Identify the Seed: When you upload your list, we identify all devices on that list that are available in the Collective. These users become the seed.
Analyze the Seed—Success/Fail Scoring: All devices within the seed are given a score of 1 (success). An equal number of devices are given a score of 0 (fail). These devices are unlike the seed and determined by Gradient Boosting, a technique for predicting models.
Analyze Device History: Device characteristics like frequency of use, screen dimensions, app ratings, locations, apps installed, and more are analyzed to determine which devices in the Collective are most like the success group, and which are most unlike the fail group.
Device Ranking: The list of characteristics gathered from the device analysis are ranked from most important to least important. For each characteristic a single device ID has, it will receive a 1. The total of the scores is averaged to create an associated gain score between 0 and 1. This gain score predicts the likelihood that a device will behave like the seed audience. For example, if a device possesses every characteristic, it would receive a gain score of 1.
NTiles: Every unique device ID within the Collective is given a gain score. Devices with a score are separated into 100 separate ntiles with even device counts. For example, if 1M devices are scored, there will be 10K devices in each ntile.
Activation: Marketers determine the size of the desired target audience and export the new device ID list to the preferred activation partner(s).