In most factories there is no single system called the supply chain; there are pieces that do not look at one another. Purchasing tracks orders in its own Excel file, the production plan runs in a separate document, the warehouse trusts its own stock count, and shipment information depends on the haulier's phone. Each of these pieces may work correctly in itself; the problem appears in the gaps between them. In the same week a line can stop because one raw material has run out while another raw material has been sitting in the warehouse for months. The two situations look contradictory, but they arise from the same cause: nobody is seeing the whole chain at the same time. The decision is taken with only as much as the person holding that piece can see, and what emerges overall is a picture nobody wanted. Where the chain is invisible, the company covers the gap in information with stock; excess stock is really the cost of a lack of confidence.
The ERP has a defined place in this picture and it fills that place well: it records the order, holds the stock, issues the invoice. But an ERP largely knows the inside of the company. It usually will not tell you whether the goods are on a ship, at customs or still waiting in the supplier's warehouse; how many times a given supplier has missed its promised date in the last six months; or whether the order that has to be placed this month is consistent with the coming production plan. For it to say these things, either information from outside the module has to enter the system, or that information has to be collected somewhere else. What happens in practice is the second: the information accumulates in the planner's head and in personal files. That is why planning effectively stops when that person goes on leave, and whoever covers cannot take the same decision.
The supply chain management layer is built on top of the ERP without replacing it. Orders, stock movements and the production plan are read from the ERP; supplier confirmations, shipment details and customs and haulage statuses are collected from outside. When the two come together on a single time axis, the answers to these questions emerge: which material will fall short in the coming weeks, which orders that shortfall puts at risk, and by when at the latest the order has to be placed to close it. The system answers these questions on a dashboard and raises an alert when a critical threshold is crossed. The decision is still taken by a person; the software's job is not to take the decision but to make the picture on which the decision rests complete and current.
The limits of this work should be drawn at the outset. The software does not shorten lead times; if an imported raw material takes six weeks to arrive, it will still take six weeks after the system is in. Nor does it turn a supplier that does not reply into one that does. What it provides is planning in full knowledge of that lead time, and learning about a delay in the week the order is placed rather than when the material fails to reach the line. The accuracy of the system also cannot exceed the accuracy of the data feeding it. If the warehouse count does not hold, if the bill of materials is out of date or if delivery dates are not entered into the system, the dashboard will show the wrong thing as well. That is why the first stage of the rollout is not screen design but establishing how reliable the data sources are; in most projects the real effort goes here, and this has to be said from the start.