08/05/2026
What can 400 customer records tell you about who is most likely to buy—or churn?
More than you might expect.
In this case study, I took the analysis from raw data to a tested predictive model:
→ Audited 400 records across five fields
→ Confirmed zero missing values and duplicate rows
→ Identified age and estimated salary as significant predictors
→ Compared four machine-learning models
→ Achieved 91% test accuracy and a 0.94 ROC AUC with an SVM model
One insight stood out: purchase rates increased sharply across older age groups, while gender showed no statistically significant relationship with the outcome.
The real value isn’t the model alone. It’s turning existing customer data into clear signals that teams can use to focus campaigns, prioritize outreach, and improve retention decisions.
What could your customer data reveal?
