Algorithms / Policies

Abstract

An algorithm (or “policy”) is a rule that suppliers follow to generate orders. The output of these policies is a set of N-curves driven by the customer demands, N₀(f). Policies that always yield non-negative orders (i.e., allow no cancellations) produce non-decreasing N-curves. They will be said to be “monotone”. Monotonicity is a desirable property in practical applications, and will be required of the algorithms developed in Chapter 5. We have already seen that monotonic algorithms avoid backorders (approximately) if (1.5) is satisfied (approximately) for all possible customer demands. This reliability condition will also be demanded of the proposed algorithms. The analysis of the bullwhip effect done in Chapter 4 is completely general, however. It applies to non-monotone and unreliable policies.