"All models are wrong, but some are useful" - George E. P. Box
I usually work on what one could call operational problems. There I usually do not have too much trouble figuring out the level of detail needed for a model to provide value. However, when I happen to work on tactical/strategic problems I struggle more to figure out the appropriate level of detail.
To give some support for the discussion, consider this made up example: Suppose that you are working on a problem where you want to determine the appropriate mix of vehicles to transport handicapped people from their home to the day-care center. You know that the demand varies every day but you want to decide which vehicles you need to buy to operate for the next 4 years.
Some vehicles have a fixed configuration while others can be reconfigured (some seats can be folded and you can accommodate a wheel chair in place of 2 regular seats, for example). In an operational model you would definitely want to take this reconfiguration into account because it can make a difference between a “good” and a “bad” route. However, when solving a more long term problem in which you only have a partial view of what your demand will look like in 2 months from now does it make sense to take this into account? Or is it a step too far which is just going to make the model more complex/slow without really adding values to the solution?
TL;DR How do you determine when your model is detailed enough to be useful? When is an excess of detail actually hurting the model?