The OHM SmallData's mathematical model automatically generates the core of application users — this is a segment of the audience that brings maximum revenue (or other target events).
The "core" of most payable users is automatically transmitted to an advertising system (for example, the Facebook) that generates a segment of similar users and filters out those who have already installed the application.
As opposed to searching users similar to all those who made target actions, the user "core" allows a more accurate forecast and a higher quality of purchasing.
For applications in which conversions are generally not made during the first two days after installation, increasing the efficiency of Adwords UAC Action campaigns may take a lot of time. At the same time, the training quality is optimal if the conversion data are sent to Google during the first three days after installation.
OHM SmallData has a built-in model that forecasts, after 1-2 days from the application installation, whether a user will perform the target action within the next 30 days. The model precision reaches 95% with a recall of 75%.
Automatic transmission of data about actual and predicted conversions to Adwords using API allows the advertising system to quickly learn to buy the most useful audience and helps boost sales without increasing the cost-per-action.