2- Digital Marketing

Background
The data for this assignment come from an experiment run at eBay in 2012. eBay is a USbased auction site, where consumers can offer new and used products for sale or auction. To
advertise its website and the individual products available for sale, eBay makes heavy use of
search engine advertising (SEA). SEA is the practice of advertising connected to keyword
searches in a search engine. That is, when a user types in a specific keyword (e.g. eBay, or
red lamp), a sponsored search advertisement might be shown at the top of the search
results directing the user to a specific URL. These keyword searches are assigned to
advertising parties through a bidding process, where the winner of the bidding process is
assigned the advertisement connected to a keyword search. eBay bids on over a 100 million
keywords a day, mainly on Google (through Google Adwords). Importantly, there are two
types of keyword searches: Branded searches, mentioning the name of the focal site (e.g.
eBay red lamp), and non-branded search (e.g. red lamp). Researchers at eBay wanted to test
if stopping the bids for branded searches would impact their revenues. They thus convinced
management to stop branded search for the month of May 2012.
Goal
In this assignment, you will analyze the impact of stopping branded search bids for the
month of May 2012 on the revenue generated by eBay. You will provide both model-free
analysis, as well as difference-in-difference (DiD) analysis of the results of the experiment.
Data description
The data cover the daily attributed revenues per Nielsen Designated Marketing Area (DMA)
for the period of April, June and July 2012. May is not included, as there were no attributed
revenues. Thus, the experimental setup consists of analyzing whether shutting down
branded SEA for a month has revenue consequences for the months following the
shutdown. Attributed revenues are those revenues that can be directly linked to a consumer
entering the website through a branded search, and making a purchase. A DMA is an area in
the United States for which Nielsen measures marketing spending. There are 210 DMAs in
total, covering large metropolitan areas (e.g. San Francisco, New York), but also smaller
areas (e.g. Butte and Bozeman). During the experiment, eBay did not spend any money on
branded SEA in some DMAs (treatment), while it maintained branded SEA in other DMAs
(control). These DMAs were selected in such a way that treatment and control groups were
comparable. Thus, the dataset available consists of the following variables:
 Date: The date of the observation
 DMA: A number designating the DMA of the observation (anonymized)