MACHINE LEARNING AND AI IN PROGRAMMATIC

Programmatic is the process of using software to buy digital advertising. Here at Advant Technology, we specialise in programmatic advertising across the real-time bidding ecosystem. Real-time bidding is the process of buying ads on an individual impression basis. The process of a publisher auctioning an impression from their website/app and a buyer buying the impression is near-instant.

At Advant Technology we’ve employed independent artificial intelligence specialists who build models using data available from our DSP platforms. These models which go beyond machine learning help to increase ROI by using bid modifiers picking and choosing the impressions that are deemed most valuable. All of this can, of course, be done by a human but AI speeds up this process further increasing efficiency and giving our clients a competitive and unique advantage. Typically, the variables that AI will use to make decisions comes from aggregated across the following areas:

1. Geolocation data (country, region, city)
2. Time partition data (hour of the day, day of the week)
3. Size of creatives
4. Data regarding Operation systems (Family, version, carrier)
5. Data regarding the browser (browser id, browser language)
6. Data regarding gender
7. Data regarding the device (device type [computer, tablet, smartphone..], device model)
8. Data regarding the placement (placement id, placement group, publisher id, seller id, domain, fold position, mobile app instance)
9. Data regarding the supply type (browser web, mobile web, app web)
10. Data regarding the viewability and the completion rate(predicted IAB Viewability Rate, Predicted IAB Video Viewability Rate, Predicted Video completion rate)
11. Data regarding the segments (when segment are owned by the seat and their data is accessible through the LLD)
12. Data regarding the frequency/recency of impressions