Deterministic Data, What's the fuss?
Smartphones are our favourite electronic appendages. Our perpetual use of smartphones results in scores of data which advertisers and marketers can use to better understand who we are.
When hearing or reading about mobile advertising data, you may have come across terms like deterministic, probabilistic, 1st party, 3rd party, last night’s party.. Ok, maybe not the latter, but you get what I mean. There are varying terms used in conjunction with data and no two people have the same definition. With this confusion, we decided to write a series of articles exploring the different types of mobile data, where it comes from, how it is captured and how it can be used. In this article we will focus on “deterministic” data and what falls under this category.
What is deterministic data?
In the mobile ecosystem, deterministic refers to data that is known to be absolute; that “X” data point absolutely belongs to “Y” user. Keep in mind that to protect privacy, personally identifiable information (PII) is always anonymized. A real user is represented by their anonymous device ID.
Deterministic data is often synonymously used with deterministic matching as per the example above. Whereby we are taking multiple absolute data points to make a conclusion. For example, “X” user logged into “Y” app on his iPhone and later in the evening logged into the same “Y” app on his Macbook. The logical conclusion is user X owns both that specific iPhone and Macbook.
However, for the sake of scalability, deterministic data is often extrapolated. A small set of users identified based on deterministic data is expanded by including users who share a trait with the original audience. This is referred to as probabilistic matching. The challenge with probabilistic matching is understanding the shared trait that qualified a match.
When drawing a comparison to a real world situation, targeting based on deterministic data is like gifting your best friend a Sephora gift card after seeing her spend two hours browsing the app. Probabilistic targeting is like giving your aunt a gardening magazine because she’s over 60 and lives in a suburb — she must like gardening, right?
How is it captured?
In future blog posts we will further explore various methods of capturing data, but for now, we will focus on how one of our partners, PushSpring, are capturing deterministic app ownership data. You can read more about the partnership here
Apps installed on a user’s device are made visible through direct integrations with apps that have device and app history permissions. Data is collected by large Data Management Platforms (DMPs) through an SDK integration with mobile app publishers. These integrations allow DMPs to record various data signals based on existing information on the users device as well as behaviour patterns of the user.
How can it be used?
To explain ways that advertisers and marketers can take advantage of deterministic data we will take the example of our partnership with Pushspring and the resulting product User App Detect. Advertisers in financial, telecom, auto and transportation, and entertainment verticals have used this data to reach awareness and conquesting objectives. As an example, a bank can promote a new savings account to their existing customers that have their app downloaded. The same bank can run a conquesting campaign, targeting users with competitors’ apps.
Interest-based audience segments can also be generated. A gaming software advertiser can reach avid gamers by targeting users with Pokemon Go, Ingress, or Real Strike downloaded.
Aligning your Goals
If your goal is to reach expecting mothers for a marketing campaign selling strollers, then a deterministic data set containing users with pregnancy tracking apps would be your best option. However, if your goal is to reach people who may buy a gift for a pregnant women, than using other data sources such that were derived through probabilistic modeling can not only give you wider scale but potentially better conversion. A good rule of thumb when choosing what tactic to use, you need to understand where the data is coming from as well as where you want it to take you. Your objectives should align with the type of data you have and should bring you closer to your goals.