Who did the offer actually persuade?
Sending an offer to someone who would have bought anyway is a waste. Uplift modelling tries to find the people an offer genuinely changes.
The question
Which customers does an offer actually persuade, and which would have bought anyway?
What I did
- Starting from the Udacity Starbucks capstone dataset.
- Sorting customers into persuadables, sure things, lost causes and sleeping dogs.
- Comparing a two-model approach: one model for customers who got the offer, one for those who did not.
- Adding an Excel cost calculator that turns uplift into money, then a Power BI or Tableau dashboard.
What I found
- Still in progress, so there are no headline numbers yet. I will add results here once the model has been properly validated.
Where it stands
Uses the public Udacity Starbucks capstone dataset. Independent project, not affiliated with or endorsed by Starbucks.
- Uplift modelling
- SQL
- Python
- Excel
- Power BI