JeanFrancoisPuget 2700028FGP Comments (2) Visits (9105)
I lenjoyed reading the following from
After airline passengers wouldn’t stop complaining about the time they spent at baggage claim (even when more staff were added and wait times fell) a Houston airport simply moved the arrival gates so that passengers spent more of their “wait” time walking to the baggage claim. The complaints all but disappeared.
This reminded of this:
A number of years ago, a company that had just built a major building realized their elevators were intolerably slow. What to do? It was too expensive to reengineer the elevators. After thinking about the problem for a while, mirrors were installed in the lobby and elevators. It turns out that people will tolerate a much longer wait if they can see themselves in a mirror.
Seems this is more of a good joke than a real story (see The
What is the commonality? In both case, the business problem is similar: reduce customer dissatisfaction due to waiting. How would it be solved using optimization? An optimization expert would map this to the mathematical problem of decreasing waiting time, because the mathematical value of waiting time is directly correlated with customer dissatisfaction. Then that problem would be studied and solved using various tools, to result in better elevator scheduling, or better operations at the airport, and the waiting times would be reduced.
This looks good, yet it is not what is exemplified by the above two stories. In these stories, customer satisfaction improved without any improvement in waiting time.
Does it mean that optimization should be thrown away? I don't think so. What needs to be revisited is how the business problem is mapped to a mathematical problem. The mathematical problem should have a more accurate proxy for customer satisfaction than waiting time. Satisfaction is a function of more than waiting time. Where the wait happens is of interest. What people can do during wait is of interest. There is probably more. When optimizing we should look at the space ranged by all variables, and not just what is spanned by wait time alone.
There is a caveat though. In order to apply optimization, we need to quantify how satisfaction depends on other variables. This would probably warrant some psychological studies and analysis of the resulting data.
Update on Nov 29, 2015:
A reader, Ehsan, commented that some operations research textbooks discuss the above topic, see the comments below.
Few days after I wrote this blog, the Washington Post published an interesting article on a similar topic.