Skip to main content

Warehouse Slotting Optimization

Quadratic Assignment Problem (QAP) — NP-Hard

Assigning 500+ SKUs to storage locations is a QAP: items frequently ordered together should be stored together. Optimal slotting reduces picker travel by 20–35%, saving €200,000+ annually in a mid-size distribution center.

Where This Decision Fits

Supply chain operations — the highlighted step is what this page optimizes

Plant Location Facility siting
Production Planning MPS / MRP
Shop Floor Job scheduling
Assembly Line balancing
Supply Chain Distribution
Warehouse Slotting / QAP

The Problem

From warehouse aisles to quadratic assignment

A distribution center has 6 SKUs (products) that must be assigned to 6 storage locations arranged on a 2D grid (2 rows × 3 columns). The objective is to minimize total picker travel distance by placing items that are frequently ordered together in nearby locations.

This maps directly to the Quadratic Assignment Problem (QAP), one of the hardest combinatorial optimization problems. The QAP was introduced by Koopmans & Beckmann (1957) and is NP-Hard—even instances with n=30 remain unsolved to proven optimality. With 6 items, there are 6! = 720 possible assignments.

Warehouse Domain QAP Model
SKU (product) Facility i
Storage location Location j
Co-pick frequency Flow fik
Aisle distance Distance djl
SKU → slot assignment Permutation π
Total picker travel QAP objective value
QAP  —  NP-Hard  (n! permutations, no polynomial exact solver known)

Mathematical Formulation

The Quadratic Assignment Problem

Objective minimize   C(π) = Σi Σk fik · d(π(i), π(k))   // total flow × distance

Where fik = co-pick frequency between SKU i and SKU k // symmetric flow matrix
d(j, l) = Euclidean distance between location j and location l // distance matrix
π : {1,...,n} → {1,...,n}   // bijection (permutation) mapping SKUs to locations

Linearized ILP (Lawler, 1963) minimize   Σi Σj Σk Σl fik · djl · xij · xkl
s.t.   Σj xij = 1   ∀ i   // each SKU assigned once
   Σi xij = 1   ∀ j   // each location used once
   xij ∈ {0, 1}

Interactive Solver

Choose a scenario, pick an algorithm, and watch the assignment

Warehouse Slotting Solver

Standard ABC classification. High-velocity A-items should be placed closest to the depot (location 0,0).

High co-pick pair (nearby)
Medium co-pick pair
Low/no co-pick
Depot
Algorithm Total Cost % Improvement Assignment Time
Press Solve to run all algorithms
Ready — select a scenario and click Solve

Solution Algorithms

Three approaches to the quadratic assignment problem

Baseline

Random Assignment

O(n)

A random permutation of SKUs to locations serves as the baseline against which all other methods are measured. With 6 items there are 720 possible assignments; the random solution typically falls near the median of this distribution.

In practice, many warehouses operate with effectively random slotting due to ad-hoc placement decisions over time, making this a realistic baseline.

Constructive

Greedy (Popular Near Depot)

O(n² log n)

The greedy heuristic ranks SKUs by total co-pick frequency (popularity) and assigns the most popular items to locations closest to the depot (origin). This implements the classic ABC analysis principle: A-items in golden zone, B-items in middle, C-items in far aisles.

While simple and fast, greedy ignores pairwise interactions—two items that are always ordered together may still end up far apart if they have different total volumes.

Metaheuristic

Simulated Annealing

O(n × iterations)

Simulated Annealing (SA) explores the solution space by performing random pairwise swaps of SKU assignments. Improving swaps are always accepted; worsening swaps are accepted with probability exp(−Δ/T), where T decreases geometrically over 100 iterations.

This allows SA to escape local optima that trap greedy methods. For the 6-item QAP, SA reliably finds near-optimal solutions; for real-world instances (n > 30), it remains one of the best practical approaches alongside tabu search and genetic algorithms.

Complexity & Hardness

  • QAP is NP-Hard: no known polynomial-time algorithm exists; the search space is n! permutations
  • Exact solvers: branch-and-bound with Gilmore-Lawler bound can handle n ≤ 30 in hours; instances with n ≥ 40 remain open
  • Greedy: O(n² log n) — fast but ignores pairwise structure, typically 15–30% from optimal
  • Simulated Annealing: O(n × iterations) — empirically finds solutions within 2–5% of optimal on QAPLIB benchmarks
  • Practical note: real warehouse slotting also considers ergonomics, weight, shelf height, and zone constraints beyond the pure QAP formulation

References

Key literature on QAP and warehouse slotting

Koopmans, T.C. & Beckmann, M.J. (1957).
"Assignment problems and the location of economic activities."
Econometrica, 25(1), 53–76.
Sahni, S. & Gonzalez, T. (1976).
"P-complete approximation problems."
Journal of the ACM, 23(3), 555–565.
Burkard, R.E., Dell'Amico, M., & Martello, S. (2012).
"Assignment Problems (Revised reprint)."
SIAM, Society for Industrial and Applied Mathematics.
De Koster, R., Le-Duc, T., & Roodbergen, K.J. (2007).
"Design and control of warehouse order picking: A literature review."
European Journal of Operational Research, 182(2), 481–501.

Need to optimize warehouse
slotting and pick paths?

Get in Touch
Data shown is illustrative. This is a simplified model for educational purposes.
 Home
ESC