Pilot wave planning on one zone and one carrier cutoff

At 16:00, three waves show “released”. The carrier leaves at 17:00, but all three compete for one packing station. The dashboard suggests progress while the physical queue makes the promised departure impossible.

A first wave-planning pilot should prove that release decisions respect the actual constraints between stock allocation and loaded departure. Limit it to one zone, a defined shift and one carrier cutoff. Keep the supervisor in control while the system prepares and tests a feasible sequence.

Define the boundary of the promise

Identify which orders qualify for the selected departure, which event confirms that they are ready, and who may approve a different commitment. A carrier booking, a packing completion and a physically loaded vehicle are separate milestones.

The pilot needs order lines and priorities, eligible inventory, handling-unit and location data, replenishment tasks, staffing/capacity, packing requirements, dock availability and the cutoff agreement. Include the people who own the constrained work, not just the person responsible for the final dispatch KPI.

Native warehouse systems may already support slotting and replenishment within release processes. Review those capabilities before introducing another planner that competes with existing work creation. Microsoft Dynamics 365 — Warehouse slotting; Microsoft Dynamics 365 — Replenishment overview.

Expose the shared constraint in a worked schedule

In this illustrative case, the carrier cutoff is 17:00. Loading requires a fictional 20-minute buffer, so packing must finish by 16:40. Three candidate waves are otherwise ready and require one shared packing station.

Total packing time is 135 minutes. From 14:00 to 16:40, the station offers 160 minutes, leaving 25 minutes of slack under these simplified assumptions. Starting the same sequence at 15:00 gives only 100 minutes and a 35-minute shortfall.

The calculation assumes the station is continuously available, the waves arrive ready to pack and the agreed loading buffer is sufficient. Now change one assumption: replenishment and picking together make B ready to pack only at 15:20.

Keeping A–B–C means A finishes at 14:45, the station waits until 15:20, B finishes at 16:20 and C at 16:50. Packing misses its 16:40 deadline by ten minutes. If priorities and handling rules permit A–C–B, A finishes at 14:45, C at 15:15 and B runs from 15:20 to 16:20. The five-minute idle gap is harmless in this simplified case; the sequence preserves 20 minutes before the packing deadline.

The alternative works because the available work fills the gap. It does not make B ready earlier. The supervisor needs earliest ready-to-pack time, packing duration and deadline for every wave, including the dependencies behind each estimate. Replenishment completion alone is not packing readiness if picking still remains.

Release work only when its dependencies are understood

A proposed release record should carry the selected orders, priority basis, inventory snapshot/version, required replenishment, capacity assumptions, planned completion and approver. When a material input changes, re-evaluate the affected work rather than preserving an obsolete green status.

Do not infer that an order was shipped because its wave closed. The final shipment record must reflect the actual eligible contents. A partially completed wave needs order-line outcomes and an owner for the remainder.

Establish one owner of work creation

If the WMS already creates picking tasks, the pilot should send an approved release instruction or use a supported planning extension. It should not create a parallel set of tasks from the same orders without an agreed control.

Use a stable release identity and version. A cancelled or changed order invalidates the relevant plan. A repeated submission must not reserve the same stock or create the same work twice. When a response is lost, query the accepted release/work identity before attempting another creation.

The supervisor may choose another feasible sequence or reduce the departure scope. Record the reason, affected orders and revised commitment. An override does not make held stock eligible or waive a physical or safety constraint; those conditions must be resolved through their authorised process. Repeated sequence changes are evidence that the model may lack a constraint or use stale capacity assumptions.

Replay failures before shadow operation

Build an independently reviewed sample of normal and peak-shift orders. Seed a reserve-stock hold, an urgent order, a cancelled order after planning, a delayed replenishment, an unavailable packing station and an offline scanner.

In shadow mode, compare recommendations with the supervisor's actual decisions without issuing work. Investigate disagreements. A recommendation that looks faster because it ignores a required handling step should fail review.

Then test approved work creation in a sandbox or otherwise controlled environment. Verify duplicate prevention, destination acknowledgements and partial completion. Proposed acceptance gates require no release that breaches a hard stock, capacity or handling constraint, and no test order allocated twice. A failed deadline test must produce a feasible revised scope or an explicit unmet commitment, rather than a supervisor approval that leaves the impossible plan unchanged.

Keep the stop procedure operationally useful

If the planner fails, stop new automated releases while supervisors use the agreed WMS process. Preserve already-created work and its identifiers. Reconcile what is planned, released, picked, packed and loaded before resuming. Simply restarting from the order list risks creating the same work again.

Measure orders meeting the actual cutoff, work left unfinished, urgent replenishments, replanning effort and labour across the constrained steps. Retain the original in-scope order population when reporting cutoff performance; removing delayed orders from the denominator would overstate success. Show cancellations and approved commitment changes separately. Compare similar order mixes and staffing. “More waves released” is not a useful success measure by itself.

Prove one dependable departure

A successful pilot makes the release decision explainable and keeps unfinished work visible when reality changes. That is a sound base for expanding to more zones or carriers without turning a planning estimate into an unsupported delivery promise.

Next actions

1. Define the order population, shift, carrier cutoff and meaning of completed dispatch.
2. Measure the constrained steps and agree stock, labour and loading assumptions.
3. Replay normal and disrupted shifts in shadow mode.
4. Test approved releases, retries, partial work and manual fallback.
5. Expand only after the supervisor and system owners can reproduce the acceptance results.

Start wave-pilot scoping with one missed departure, the constrained work area and the available order/task records.