Daily Archives: November 6, 2006

Everyday Lessons from Operations Research

Although many of the presentations at the INFORMS Annual Meeting in Pittsburgh are very academically focused, Mark S. Daskin‘s Presidential Address Everyday Lessons from Operations Research had a lot of great lessons for those of out us in the field. The following are Mark Daskin’s Top 13 lessons from operations research:

  • 13. Service Gets Worse as Utilization Increases
    Think about customer service lines.
  • 12. Performance Degrades as Variability Increases
    … but variability can be reduced through risk pooling. Think about global sourcing.
  • 11. Variability is necessary.
    Relationships cannot be determined without variability.
  • 10. Expect the unexpected in today’s world.
    There are 300 M people in the US, which means 800 K will be at least 3 standard deviations from the mean in any study you are conducting. There are 1 B people in China, which means at least 2.7 M will be 3 standard deviations from the mean. Globally, there are over 400 K people over 4 standard deviations from the mean in any study you are conducting.
  • 9. If it is too good to be true, it probably is not.
    Learn from experience – and samples.
  • 8. Life is full of errors.
    Both Type I (false positive) and Type II (false negative) – and there is no free lunch. As you decrease one type of error, the other type of error increases – unless you want to pay for more data.
  • 7. A good decision can result in a bad outcome.
    C’est la vie.
  • 6. If you are not using all you have, don’t pay for additional quantity.
    It’s not savings if the item perishes in industry.
  • 5. You can never do better by adding a constraint.
    Adding a constraint can never improve the objective function – that’s optimization.
  • 4. Keep it Simple
  • 3. Think about problem formulation
    • What do you know?
    • What do you need to decide?
    • What do you need to achieve?
    • What inhibits you?
  • 2. Look for compromise solutions
    Sometimes optimal is not good enough – do a tradeoff and robustness analysis before accepting a solution since your optimal solution may be very susceptible to change.
  • 1. Data is not information.