EOQ vs Wagner-Whitin: time-varying demand
Wilson assumes constant demand D. When demand is known but lumpy across time buckets (typical MRP setting), Wagner-Whitin dynamic programming finds the cost-optimal lot plan and routinely improves on EOQ by 5 to 15 percent. Silver-Meal is the standard heuristic when DP is overkill[1].
Why Wilson fails on lumpy demand
Wilson averages annual demand and assumes the cycle repeats. If your MRP buckets show 50, 50, 200, 0, 50, 50 units across six weeks, an annualised D smears the 200-unit spike across the year. The right answer is often to order ahead of the spike, batch the small weeks together, and skip the zero week. Wilson cannot represent that.
Silver-Meal heuristic
Silver-Meal computes the average period cost if you cover periods j through j+k with one order. Stop when adding another period raises the average. Near-optimal in practice and a fraction of the DP compute. Most ERPs implement it as the default lot-sizing rule when the planner picks "least period cost."[2]
When to use which
- EOQ: stable demand within a 20 percent CV across buckets, no setup-cost step change.
- Wagner-Whitin: known deterministic demand, MRP-driven, small number of buckets per horizon (under 52 weeks), willing to recompute on plan refresh.
- Silver-Meal: the practical default inside an ERP. Within 1 to 3 percent of W-W in almost all real cases, much faster to compute.
- (s, S): stochastic demand. Different tool, different page.