Digital advertising has transformed the way companies reach potential users of their services. In this work, we define the problem of profit optimization in digital advertising for a company that develops mobile applications. Based on the analysis of data from an imaginary company, we develop an innovative model for determining the optimal control settings for digital advertising campaigns, specifically the cost per install (CPI). In the course of the research, we test various machine learning models. The developed model enables the prediction of the number of paid and organic app installs based on the set CPIs, allowing the company to find the optimal balance between advertising revenues and costs. The optimal values of control variables, i.e., the set CPIs for all used advertising networks, are determined using Bayesian optimization separately for each application and each segment, which represents a combination of country and user's phone operating system.
|
