Slotting software should help a warehouse team compare better choices, not pretend the building is a clean optimization problem. Travel distance matters, but so do replenishment work, ergonomics, product handling, congestion, automation limits, and the cost of moving inventory.

Build a trustworthy facility model

The model needs a current view of locations, dimensions, storage rules, pick paths, inventory, and demand. It also needs the constraints operators already use: hazardous-material separation, temperature zones, weight limits, product affinity, pick-face capacity, and replenishment access.

If these records are incomplete or disputed, the first product should expose and correct that problem before recommending moves.

Score the tradeoffs

A useful objective function can compare expected travel, replenishment frequency, congestion, handling requirements, and relocation effort. Each factor should be visible to the user. Operators should be able to understand why a move is recommended and adjust the weight placed on each tradeoff.

Keep the WMS in charge of inventory

The optimizer should propose a plan and send approved work through existing inventory controls. The WMS remains the system of record for location and quantity. A warehouse execution or robotics layer can use the same approved plan to sequence work, but it should not create a second inventory truth.

Coordinate people and automation

Robotics coordination adds another set of constraints: charging, queue depth, eligible zones, handoff points, and shared aisle capacity. Recommendations should account for those limits without assuming that every robot or conveyor exposes the same controls.

The operating team still needs a way to pause, reject, or resequence work when the building changes faster than the model.

Pilot one area

  • Choose a stable product family or zone.
  • Compare the current slotting plan with the proposed plan before moving stock.
  • Review recommendations with supervisors and inventory control.
  • Track travel, replenishment work, congestion, and exceptions together.
  • Record why operators accept or reject each recommendation.

The first pilot should prove that the recommendations are understandable and operationally safe. A smaller model that earns trust is more valuable than a complex optimizer the floor will not use.