🔗Overview
DADI is unique in its approach to data: the entire platform is built to facilitate Data Driven Experiences.
At the heart of this is a series of cognitive apps that provide predictive analysis to enable the creation of unique experiences targeted at the individual.
🔗Cognitive insight with DADI Predict
DADI Predict is the first of our machine learning apps. It is an API that simplifies audience-based predictions by using specific machine learning techniques.
Very basically, events go into DADI Predict and predictions about future events come out.
🔗A working example
DADI Predict uses the concepts of events - familiar to anyone that has used GA - to allow for the simple collection of data points to model against.
An event is a grouping of four types:
- Person
- Action
- Object
- Weighting
For example: user A (the person) may add a Porsche 911 (the object) to their wishlist (the action).
Predict uses our identity tools - specifically the guarantee of an individual and an issued UUID - to populate the identifier for the person in the event.
The actions and the objects are based on the predictions that we want to make within a product.
Going back to the example:
- User A adds a Porsche 911 to their wishlist
- User B adds a Porsche 911 and a Mercedes S-Class to their wishlist
This enables us to predict that User A is more likely to be interested in a Mercedes S-Class than in a randomly selected model.
The analytical approach being used is called collaborative filtering: the prediction of interests for a single user (filtering), calculated based on the interests of many users (collaborating).
The system works best at scale. Generally speaking, it needs at least 20x more data points than variables.
And the more data you throw at it, the better it gets.
