Balancing Speed and Accuracy in Digital Order Fulfillment: The Next-Gen Solution

8 May 2026

Problem

An online retailer must decide, within seconds, which fulfillment centers supply each item of an order and how items are grouped into shipments. The goal is the lowest total ship cost that still meets the expected delivery timeline. Exact enumeration gives the best answer but is too slow for large orders. A greedy rule that minimizes the number of shipments is fast but often costs more, because real shipment cost is stepwise and jumps after certain thresholds.

Example from the article: a guest orders diapers, paper towels, headphones and sneakers. One box from a single location may be cheapest, or splitting into two shipments may arrive sooner and cost less overall.

Solution in brief

Treat allocation as a set-covering problem. Start from the greedy allocation, then improve it with local search using two moves: reassigning items between shipments and merging shipments. Small orders still use brute force. Target reports that the hybrid recovers ~91% of the cost gap between greedy and brute force at ~8% of brute-force compute time.

Questions this case answers
  • How to assign items to capacitated sources so that total cost is minimal under a hard latency budget, when exact enumeration is combinatorially infeasible and a greedy heuristic is myopic about step-shaped costs
  • How to choose which fulfillment centers ship which items of an online order at minimum total ship cost while still meeting the delivery promise?
  • How to keep order allocation fast when brute-force enumeration explodes for large orders?
  • How to improve a greedy shipment-minimizing allocation that ignores stepwise cost jumps?

Tools and methods

  • Brute force — enumerate all shipment groupings and pick the cheapest; used for small orders
  • Greedy by availability factor — fast initial allocation that minimizes shipment count
  • Local search (Set Reassignment, Set Merging) — iterative cost reduction starting from the greedy result
Continue reading — it's free

Google or a one-time email code. No password.