The Thesis Experience

In 1971 I was back at MIT in the Sloan School.  I had finished my EE degree in 1968 in a dual master’s course of study with Management before going on active duty with the Navy.  But I was unable to complete the Management portion.  There was a thesis requirement plus two courses to go.  I’ll describe the concept of the thesis.  But the real story is the process of writing it.

The MIT program was heavy on mathematical tools for management and one of the professors, Warren Hausman, had a problem that Polaroid corporation faced.  Polaroid was based in Cambridge nearby.  They made the very popular cameras that developed photos as they came out of the camera.  The problem was simple to state.  The big sales season is Christmas, and they estimate how many to make.  The guess is always wrong, of course.  But they wanted to optimize the core trade-off – Make more so that they never lose a sale for lack of inventory, yet possibly end up with a lot of excess that cannot be sold, or make fewer to reduce the possible excess, and risk losing sales for lack of product to sell.

In the newspaper business, this question is an everyday problem, with a neat statistical solution.  It’s called the “Newsboy problem”.  The classic solution is based on the concepts of the cost of overage (scrap), and the cost of underage (lost sales).  You estimate a probability distribution of sales.  For newspapers, there’s a lot of history to draw from.  Then you print a quantity so that the probable cost of overage balances the probable cost of underage.  Simple enough.

For Polaroid, and for many other seasonal producers, the problem can look more complicated.  They assemble many complex and expensive parts in multiple stages over an extended time of many months.  Ideally, they are also learning about the likely sales demand as that time goes on.  The better the forecast, the better the production plan can be.  That has real money consequences for the company.

So, my thesis started on “Multistage Production with Stochastic Seasonal Demand”.  Whew!  Prof. Hausman suggested that I keep him informed of my progress on a regular basis.  After a few weeks, I set up a time to brief Prof. Hausman and went to describe my progress.  I felt reasonably prepared and was working on some decent ideas.  He listened politely, asked some good questions, suggested some things to consider, and I was ready to leave.  But just before I got up, he said, “by the way, Mr. Kampe, it would be helpful when we have these talks for you to type up the material you want to discuss.”  I nodded an OK, and left, thinking that he was being rather demanding. 

I also knew that he had to be happy with me for me to graduate, so for the next meeting, I did just what he asked, and typed up the concepts I planned to discuss.  I remember that I was making a key assumption about the demand model, so there was some text, liberally peppered with equations.  I submitted it in advance, then went to see him.  At the start of the conversation, he handed by paper back, and it was littered with his edits, in red ink.  I was quietly incensed.  Still, we had a good talk about the topic.  Then I left, with my wounded and bloodied paper in hand.  I was sure I had been very clear in what I had written, so I rushed home to take a closer look.  Much to my chagrin, point after point, I looked at his edits, looked at my draft, then realized my words were not as clear as I imagined.  Every one of his edits made sense and made a better document.

That update process continued for about 6 cycles.  Somewhere around the halfway mark, we were at the very core question of just what would make the production planning process work better.  Where was the key insight?  I been wrestling with it and had some intuition on where I felt I needed to get to.  But it wasn’t exactly singing on paper, or in the computer models I had created.

Professor Hausman reviewed the progress, discussed the implications of what I had proposed so far, then calmly pronounced “that doesn’t quite seem to grasp the issue, does it Mr. Kampe?”  I knew that was true, and it still felt like an arrow thru the heart.

The next two weeks were a flurry of very focused creativity and testing of ideas.  And then the math started to work, and the computer simulation models began to show the needed improvements.  I now had a methodology that produced better outcomes than any of the currently documented approaches, at least the ones known to Prof. Hausman and me.  By then, Prof. Wallace Crowston was also advising, and seemed intrigued.

It was time to write the thesis, and writing was never comfortable for me.  Under a cloud of doom, I made a list of the topic headings that might make sense as chapters.  And then the amazing realization hit me.  Those chapters were almost exactly the topics of the briefing papers I had prepared for Prof. Hausman.  He had led me to write the thesis already!  Now the task was more a case of editorial integration, and that’s much easier to do.

I had also been learning from Prof. Hausman’s editorial feedback and was starting to get much better at self-editing.  It was a realization that good writing doesn’t just spring full-blown in final form.  Rather, it’s the hard work of crafting the words, sentences, paragraphs, the concepts, the data, and the equations, into something that conveys useful and interesting information clearly to a reader.

At the oral defense of the thesis, Prof. Hausman and Prof. Crowston asked some challenging questions.  The critical ones, though, involved 3 situations where we assumed a simplified case for each of 3 key variables.  My method had to simplify to the known optimum solution.  And in all three cases, it did.  Prof. Crowston observed, of course, “that doesn’t mean it’s optimal under all conditions in this framework.”  True, yet at that moment, I knew, and I think we all knew, that no one else had a better approach.

My advisors suggested we write a paper to submit to the periodical Management Science.  We met one day shortly after that to create the paper.  I was amazed to see them work.  They knew how it needed to be structured, assembled the pieces deftly and sent it off.  I had expected to be tasked with a lot of the effort.  Instead, we drew mostly from what existed and made decisions quickly.  I was there to make sure the concepts were accurately represented as we condensed a whole thesis into a few pages.

They let me the know the paper was accepted.  It was the first time they had submitted a paper and it had been unconditionally accepted without a redo.  It was published in the April 1973 edition of Management Science.

Prof. Hausman eventually moved from MIT to Stanford, where he became Department Chair for the School of Industrial Engineering.

The concepts from that thesis have been valuable to me for virtually my entire career.  The essence of the issue, is “How do you make decisions when outcomes are uncertain?”  How boldly do you commit money and resources?  What are the consequences of being wrong?  How do you balance that against the opportunity ahead?