Remarketing Dataset
March 6, 2025
Your marketing partner would like to retarget users on Facebook who have lapsed from playing a mobile game, believing one can create ROI positive campaigns. Unfortunately, the marketer doesn’t know when to call a user “lapsed” and how to calculate lift. As an analytics guru, you stepped in and took the project over with confidence
1. Let’s design an experiment to determine the effectiveness of the retargeting campaign
- How would you define a “lapsed” user?
- What variables (i.e. indicators/features) would be important to measure the effectiveness of the campaign?
- How would you calculate the revenue lift? How would your answer change if you have data on all three probabilities below
- probability of exposure of users to a marketing campaign
- probability of opening game given that a user is exposed to a marketing campaign
- probability of opening game given that a user is not exposed to a marketing campaign
2. The result of a split test that we ran over FB over 14 days is provided here. The first group, gameofwar_rocks_1 saw a video that incentivized users to come back to the game via in-game virtual items. The other group, gameofwar_rocks_2 saw another video but they were not incentivized via virtual items. Feel free to choose your weapon (i.e Excel, R, Matlab, Python, etc) and please provide any code/output file/model you produce.
- What derived/transformed variables do you potentially need to analyze the data?
- What variables show a significant difference between the control group and the test group?
- How do you know if your sample is biased? What would you do if your sample is actually biased?
- Are there any other potential caveats in this study? If so, please explain.
- The marketer believes that the incentivized campaign has positive impact on tutorial completion. What would be your response? How would you justify it?
- What is your final recommendation about the split test? (i.e wait/kill/roll out)
The data you’ll get:
