The building knows nothing
Five systems decide the work and a ring of others feeds it. Where the boundaries actually run — and why the machine is almost never the one choosing.
The argumentA correct warehouse that still stops
Parts one to three each end with something you cannot change. Seven numbers you either pulled or guessed at, a method built around, a shell with a grid and a fire compartment line drawn through it.
None of that decides what happens inside on a Tuesday. A warehouse can be correct on paper and still stop — because the ERP goes down at eleven and nobody knew the pick list came from there, or because two systems each believe they hold the stock figure and neither is wrong about its own.
Part three said you get the shell once and adapt the rest. That is true of racking and false of software. A connection you have built for you is as permanent as a roof — you simply do not see it coming, because it looks like software and software feels soft.
What it coversThe stack, the ring, and the boundary between them
Five systems make the decisions, and most operations run two or three of them without knowing the rest have names.
Two layers are rarely in dispute. The ERP owns the commercial record and the WCS came with the machine. The argument is always about the three in between.
Around that stack sits a ring that decides nothing about the floor and without which the floor has nothing to do: webshops, marketplaces and EDI bringing demand in; carriers, a TMS and customs taking goods out; a PIM feeding master data and the ledger receiving the record.
Of the stack you ask who decides this. Of the ring you ask who holds the truth, and how often do they speak.
The part worth reading twiceThe boundary runs along the decision, not the machine
People draw the line around equipment. If the sorter separates the cartons, the sorter is doing the sorting. That is not true, and the mistake costs money, because it leads to buying a machine for a problem the machine was never going to solve.
| The work | Decision | Execution |
|---|---|---|
| Batch forming | WMS | Sorter, put wall, a person with a trolley |
| Emptying a location | WMS | The crane retrieves it |
| Release sequence | WES | Each machine and person picks up their part |
| The divert at the scanner | WCS | The belt and the pusher |
| Who gets the last unit | OMS | No machine is involved at all |
| Stock value at month end | ERP | Knows nothing about any of the above, correctly |
The full paper works through five situations where that boundary was drawn wrong — the webshop order that passes three systems before anyone can pick it, the vertical lift with its own picking software, the consultant who insists every location lives in the ERP, the ERP salesperson who is not lying, and the customer who would never do retail and rang three years later to ask about EDI.
The conclusion"We support that" is not "we can direct that"
An ERP vendor will tell you they have warehouse functionality, and they are usually telling the truth. There is a stock module, there are location fields, you can print a pick list.
Whether it does what parts one to three require is a different question, and you find that out on the floor rather than in the brochure. Four questions that cannot be answered yes:
- Show me what a picker sees at a location the system says is full and is empty. Who resolves it, on what screen?
- Trace a batch back to the customers who received it. Now, with the mouse. How long did that take?
- Show me one order that crosses three zones, and who is given which line in what sequence.
- After this is installed, how many systems hold a stock quantity for the same item? Name them.
The gap between support and direct is where implementations disappoint. Not because anyone lied — because two different things share one word, and nobody tested the difference before signing.