Warehouse Layout Simulation
Compare rack arrangements and SKU placement before changing your picking operation.
Why use this tool?
Which proposed layout reduces single-picker travel for the same demand?
Per-order travel, picking time, location visits, and relative improvement.
Two 28 by 24 m warehouses with 30 bays A1-F5, connected aisles, shared induction, and identical picking demand.
Lower travel with unchanged pick time suggests a movement saving, not a guaranteed labour saving.
1. Layout Editor
Layouts and orders stay in your browser. Saving is explicit, device-specific, and may be cleared with browser data. JSON preserves all layouts and simulation settings; keep a copy for long-term reuse.
Pale green cells show walkable aisles and cross-aisles after clearance around racks and walls.
Run a simulation to inspect the first 20 sampled routes.
Shared SKU Demand
Frequency weight represents relative unit-pick demand: a SKU with weight 40 is drawn four times as often as one with weight 10. Probability = SKU weight / sum of all SKU weights. Higher-weight SKUs are fast movers. Enter historical unit picks for a representative period, or relative estimates; zero excludes a SKU. Adding another storage location does not change its demand.
CSV columns: SKU,FrequencyWeight. Matching SKU names update their weights; new names are added. Other demand records and location assignments are retained. Up to 30 unique SKUs; files up to 1 MB. New SKUs must be assigned to locations before simulation. Files are processed in your browser.
| SKU | Frequency weight | Pick probability | Action |
|---|
2. Simulation Settings
No simulation has been run.
3. Results and Comparison
Simulate one layout or compare alternatives using identical SKU weights and seeded orders.
| Layout | Avg distance (m) | Min / max (m) | Avg travel (s) | Min / max travel (s) | Avg total (s) | Min / max total (s) | Distance improvement | Time improvement |
|---|
Average Travel Distance
Shorter routes reduce movement under the same sampled demand; this is not an optimality certificate.
Average Order Time
Total time equals travel plus unit-picking time. Fixed picking time limits percentage time savings.
Location Visit Frequency
Compare the concentration of picking activity across locations. The table below provides exact unit-pick and order-visit counts.
Actual Visits and Unit Picks by Location
| Layout | Location | Unit picks | Order visits |
|---|
Methodology and Limitations
A* finds shortest orthogonal paths on a 1 m cell grid around storage locations. Obstacles are expanded by (aisle width - 1 m) / 2; cell centres also retain half the aisle width from walls. Pick faces use the nearest conservative grid cell. Start/end coordinates identify cells, whose centres are at X + 0.5, Y + 0.5 m. Narrow or disconnected access is rejected.
Each unit pick is drawn independently with replacement, with probability equal to SKU weight / total weight. Demand is stored once per SKU in the shared catalogue. Every layout receives the same orders. Available quantities reset before each order. An order that exceeds the total available quantity fails explicitly; replenishment and stock depletion across orders are excluded.
Nearest mode chooses each next location by distance and remaining stock. Fixed mode visits the required locations alphabetically from A to Z, with numeric bay ordering (A2 before A10). The baseline allocates units to stocked locations once; repeated picks at each location are combined, then all layouts follow the same sorted stops and quantities. Matching locations and stock are required. Neither method guarantees a globally shortest tour. Travel seconds = metres / speed; total seconds = travel seconds + units x pick time.
The shared SKU catalogue, generated orders, speed, and pick time apply to every layout. Location IDs, positions, assignments, and start/end points can differ. Keep dimensions and total available quantities comparable when isolating a placement change. Improvement = 100 x (baseline mean - alternative mean) / baseline mean; negative means worse. Min/max are sampled extremes, not future bounds.
One picker, constant speed, flat floor, and unrestricted two-way aisles are assumed. Congestion, turning delays, ergonomics, rack height, safety regulations, and labour scheduling are excluded. Aisle clearance is a geometric approximation, not a safety certification. Validate with measured warehouse trips and several seeds before committing to a layout.
Limits: six layouts per project, thirty locations per layout, thirty shared demand SKUs, 80 x 80 m, twenty thousand orders, and thirty unit picks per order. Saved results include inputs, seed, method, diagnostics, and sampled metrics.
Related decisions: ABC/XYZ inventory prioritization and facility-to-customer network allocation answer different questions.
FAQ
Can the same SKU be stored in multiple locations?
Yes. Create unique location IDs and assign the SKU and available quantity to each. Its frequency weight appears once in Shared SKU Demand. The picker chooses the nearest location with usable stock and can split a pick across locations.
What does frequency weight mean for a fast mover?
A higher weight gives a SKU a higher probability of being requested. Weights 40, 10, and 50 give probabilities 40%, 10%, and 50%; weights 4, 1, and 5 give the same probabilities. Weight is demand, not a location score or stock quantity. Use historical unit picks from the same period where available.
Can induction and drop-off be in different places?
Yes. Clear Start and end are the same point, then position each independently. When they are shared, moving the induction point also moves the drop-off point. Routes finish at the configured end point.
What do the two sample layouts represent?
Layout 1 has longitudinal rack banks and three picking aisles linked by front and rear cross-aisles. Layout 2 has transverse rack banks with side cross-aisles. Both have 30 bays A1-F5, the same 28 by 24 m footprint, matching SKU assignments, and the same induction point. Dimensions and demand are synthetic teaching inputs.
How does Fixed pick sequence compare layouts?
Order contents are randomly sampled, but location visit order is alphabetical. For example, required stops B5, B1, D4, C2, A2 become Start > A2 > B1 > B5 > C2 > D4 > End. Bay numbers sort numerically, so A2 precedes A10. Picks at the same location are combined into one stop, retaining all quantities. Every layout uses the same sorted stops. Nearest eligible location instead chooses each next stop by travel distance.
Does this optimize the warehouse layout automatically?
You create candidate layouts; the simulation compares their picking performance. It does not automatically search all possible placements. Use the results to review proposed arrangements under representative demand.
Why might a wider aisle make a location unreachable?
The model reserves clearance around locations and walls. Move the location, change its pick face, or widen the physical gap. Blocked start/end points and disconnected pick faces cannot produce a valid simulation.
Can I open this without a server?
Yes. Libraries are bundled locally, with a cancellable batch fallback when a local file cannot use Web Workers.
What is stored or uploaded?
No layout data is uploaded. Browser saving is explicit and device-specific. Project JSON preserves editable layouts and the shared demand catalogue; result JSON captures the calculation-time snapshot. Legacy version 1 projects are converted to unique location IDs with their original SKU assignments and ample per-order stock.