The goal of the project is to optimize the BuzzFeed’s revenue from running programmatic ads by conducting in depth analysis on programmatic data and build models to predict the eCPM (effective cost per mille) prices by using machine learning models. Through this internship, I obtained domain knowledge in programmatic ads on publisher’s side, and understood how to solve business problems by applying data science skills.
(i) An exploratory analysis on what are the factors that drive effective Cost Per Mille (eCPM) prices of programmatic ads
(ii) From conclusions in the exploratory analysis, fit machine learning models that would predict eCPM prices
(iii) From conclusions (i) and predictive model (ii), propose a way to optimize the eCPM
I am in charge of analysing programmatic ads data from Ad Exchange to figure out what are the key factors to drive the eCPM prices, as well as predict eCPM of ads. To find out key vectors, I applied lasso regression and random forest regressor to predict eCPM prices to catch main factors driving eCPM prices.