The mobile advertising industry loves buzzwords and, lately, there’s been a set of new, much-buzzed-about terms. Since the deprecation of the IDFA, adtech players have discussed ad nauseam the difference between deterministic and probabilistic data and how to market to users using a blend of both. Below, we define each type of data. We also discuss how the latter type — probabilistic — can be used to effectively retarget in accordance with Apple’s ATT Framework.
What is deterministic data?
In mobile advertising, deterministic data is data that links directly to information about a user that is known to be true and accurate. This accuracy can be 100% verified since the data is provided directly from people. For example, when a user signs up for a one-year subscription and they give their age and email address, those details are deterministic. In addition to demographic information, deterministic data can also take the form of a user’s interests or commonly visited mobile websites or apps. However, for mobile browsing data to be verified deterministically, a user’s device ID like the IDFA must be accessible.
How is deterministic data used in mobile advertising?
Deterministic data can be used to track an individual across mobile websites or apps for ad measurement and attribution. Since Apple’s ATT Framework , deterministic identifiers like the IDFA must be provided with consent by users on iOS devices. For brands that obtain opt-in from a user, deterministic data can be used to build targeted mobile marketing campaigns.
Is deterministic data the same thing as first party data?
In certain situations, yes. First party data is information a brand collects directly from its customers and, therefore, owns. So, first party data that’s gathered directly from people by a brand such as names, emails or phone numbers is deterministic. But, if a brand collects first party data through other means, it might not necessarily be deterministic. For example, a brand might collect first party data in the form of actions taken on a mobile landing page, articles read, purchase transactions or other behavioral data. Since this data is not supplied directly from a person themselves, then it’s not deterministic. Remember: deterministic data is typically information someone supplied herself, usually by logging in with a name, email address or phone number.
What is probabilistic data?
Probabilistic data consists of contextual signals. These contextual signals can include the operating system of a user’s device, their IP address, page views, time spent, etc. These individual pieces of information are compiled, grouped and analyzed to make conclusions about a users.
To generate probabilistic data, marketers can use algorithms and machine learning to identify behavioral patterns within a group of users. From there, marketers can group users according to specific behavioral patterns to serve them more relevant advertising. For example, a brand might generate probabilistic data that groups users by the media they’re most likely to consume. Another grouping could be according to the type of device they’re most likely to use to access a touchpoint.
Do you need device IDs to obtain probabilistic data?
The short answer is no. Probabilistic data is developed from contextual signals which are privacy-compliant data points that relay useful information about a user. This information can include their location, device type, and the characteristics of the app or mobile website on which the ad is shown. Using these signals, marketers can match an ad to an impression that accurately assesses the probability of a user engaging. From here, they can generate probabilistic data to determine the amount to bid for each impression. This helps iterate campaign spending over time.
How is probabilistic data used in mobile advertising?
Using machine learning, probabilistic data can be developed from contextual signals. From there, contextual signals can be combined with other metrics to create algorithmic models that predict the probability of future desired outcomes. For example, you can combine contextual signals with information about the number of interactions made with a specific ad element, like a CTA button, to understand what parts of a creative are driving performance. From there, you can determine the real time value of each ad impression with close to the same level of accuracy as device ID-powered advertising. New contextual signals can be added to the probabilistic data model and iteratively tested against a holdout group. If the signal doesn’t improve performance, then it’s not used. But if the signal has a positive impact, it’s added to the existing model and tested on live traffic to continue enhancing the predictability of the model.
Can you use probabilistic data to retarget mobile users?
Absolutely! While device IDs like the IDFA were key to helping brands capture user data across platforms for targeted mobile advertising campaigns prior to iOS 14.5, many DSPs have innovated alternative ways to track and predict users’ mobile behavior in today’s age of enhanced data privacy. For example, YouAppi’s DSP has developed a way to support targeted campaigns by using its proprietary contextual targeting algorithms. These proprietary contextual targeting algorithms leverage machine learning to identify connections between a diverse set of contextual signals and build probabilistic data models. As mentioned above, this type of probabilistic modeling is compliant with Apple’s IDFA changes.
What’s the difference between deterministic and probabilistic data?
- Deterministic data. Collected from users when they input their information when filling out surveys, using social media platforms or make a purchase. Deterministic data can be used to track an individual across mobile websites or apps for ad measurement and attribution. Since Apple’s ATT Framework, deterministic identifiers like the IDFA must be provided with consent by users on iOS devices.
- Probabilistic data is generated from contextual signals gathered from a user’s browsing behavior. Contextual signals are privacy-compliant data points that relay useful information about a user. This information can include their location, device type, and the environment in which an ad is shown. Many DSPs have innovated technology to build ATT-compliant probabilistic models using machine learning and contextual signals.
Use probabilistic data to re-engage with your best customers
YouAppi’s DSP uses a blend of machine learning, contextual signals and probabilistic modeling to build hyper-targeted retargeting campaigns. Reach out to us to get started today.