Our approach began with a breakdown of total store-level sales into predictable pieces: sales from repeat customers and sales from first-time customers. Multiple data streams — such as point of sale, customer relationship management, customer satisfaction score, regional demographics, weather forecast and competitive landscape data — were combined into a staging database.

We then used machine learning algorithms to build a set of key drivers that predict visitor frequency, average ticket size, traffic of new customers and other customer behaviors. Our Action-Planner module enabled decision makers to set targets on each of these key drivers, while they saw the predicted lift in sales.

Our solutions helped reduce the variance between forecasted and realized sales. As a result of our aggressive target setting and monitoring, the average store saw a YoY sales growth of 8 percent.