05 / Case
Predicting who sells, who stays, and where they fit
Machine-learning models and a new operating process matching the right salesperson to the right category and store, built in four months for a homeware group of 11 brands and 2,000-plus points of sale approaching $1 billion in revenue
- Industry
- Retail
- Client
- Major homeware retailer
- Service area
- Sales force AI
- Status
- In production
- Published
- June 2026
- Reading
- 2 min
A major homeware retailer runs 11 brands and 305 stores, more than 2,000 points of sale across 43 countries. At that scale the sale still turns on one floor conversation. Two problems: staff left fast, and the range was complex enough that who stood in which aisle mattered.
Homeware is a considered purchase across a wide, technical range. A salesperson who knows the category sells more of it; one in the wrong aisle sells less and leaves sooner. With turnover high and 2,000 points of sale to fill, the matching was done by instinct and restarted whenever someone resigned.
DA built machine-learning models around the salesperson. One predicts a new hire’s potential before a track record exists. One predicts the risk that a salesperson will resign. A third finds the category and store where a person will perform best. Three calls once made blind became predictions the business can act on.
Models alone become a report no one uses. So the engagement redesigned the staffing process around the predictions and stood up the governance to run it, so a potential score, an attrition flag, and a best-fit recommendation each arrive with an owner and a decision attached.
Four months from kickoff to a working system: the process, the governance to operate it, and the models underneath. Matching of person to category to store is now made the same way across the network and refreshed as people and predictions change, instead of reinvented by hand after every departure.