Bayesian estimation of the effect of aggregate advertising on web metrics

A method for data analytics, comprising:

• receiving aggregate marketing data comprising geographic information, product information, and time information of a marketing activity;
• receiving outcome data for the marketing activity;
• identifying prior distributions corresponding to a geographic effect coefficient, a product effect coefficient, and a time effect coefficient, wherein the prior distribution for the time effect coefficient decays to zero as time increases;
• generating a regression model representing a relationship between the aggregate marketing data and the outcome data using Bayesian regression based on the prior distributions, wherein the regression model comprises a geographic effect term based on the geographic effect coefficient, a product effect term based on the product effect coefficient, and a time effect term based on the time effect coefficient;
• and
• predicting a treatment effect for the marketing activity based on the regression model.

An apparatus for data analytics, comprising:

• a data component configured to receive aggregate marketing data and outcome data for a marketing activity;
• a regression component configured to generate a regression model using Bayesian regression based on the aggregate marketing data and the outcome data and further based on a prior distribution of a time effect coefficient that decays to zero as time increases;
• and
• an estimation component configured to estimate a treatment effect for the marketing activity based on the regression model, wherein the estimated treatment effect comprises a rate of effect decay.