“Modeling Conversions in Online Advertising”

Document Type

Presentation Abstract

Presentation Date

4-20-2010

Abstract

This work investigates people who purchase online and how to predict such sales. Advertising as a field has long been required to pay for itself–money spent reaching potential consumers will evaporate if that potential is not realized. Academic marketers look at advertising through a traditional lens, measuring input (advertising) and output (purchases) with methods from TV and print advertising. Online advertising practitioners have developed their own models for predicting purchases. Moreover, online advertising generates an enormous amount of data, long the province of statisticians. My work sits at the intersection of these three areas: marketing, statistics and business. Academic statisticians have approached conversion modeling through a proportional hazard framework. Here we seek to provide new statistical learning tools to practitioners. We investigate a data visualization technique that allows cookie histories to be compared easily. We also provide a framework to use existing clustering algorithms to better understand the paths to conversion taken by consumers. Finally we marry the leading edge of online advertising conversion attribution (Engagement Mapping) to the proportional hazard model, showing how this tool can be used to find optimal setting for advertiser models of conversion attribution.

Additional Details

Doctoral Dissertation Defense. Link to the presenter's dissertation.

Tuesday, April 20, 2010
3:10 pm in Math 108

This document is currently not available here.

Share

COinS