I have a modelling project using a Kelly network.

An observation I have made is that the bottlenecks based on the marginal distributions of service times is not always the same as the bottlenecks based the conditional distribution of service times where the conditioning event is the customer class.

It is an analogous problem to Simpson's paradox except we're considering the set of bottlenecks instead of the reversing of trends.

There may exist a Pareto front of allocation options that involve either exploiting the marginal vs conditional constraints of a system.

Does TOC weigh in on how I should consider marginal bottlenecks vs conditional bottlenecks?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.