Source: splitmetrics.com

1. Randomly assign customers to treatments
2. Measure response(s)
3. Compare groups to determine how the treatment changes response

Source: Optimizely Blog

By randomizing over a large number of customers, we create groups that are the same, on average.

Any behavioral differences between these groups is caused by the treatments we randomly assigned.

• Professor at Drexel University
  • Teach data-driven digital marketing and marketing experiments
  • Develop new marketing analytics tools
    • Measuring Multi-Channel Advertising Response
    • Test & Roll: Profit-Maximizing A/B Tests
  • Make analytics accessible to practitioners
    • R for Marketing Research and Analytics
• Previously
  • General Motors
  • The Modellers
  • Wharton Customer Analytics

Slides and code at https://eleafeit.github.io/ab_test/.

R Markdown source is at https://github.com/eleafeit/ab_test/.

We will walk through a set of examples that I created using the R (a statistical programming language).

• If you don’t know R, let me drive the R syntax so that you can focus on where we are going. Download the slides and follow along.
• If you are learning R, you should also let me drive. I can answer some R syntax questions along the way, but I don’t want to get stuck in the syntatical mud. The code will be there later when you want to review.
• If you know R well, download the RMarkdown files and run the code as we go along.