- A business-to-business company with many high volume, high-quantity orders from customers was concerned about declining sales for a sizable number of accounts. They wanted to use their invoicing data to diagnose the causes and to seek remedies.
- Our analysis of the invoice data indicated that the issue was not with order quantities, but rather, with timing; many customers would slow down or delay their ordering patterns.
- We developed a model of the purchase timing cadence of each customer based on past sales patterns. The model produced a “likelihood score” of the odds that a purchase should occur each day. As days would pass without a purchase, eventually a probability threshold would be crossed, and the model would produce an alert that a purchase should have been expected by that time.
- The alerting system was able to reveal potential lost or deferred sales situations. Though the sales force has regular contact with a handful key accounts, the middle tier of customers, large in number, that are responsible for a significant volume of sales, are harder to monitor manually. This system alerted the sales executives where to direct their calls to help increase the buying rates of those customers.