Featured · Predictive analytics
Predictive Analytics for Customer Targeting
A company wanted to limit an expensive free-trial campaign to its most valuable customers. Our team took raw point-of-sale data for 256 key customers, built recency, frequency and monetary (RFM) metrics, tested clustering solutions in Tableau, and used regression in Excel to pick the segment the firm should target. I was the project coordinator and presentation lead.
My role
- Split the work into four roles (data, Tableau segmentation, regression, coordination) and kept the team on schedule
- Pulled the data, cluster and regression work together into one storyline
- Built the executive deck for a non-technical audience, with Tableau and regression output in the appendix
- Reviewed and submitted the final package: deck, Excel file and Tableau workbook
Result
- Frequency vs. Monetary gave the clearest clusters: frequent buyers, moderate buyers and inactive customers
- Top-segment customers bought about 18 times on average, compared with about 10 in the lowest segment
- Segment membership explained 63% of the variation in spend, and frequency alone explained 88%
- Recommended targeting Cluster 1, the frequent, high-spending customers
256 key customers, by recency and spend
Each dot is one customer from the project data, colored by the segment our team found. Hover a dot for details.
- Top value · 131
- Steady · 79
- Lapsing · 35
- Inactive · 11
| Segment | Customers | Avg. spend |
|---|---|---|
| Top value | 131 | |
| Steady | 79 | |
| Lapsing | 35 | |
| Inactive | 11 |
256 → 131Targeting only the top-value segment roughly halves the free-trial list while keeping the customers who buy most often and spend the most.