Summer Reading, 2011, Part II. Little did I know it, but the groundwork for my second piece of summer reading was set in a parking lot in Ewing, as I was trying to decipher a U.S. 1 delivery list. I was lost multiple times as I tried to follow the list that supposedly would guide me or any of our deliverers through the business parks and light industrial centers of Ewing Township.
Not only that, when I got back to the office and consulted some Google maps of various stops along the way I was further confused. Our Ewing route was taking a deliverer all the way out to New Jersey Manufacturers’ facility on Sullivan Way. But another deliverer went from Pennington through the territory covered by the Ewing list and ended up at Mercer County Airport. To make matters worse, the Ewing deliverer began his list on Lower Ferry Road, which — I discovered — is actually an extension of Reed Road. To start the list the Ewing deliverer had to drive past 20 stops already being covered by the Pennington deliverer.
So a few days later, sitting at the edge of the classic mountain lake, I quickly find myself absorbed in “The Optimization Edge,” the just published book by Steve Sashihara, founder and CEO of Princeton Consultants at 2 Research Way (see page 30).
At first it is a one-way street. Sashihara fascinates me with his accounts of how optimization helped Walmart beat Kmart, Marriott conquer Howard Johnson, and Google run past off Yahoo, among other classic examples. His succinct history of modern-day computing explains how optimization methods accelerated thanks to advances in mathematics, computer science, and decision theory that were made under the life-and-death deadlines of World War II.
But when Sashihara, a philosophy major at Princeton, Class of 1980, moves into a section on “linear programming and the simplex solution algorithm,” I began to think that this optimization stuff is great, but it’s for the big boys, not for me. Sashihara takes what he calls “a very simple example” to illustrate the concept. Farmer Jones has a tract of land and has to decide which combination of wheat and barley he should plant. Hmmm, I wonder, where to begin.
And that’s the simple example. Complex problems require (and many of you will appreciate the choice of words) “the simplex algorithm.” As Sashihara explains: “Conceptually the simplex method involves representing the set of equations as a geometrical shape in an x-dimensional space. A 3-simplex, for example, is a tetrahedron.” Throw all that in with a little game theory and power it first by mainframe computers (and now chips that can fit inside your cell phone) and you have the full force of optimization.
But — and this is evident in both Sashihara’s book and his company — the human factor enters at every turn.
Sashihara describes UPS’s herculean task of routing packages from offices and collection boxes to trucks and planes and back to offices. Among the simpler strategic considerations: Minimize the number of left hand turns a truck has to make. But other factors have to be weighed: “If a truck dispatcher always works with the same drivers, he may or she may know that one driver avoids a particular route, while another prefers it. So the dispatcher factors these preferences into the scheduling, building artificial constraints.”
The more I read Sashihara’s book, the more it applies to our little operation. Assuming that U.S. 1 is too small to profit from some linear programming and simplex algorithms, I focus on the “artificial constraints” of human bias. I recall that while our Pennington deliverer was driving through Ewing’s territory to get to the Mercer Airport, our Ewing deliverer was driving across the Pennington route to get to Ludlow Drive. Bizarre. Then I realize that the last two Ewing deliverers also worked for the Trenton Times, and the Times has a facility there.
Even a cursory study of our delivery system reveals some other human constraints that have to be considered in any optimization process. We realized long ago that because the lists are printed by a computer, some of our deliverers have been reluctant to make changes. The reasoning: If it’s this way in the computer who am I to change it?
This summer I saw something new. A first time deliverer had been assigned to a list that began at the Dutch Neck Commons, a strip mall on Route 130 just south of Route 571 in East Windsor. At the very top of the list were carefully written directions to lead from our parking lot to the retail center. But the deliverer wasn’t reading the directions. She was fidgeting with her GPS, wondering why Dutch Neck wasn’t coming up.
For years I have considered buying software that promised to lay out the ads in a newspaper to conform to various space and placement requirements. But every time I have rejected the idea, believing that the subjective factors were so great that only very sophisticated and highly customized (and prohibitively expensive) software would do it. I found vindication in Sashihara’s book, which describes the ad layout challenges faced by a major international daily newspaper. In that case the decision-making ended up in the hands of two — only two — veteran employees who juggled 64 different pre-set scenarios.
Here at U.S. 1 we have a much smaller operation, and fewer scenarios. But we still have some techniques to optimize the process. One innovation: We let the editors move ads around to accommodate their space requirements. The ad department still reviews the placement before the paper is printed, but there is a noticeable efficiency in having the editor involved in the ad placement process.
As U.S. 1’s Barbara Fox observed after her interview with Sashihara, “to be dedicated to optimization is almost a curse” — everything looks like an opportunity.
I agree. This Wednesday I am going out again to deliver that Ludlow Drive route. Let me re-phrase that: This Wednesday I am going out to optimize that Ludlow route.