- Connects known and unknown devices (declared and predictive)
- Create a CDIM segment directly from the report
- Aggregate audiences and their data across devices
- A choice of similarity levels allows you to consider the trade off between accuracy and scale
- Improves reach, relevance, and inventory monetization
- Increases your opportunity to sequence ads, at optimal frequency, for consistent messaging and experience across channels, throughout the journey
- Delivers accurate analytics for segment profiles and campaign conversions
CDIM is used in many different places within the Audience Studio, each with different timing considerations.
- Probabilistic CDIM - Once this is enabled for your account, it will start working immediately.
- Deterministic CDIM - Immediately after you implement this component correctly, segment populations (and other parts of the platform) will start to use these relationships in conjunction with probabilistic CDIM.
- Grow My Population (Extension) - Growing a population with CDIM is a complex model that is run only once a week. As a result, you should not expect this to be available until the Monday after you have extended your segment.
The matching model is global, but it is applied specifically to the devices of each customer.
The core of the model is designed to uncover the devices that stay together in various places over long periods of time. To do this effectively there are three important considerations - scale, training, and validation.
- Scale - Because the success of the model is predicated on seeing lots of devices and how they move over time, it is essential to have massive reach into the device world. With one of the largest device footprints on the planet, Audience Studio has an advantageous position from which to deliver accurate results.
- Training - Audience Studio uses deterministic data coming from authenticated events (logins, purchases, etc) to provide a truth set for training the algorithm.
- Validation - From the deterministic data set, Audience Studio reserves some data to validate the efficacy of the model.
Between these three items we are able to continuously improve the accuracy of the predictions and deliver a better product.
To better understand the view of devices over time, consider a user's work and home movement. During the work day, all of the user's devices are likely to be tied to the same IP from work. However, colleagues are also tied to the same IP and in that single snapshot of time, it may be assumed the user and their colleagues are the same person.
Looking at that same IP address at night, there is likely no activity, as everyone has left the office. When looking at the user's home IP at night, all of the devices that belong to that user and their family are tied to the same IP. That single snapshot might imply that the user and their family are one "person".
These intervals are benchmarks designed to bucket the trade-offs between accuracy and scale.
No - not every single device is mapped to a person (probabilistically or deterministically). As a result, the unmapped devices which would skew this calculation, need to be excluded.
The predicted set of devices come from the global Audience Studio universe.
Audience Studio applies logic to ensure that we credit the most rational ONE deterministic ID to each KUID. We take into account multiple factors when making this determination, but frequency and recency are highly weighted aspects.
For example, if you and your significant other have both logged into a major news site on the same device (KUID1), but you have done so 10x more times and you were the most recently logged event in association with that device, the credit would be applied to you, and not your significant other.
The CDIM solution in Audience Studio does not currently have a household component at this time.
No, Audience Studio does not use 3rd party data to deliver CDIM.
Yes, If a match table is provided, it can be incorporated and the matches will be only available to that account.
No, the device graph itself is not exportable.
There will be no impact in the immediate near term. However, we look forward to the future where the worlds largest offline CRM system and the Audience Studio are fully integrated and can deliver unprecedented scale in person-based marketing.
Currently, all data is used for the CDIM algorithm. Region-specific versions are not available. Audience Studio is currently and will continue to be compliant with all global and regional privacy policies.
Audience Studio can map ALL devices. However, this is a question of accuracy and scale. The higher the accuracy, the fewer matches expected.
Audience Studio is not able to capture Safari in CDIM, as we do not have the ability to drop 3rd party cookies. We are working hard to identify a solution that will address the limitations of Safari in CDIM.
CDIM includes app and web. Audience Studio supports IDFA, AAID, and Cookies. We will soon be adding additional device categories to CDIM.
The quality of the probabilistic matching algorithm is improved by deterministic data coming from all clients. If a customer does not have deterministic data, it does not mean that CDIM is not a valuable feature. It simply means the entire set of devices in the account must be probabilistically matched. Several factors will determine the validity of this approach for any given client. These factors are related to the level of uniqueness of the audience relative to the general device pool.
Is there a minimum threshold of segment size to apply CDIM for a segment?
No, there is no minimum threshold in segment size to apply CDIM for a segment.
How does GDPR impact CDIM scalability?
GDPR could potentially impact CDIM scalability. We use devices that are analytics consented only to build a CDIM predictive model. Then we use that CDIM predictive model to score the devices that are cross device consented only. As such, lower cross device consent has led to lower scalability with the CDIM predictive model.