An irresistible supply-chain story

Want to hear what could be one of the best supply-chain success stories ever? Take the $4 billion commercial and consumer equipment division of a $20 billion company, reduce inventory by $500 million, and as sales grow, keep inventory constant -- thus avoiding an additional $500 million in inventory. This is what John Deere did starting in 2002, with the help of supply-chain optimization software vendor SmartOps.

When I spoke with SmartOps CEO Sridhar Tayur, he told me that the key concept behind Deere’s success is the use of stochastic theory when building forecasts. According to Tayur, stochastic theory dictates that what you know today is quite likely to change tomorrow. How likely it is to be different is what you capture in the model.

Whereas most BI tools take a deterministic view of data -- data is known and unchanging -- stochastic modeling anticipates change and captures characteristics of the variation mathematically. One way of discovering variability in the future is by looking at past variability. But what stochastic analysis is really doing is finding comparable variations.

Consider the complexity of the task at John Deere. Reduce the inventory-to-sales ratio by half, and as sales increase, keep inventory levels stable with 2,500 dealer/owners and 100 product families, each with 10 to 15 configurations. Products include everything from ride-on lawn mowers to golf course maintenance equipment, aerators, and utility tractors. Plus, 65 percent of all retail sales occur between March and June.

John Deere’s director of order fulfillment, Loren Troyer, tells me that Deere executives knew they could bring inventory down, but didn’t know whether that would cause shortages that would hurt service levels, given that John Deere sells “impulse-purchase products.”

(There must be more fearless folks out there who would buy a ride-on mower on impulse than I thought. But that’s another story.)

Before turning to SmartOps, Deere used a “crude rule of thumb” to plan target inventory levels for each dealer. The goal was to maintain 30 percent of annual sales in inventory for each dealer. But this rule of thumb didn’t take into account seasonality  or the specific requirements of individual dealers -- in other words, variability. If a dealer sells 1,000 units in May, he or she might sell anywhere from 800 to 1,200 units in June.

Replenishment cycles represent the other variable. If a product takes two weeks to replenish, the variability might be so great a dealer could run dry. If the same product took two days to replenish, the dealer could carry much less inventory because there is less variability in two days than in two weeks.

“The question is, How much inventory do you have to have to be able to cover the variability in both supply and demand?” Troyer says.

SmartOps loaded the data from three John Deere plants and 25 dealers into its MIPO (Multistage Inventory Planning and Optimization) software. By designing a model that forecasts what products should be at which locations each week, the MIPO software showed how Deere could improve service levels while maintaining less inventory.

Since the project began in 2002, Deere has reduced inventory to the tune of $1 billion. It went from the traditional push model of inventory to a pull model. It took an investment of between $1 million and $3 million dollars, but the ROI has been dramatic. I find that simply irresistible.