we see today are still around mobile, desktop and tablet but there's continuous innovation happening to bring different types of devices into the mix.
The Adobe device graph comprises of multiple devices linked together via two methods namely deterministic and probabilistic. This Adobe article explains this concept really well but at a high level, probabilistic device link allows us to predict a person's identity (John Doe in our case) based on IP address, operating system etc. and deterministic device link allows us to identify a person based on their encrypted user ID captured on the website sent over to the Experience Cloud ID Service across devices. Please note that this matching and attribution happens
The visual below is my attempt to explain the device graph at a high level where John Doe visits a website or mobile app using his iPhone, iPad and Mac. There are prettier visuals available in the Adobe documentation but I just wanted to create something simple to convey the concept.
Probabilistic and Deterministic Linking
The Adobe device graph comprises of multiple devices linked together via two methods namely deterministic and probabilistic. This Adobe article explains this concept really well but at a high level, probabilistic device link allows us to predict a person's identity (John Doe in our case) based on IP address, operating system etc. and deterministic device link allows us to identify a person based on their encrypted user ID captured on the website sent over to the Experience Cloud ID Service across devices. Please note that this matching and attribution happens
The visual below is my attempt to explain the device graph at a high level where John Doe visits a website or mobile app using his iPhone, iPad and Mac. There are prettier visuals available in the Adobe documentation but I just wanted to create something simple to convey the concept.
Types of Device Graphs
There are primarily two types of device graphs which Adobe supports and these are ways by which you can create a true identity of your customers across multiple devices and sites. There's also an external graph option as well which allows companies to leverage 3rd party device graph data which is explained in detail here but it's not in scope for this post.
Device Co-Op
The Device Co-Op (available in US and Canada) allows companies to participate in a device graph which allows them to identify their "linked" customer devices (at a person and household level) across a magnitude of channels and websites in near real-time. The last time I heard about the scale of the Co-op, there were 100+ companies, about 300 Million users and 2 billion devices part of the Co-op device graph but this number is obviously higher now.
The concept of the Co-op can be better understood based on the visual (above) taken from the Adobe blog:
- A customer John Doe visits the travel website and authenticates on both his mobile device and laptop.
- He then goes to a retail website but doesn't authenticate so the retailer doesn't know who this customer is.
- Device co-op enables the retail website to "link" the customer's devices assuming both websites (companies) participate in the Device Co-op and identifies the anonymous user as John Doe.
- This in turn allows both companies to personalize the experience for this customer on both devices and websites. As you can see, the Device Co-op makes use of both the probabilistic and deterministic linking methods.
In contrast, if these companies didn't participate in the Co-op, then the retail company wouldn't be able to know what all devices did its customers use before purchasing something and treat it as 2 unique visitors instead of 1.
Below are some common use cases of the Device Co-op:
- If you want to access to a large pool of users for prospecting related use cases.
- If you want to perform frequency capping across devices.
- Some other use cases are covered here.
Here's a link to a document which covers the membership and eligibility criteria for companies to participate in the Device Co-op. One other requirement of the Device Co-op is for the company to make changes to its privacy policy so it takes into account all the necessary privacy requirements and it does not collect any kind of PII or behavioral data. You can view your all your devices linked to the Co-op here and can unlink your devices from it anytime.
Private Device Graph
The Private Graph is another way for a company to link their customer devices in a device graph but visible and accessible only for their own organization and not any other company.
Using the previous example, if we just look at the travel website alone we can treat it as a participant of the private graph as the data will simply be available within the context of that company. Private Device graph primarily makes use of the deterministic link as it works best if encrypted user ids are captured. Probabilistic matching will also be available to Private Graph sometime this year.
Below are some use cases of the Private Device Graph:
- If you want to reconcile user identities across multiple devices into one.
- If you want to perform frequency capping across devices thereby optimizing ad spend.
- If you want to establish a common identity across online and offline channels.
In order to participate for the Private Graph, the customer must be on Analytics Ultimate or have Audience Manager or have Target Premium and also be an Adobe Experience Platform customer. If you qualify for this as a client, subscribing to this should be a no-brainer for you given that no privacy policy updates need to take place. However, Device Co-op does give you access to a much larger amount of users which you wouldn't get with a Private Graph so you'll have to weigh your options on which one to choose.
Device Graph and Adobe Experience Cloud Solutions
In this section, I'll cover how to leverage some of the Adobe solutions I've used in conjunction with Device Co-op. Please note that I've only used Adobe Analytics and Audience Manager for customers participating in Co-op but there are use cases pertinent to Adobe Experience Platform, Adobe Target and Ad Cloud as well.
Adobe Analytics
Device Co-op is a natural fit for Adobe Analytics given the availability of the "People" metric which deduplicates the user count tied to multiple device and provides a true representation of a user as opposed to a device. This was introduced back in 2017 so it's been around for a while now and you can find more information about it here. Below is an example of what this looks like in a sample taken from the Adobe documentation.
The other service you can leverage (I've not used it yet) is Cross Device Analytics (CDA) which allows you to analyze cross device behavior in Analysis Workspace using Adobe Analytics data (needs a cross device report suite). This is not a default service and all the details and eligibility requirements are included in this Adobe Spark page.
Audience Manager
If you are a member of the Co-op and use Audience Manager, you get a lot of benefits primarily around prospecting and cross device frequency capping in offsite marketing. The first thing to do is to setup your Profile Merge Rule similar to how it's shown in the screenshot below taken from Adobe's documentation.
The other thing which is very consistent with the screenshot is that your Co-op devices per Person count should be lower than the Person count of other data sources. Finally, you should expect to see a much larger count of users (potential reach) in the Co-op Person number compared to any other data source.
The other thing which is very consistent with the screenshot is that your Co-op devices per Person count should be lower than the Person count of other data sources. Finally, you should expect to see a much larger count of users (potential reach) in the Co-op Person number compared to any other data source.
Adobe Experience Platform, Adobe Target, Ad Cloud
Device Co-op also provides additional capabilities to Adobe Experience Platform for id stitching, Adobe Target for personalizing experiences across devices and to Ad Cloud for media uses cases around retargeting across devices. I haven't personally leveraged Co-op for AEP, Target and Ad Cloud so I'm unable to provide any additional context as I don't have much exposure to these three in the context of Device Co-op.
So, that was a high level overview of the Device Graph but I'll be writing more about it as I learn more about how it's used with other solutions of the Adobe Experience Cloud stack. Please feel free to share your feedback!