Customer models on visit behaviour, test profiles, and payment behaviour, a customised offer for each segment, and an A/B test before any offer was scaled.
The problem
Campaign audiences were picked with fixed rules, so every customer in a list received the same offer whatever their history.
What I did
- Replaced rule-based list selection with ML customer segmentation.
- Built new customer models on visit behaviour, disease and test profiles, and payment behaviour.
- Gave each segment a customised offer, and A/B-tested every offer against a control group before scaling it.
- Automated the impact reporting, so each campaign's result was visible without manual work.
- Supervised the junior data scientists on the team.
Result
- Monthly campaign impact from about EGP 10M and 15K visits to about EGP 40M and 45K visits (about USD 0.8M a month) within 3 months.
- An annual incremental-revenue run-rate above EGP 300M (about USD 6M).
- A related House Call propensity model, targeting branch customers likely to book home visits, added about EGP 3M (about USD 60K) a month.
Confidentiality
Employer work: no data or code is shared. Figures are the business's reported campaign results.
Tools
Python, ML segmentation, Oracle SQL, Power BI, A/B testing