Chapter 5 Marketing Analysis I

Simple Problem: Examine whether baseball social media ad campaign had a causal effect on the number of attendee.

Data:

Suppose that we have data on each fan who was served promotional ads (treatment group) and each fan who was not served promotional ads (control group). Specifically, the data has a record of how many games each individual attended per month and how many ads each individual was served in a given month.

Note that experiment was run for some months, not for all.

Method:

• Calculate average treatment effect. The simplest starting point is to calculate mean values in only treatment group:

  • Examine how whether individuals in the treatment group went to more baseball games during the months of the experiment (July, August, and September). Compare mean attended number for the treatment group during the experiment months to the average of attended number prior to the months of the experiment.

• Think about potential confounding factors.

  • Weather: Did the weather affect attendance too?

    • You may want to plot the total attended number for different temparatures for the whole period.

  • Ranking: Did the ranking of the team affect attandance too?

    • Similar plot may be drawn for ranking and attended relationship.

The general idea can be that people also do go to more games as the weather gets warmer and their team plays better, so we do have potential confounders.

• Possible solution is to run the ad campaigns in a month that has a similar weather condition compared to other months when the campaign is not in effect. Moreover, running the campaign in a month when the team has a same ranking as he has in another month when the campaign is not in effect.