machine learning
Innovate, Experiment & Automate
We usually commit to 30% better ROI for any in-housing deployment that uses our Buy ML stack. The CDP and ML stack can be plugged via our Java libraries into any Java based bidder. The SPO, CPM and CPC based integrations are fairly straightforward. In reverse our bidder can work with other CDPs by providing a scale down version of ION CDP for just ML Engine implementation.

Some of the out-of-box models support SPO, CPM and Clicks based optimisations.
Key features of our bidders
  • We provide blacklisting options for exchanges, exchange in combo with publishers and exchanges in combo with pub and ad units. This allows us to able to select the media path at a much granular level
  • The path identification in conjunction with media forecast is done before finalising on the path
  • CPM needs a bidder level bid range to be defined for its models to work. This method is required to evaluate a fair price of the media and audience and have a competitive winning. 
  • With advent of first price auctions and expand hb as header bidding with dynamic floor prices for auctions the regular bid shading mechanism doesn’t work. This calls for equitable media and audience pricing for a sustained media buy trading desk to function
  • CPC models typically provide with a robust filter to evaluate a bid request based on its context and theaudience profile and similar historic performance. Beyond a score provided by the ML Engine the bid is cleared for CPM model to take over. 
  • CPC Algorithms try to constantly strike a balance between scale and price. Some CPC models can also give scope to configure the scale of price and scale.
Download the latest Report summary by ISBA*