WMS implementation: getting a warehouse live without losing a week

Profile the operation before configuring anything

Design decisions depend on measured reality: order profile by lines and units, product dimensions and weights, movement classes, receipt patterns, seasonality and the physical constraints of the building. Configuring picking strategies without that profile means guessing at wave sizes, pick face capacity and replenishment triggers. The profiling exercise also tends to surface data quality problems in product master records, which is helpful early and painful late, since dimensions and weights are prerequisites for almost every automated decision the system will make.

Location and slotting design comes before build

Location naming, zoning, capacity definitions and the initial slotting of products into pick faces determine daily productivity and are awkward to change once stock is placed and labels are on beams. Naming must sort into the pick path, capacities must reflect what physically fits, and slotting should be based on movement rather than on where products have always been. Doing this properly requires floor time with the people who work the aisles, and it is the design activity most often compressed when a programme is behind.

Standard process against configured exception

Every site believes parts of its process are unique. Some genuinely are, driven by customer requirements, regulated goods or the building itself; many are habits formed around limitations of the previous system. Separating the two determines how much configuration and custom development the project carries, and therefore how expensive every future upgrade will be. A useful discipline is to require a written reason, owned by a named person, for each deviation from the product's standard flow, reviewed after go-live to see whether it was needed.

Cutover: stock, timing and the fallback

The choice is between a full stop with a complete count and a clean start, a phased approach by area or product group with a period of dual running, and a big-bang switch relying on transferred balances. A full count is disruptive and gives the cleanest opening position; phasing reduces risk but requires both systems to hold part of the truth simultaneously, which is operationally demanding. Whichever is chosen, the fallback plan needs to be written and rehearsed, including the point of no return after which reverting is no longer viable.

The productivity dip is real and should be budgeted

After go-live, experienced staff become temporarily inexperienced. Throughput falls, exceptions rise, and supervisors spend their time answering questions instead of managing flow. Plans that assume normal output from the first shift create a backlog on day one that is then blamed on the software. Sensible preparation reduces inbound volume for a period, brings in additional supervision and floor support, agrees relaxed service expectations with customers in advance, and defines when the site returns to normal targets. Stabilisation is a phase with an owner, not a gap between the project and business as usual.

Frequently asked questions

Is a full stock count necessary at cutover?

Not always, but it is the only way to start with a position you know to be accurate. Sites that transfer balances instead should expect to spend the first weeks reconciling, and should plan counting capacity accordingly rather than discovering the need under pressure.

How should radio coverage and hardware be handled?

Survey and test coverage in the racked, stocked building rather than in an empty shell, because stock absorbs signal. Device readiness, charging, spares and printer configuration cause more first-day disruption than application defects, and they are entirely preventable.

What if the site holds goods under a customs procedure?

The authorisation usually specifies how records are kept and how movements are reported, so migrating stock into a new system affects an arrangement with the supervising authority. Raise it with that authority during design, since retrofitting the required record structure after go-live is far more disruptive.