In light of the May 6, 2010 U.S. market plunge I wanted to ask a question about bad OR solutions to a problem. Often times we only hear of the good solutions to problems. Yet we can also learn from our mistakes. Do you have an example where something went wrong trying to implement an Operations Research solution?

asked 11 May '10, 13:18

larrydag%201's gravatar image

larrydag 1 ♦
3.2k51326
accept rate: 9%

retagged 26 Nov '11, 15:09

fbahr's gravatar image

fbahr ♦
4.6k716


This is basically a GIGO story. I was part of a team of consultants working for a US auto manufacturer (to remain anonymous) to streamline their system for delivering completed vehicles from plants to dealers. My job boiled down to coding Dijkstra's shortest path algorithm on a PC. The code would then suck data from their database of rail shipment records (which contained carrier, start and end node, distance and some other fields for every segment of every rail shipment in recent history), and use it to route rail shipments from plants to "rail ramps". The objective metric was travel distance. This would seem to be a no-brainer.

The day came to demo the code, and the first test case was a shipment from a plant in Georgia to a ramp in Massachusetts. My code routed it (roughly) plant -> Atlanta -> Chicago -> Salt Lake City -> Seattle -> Van Nuys (CA) -> Milpitas (CA) -> Oklahoma City -> Cincinnati -> Framingham (MA). It's a very scenic route, but probably not very efficient. I couldn't figure out how I could muck up a simple algorithm that badly.

Turned out I hadn't. The data was entered into the database by some apparently unmotivated employees, who took to leaving the distance field zero if they couldn't be bothered to type it in. Thus the route above appeared to be only a couple hundred miles to the code. (The employees also had to enter a four letter alpha code for each rail junction. It turned out that 'D' meant either Detroit, Denver, Dallas or delivery point, depending on where Mars was relative to Venus that day.)

AFAIK my code was never used.

link

answered 11 May '10, 14:19

Paul%20Rubin's gravatar image

Paul Rubin ♦♦
14.6k412
accept rate: 19%

I think we have all had to deal with "garbage in = garbage out" problems. Thanks for sharing.

(11 May '10, 15:06) larrydag 1 ♦

This is my humble opinion. I have seen someone doing some manipulation of source data in order to make their solution look better than another solution. Without the manipulations, the enhancement provided by the optimization is small. With the manipulations, the enhancement provided by the optimization looks more attractive. Since there is no cooperation from the other team, there's no way to know how much improvement to our REAL current solution is... In my opinion, that's not okay. If O.R. does not offer considerable improvement in all instances, what's the problem with that? I think it's fair game to acknowldege it's not a magic ball or magic cauldron that solves it all... thoughts from a humble analyst...

link

answered 11 May '10, 18:14

lucecita's gravatar image

lucecita
511
accept rate: 0%

We were preparing a scheduling and automated time tabling tool for our business school. The objective was to schedule a large number of classes quickly. My code that took care of times and dates had some bugs that I was not aware of. In the first demo that we showed. Many classes were scheduled for 3 in the morning.

It was embarrassing but at least no stock market crashed that day.

link

answered 11 May '10, 21:03

Mark's gravatar image

Mark ♦
3.6k22350
accept rate: 9%

1

It would have been interesting to pilot test the solution and read the student evaluation forms. They'd probably complain about being yanked out of the bars an hour early.

(13 May '10, 15:38) Paul Rubin ♦♦
1

@Paul you have certainly forgotten your undergrad days. Undergrads don't go to happy hours! Drinking is typically done by large amount of cheap beer obtained from convenient stores late at night! considering all the budget cuts even grads cannot afford to go to happy hours that much.

Joking aside, we are planning to use the code for scheduling classes for courses in two semesters from now. I'll be happy to hear some ideas.

(17 May '10, 03:45) Mark ♦

Bad stories are actually good stories because they tell what went wrong. I have never applied any OR to real life applications (even though I am a PhD in OR) but I find all the above stories useful.

link

answered 14 May '10, 14:09

anonymous's gravatar image

anonymous
691311
accept rate: 8%

Your answer
toggle preview

Follow this question

By Email:

Once you sign in you will be able to subscribe for any updates here

By RSS:

Answers

Answers and Comments

Markdown Basics

  • *italic* or _italic_
  • **bold** or __bold__
  • link:[text](http://url.com/ "Title")
  • image?![alt text](/path/img.jpg "Title")
  • numbered list: 1. Foo 2. Bar
  • to add a line break simply add two spaces to where you would like the new line to be.
  • basic HTML tags are also supported

Tags:

×29
×14
×5

Asked: 11 May '10, 13:18

Seen: 2,431 times

Last updated: 26 Nov '11, 15:09

OR-Exchange! Your site for questions, answers, and announcements about operations research.