Model helps Deere save big bucks

“Implementing Seasonal Logistics Tactics for Finished Goods Distribution at Deere & Company’s C&CE Division” by Valerie Tardif, Sridhar Tayur, James Reardon, Reid Stines and Pete Zimmerman, Operations Research, Vol. 58,No. 1.

There may very well be a new incentive model for consulting groups that collaborate with large corporations: the more you save your client, the more you earn.

SmartOps, which collaborated with a division of John Deere, accepted this model and was all the better for it, proceeding to save Deere $10 million in logistics costs in the first three years of a project that began in 2004.

That year, the company’s Commercial and Consumer Equipment Division responded to a call to speed up delivery of lawn mowers, golf course maintenance equipment, tractors and other equipment to 2,500 independent dealers in the United States and Canada, especially when demand surged at the beginning of the year.Deere marketing staff perceived that customers were itchy to get what they wanted when they wanted it and might easily switch to a competitor if a Deere product was out of stock. They established twin goals of faster,more reliable replenishment of retailer inventory and keeping additional delivery costs as low as possible.

SmartOps worked with C&CE to customize delivery for each of the two seasons: peak, from the winter thaw through summer (February to July) and off-peak. Throughout the collaboration, the two companies monitored actual realized savings, making this collaboration a rarely documented example of a successful O.R. reward-sharing project.

The team created detailed models of current and potential distribution systems, using mixed-integer programming to work with potential delivery decisions and constraints.

This allowed them to use already available software to come up with recommendations and options. They also developed a highly
useful what-if analysis features.

The results: over the three years of the project, Deere significantly improved service to 82 percent of their dealers while reducing logistics costs by more than $10 million.

“Merging AI and O.R. to Solve High-Dimensional Stochastic Optimization Problems Using Approximate Dynamic Programming” by Warren Powell, INFORMS Journal on Computing, Vol. 22,No. 1.

Operations researchers, particularly those working in transportation and logistics, are sometimes daunted with the task of optimizing
hundreds of thousand of discrete entries with complex attributes – and in the presence of uncertainty, to boot. This is the challenge
for computerized programs that must manage fleets containing hundreds of trucks that move loads from one location to another.

The operations researcher’s challenge is to maximize profits while achieving other goals such as minimizing empty miles, obeying work rules and concluding each truck’s run during the driver’s assigned number of hours per tour. Company staff must also fit the information about one run into the matrix of all the current and upcoming drives.

Enter veteran operations researcher Warren Powell of Princeton’s Castle Laboratory with this piece on approximate dynamic programming. Approximate dynamic programming is a way to tame what he describes as the “curse of dimensionality” that is inherent in large problems in this class. It uses ideas from traditional O.R. and artificial intelligence. Powell comes to the problem as the author of a well-regarded book on the subject:

“Approximate Dynamic Programming,” a 2007 Wiley publication.

The journal accompanies the Powell piece with critiques by MIT’s John N. Tsitsiklis (“Perspectives on Stochastic Optimization Over Time”), Rutger’s Andrzej Ruszczyski (“Post-Decision States and Separable Approximations Are Powerful Tools of Approximate Dynamic Programming”) and a final reply by Powell, himself.

“Hybrid Entrepreneurship” Manby Timothy B. Folta,Frédéric Delmar and Karl Wennberg, Management Science, Vol. 56,No. 2.

Those who study entrepreneurs historically view a direct transition from employment to self-employment. The authors of this study, drawing on a European dataset, conclude that large numbers of entrepreneurs spend a long time in a limbo state, retaining their day jobs while developing their dream companies.Theirs is apparently the first study to systematically document the prevalence and influence of this transitional phase of entrepreneurship.

The authors find three broad implications for study of self-employment.

1. Hybrid entry influences self-employment entry but does not determine it. Rates of transition hinge on financial performance as a hybrid. The authors’ results show that a key benefit of hybrid entrepreneurship is reducing uncertainty through learning about self-employment.
   
   
2. An individual’s decision about whether or not to become a full-time entrepreneur is influenced by his or her switching costs, uncertainty around the entrepreneurial context and the quality of their workforce skills.
   
   
3. A serious research error occurs if hybrid entrants are classified as selfemployed entrants. When this occurs, the authors write, coefficients are substantially different from what would obtain if the two types of entrants were distinguished.