15 Sep Virtual warehousing: the network as warehouse

Where stock sits no longer matters

Companies with stock in more than one place eventually face the same question: centralise or spread out. Is the combination possible too — a network of stock points you manage as if it were a single warehouse? Where the goods sit then no longer matters, as long as they reach the customer on time.

Those stock points can be distribution centres, but equally stores fulfilling web orders, a regional network of service locations, or stock still sitting at your supplier. The scale differs, the question does not.

That is called virtual warehousing, and it solves a problem the classic structures run into.

Why centralising collides with speed

The classic setup holds stock close to the customer, with the full assortment at every location. That ties up a lot of inventory, because each point carries its own buffer. And if a customer orders something that has just run out in their region, they cannot be served even though the product sits elsewhere.

The alternative is centralising: one distribution centre, far less stock, the same service level. But then you get long transport movements. A customer in Spain orders an item produced in Spain, and it shuttles to the European distribution centre in the Netherlands and back.

Both structures run into the same development. Orders are getting smaller, the sales area is getting larger, and lead times have to come down. Thin flows over long distances — neither is built for that.

The network as warehouse

Virtual warehousing treats stock points as a network. Stock preferably stays close to the source. When an order comes in, the system determines which location ships it and how the goods move through the network most efficiently.

Three mechanisms make that workable:

Cross-docking. If there is not enough volume for a full truck from Finland to Spain, you cross-dock in the Netherlands. That way you run full on both legs.

Consolidation. Goods for a single order come together from different locations at a point near the delivery address, where they can be made customer-specific before going out.

Forward stock. Without stock near the customer, lead times go up. A limited buffer near sales areas bridges the replenishment time. Small compared with full local stocks, so total inventory stays low.

What it delivers

Local stock is the fastest, but expensive: each point carries its own buffer, and a customer can still come up empty because the item happens to sit elsewhere. Central stock reverses that — low inventory costs, complete availability, but a lead time that grows with distance.

A network takes the best of both. The lead time approaches that of local stock, while inventory costs stay at central level. Against that, warehousing costs rise, because you need more locations. That investment has to come back through higher service and better load factors.

Not every stock point is yours

The principle reaches beyond your own warehouses. A retailer fulfilling web orders from store stock applies it on a regional scale: the store becomes a forward stock point with a short line to the customer and access to stock at sister stores. A wholesaler shipping directly from a supplier’s stock is doing effectively the same thing.

As soon as locations belong to different parties, something is added. You need visibility of stock you do not manage yourself, and agreements about who ships what and when. That calls for clear SLAs and deep system integration.

The business case does not change. Every additional stock point raises stock availability. The question is whether that outweighs the cost.

Amazon: the network as business model

Amazon has taken this to the extreme. Germany alone has 23 fulfilment centres. Which centre ships an order is an outcome of the system, not of a fixed regional split.

If the whole order sits in one FC, that is preferred: one pick location, one shipment, no consolidation. When I walked through an Amazon fulfilment centre, what struck me was the stream of totes arriving from other FCs. If an order has to come from several FCs, that costs an extra day of lead time to bring the order lines together. The system also weighs the available processing capacity per FC, spreading the work across the centres.

For Prime customers it works differently. They are entitled to next-day delivery, so there is no time to consolidate across FCs and the customer may receive several parcels at the door.

Kramp: the same principle at a different scale

Kramp does the same at European scale. The technical wholesaler supplies a broad assortment of parts to the agricultural sector across Europe, with small orders and short lead times.

Kramp uses an order management system that centrally determines which DC fulfils each order. At night, shuttles run to exchange goods between the DCs. That combines high availability with low inventory costs.

The question is not whether it can be done

Order management systems have matured, with vendors offering this functionality as standard. The orchestration layer between ERP and WMS is something you buy these days.

Is a network structure worth it for you? It pays off with multiple stock points, a broad assortment, small orders with several lines, short lead times and a large sales area. With large orders and generous lead times, central stock is almost always cheaper.

About the author
Jeroen van den Berg is the author of Highly Competitive Warehouse Management and holds a PhD in warehouse algorithms from the University of Twente. He has been advising companies on warehouse optimization and WMS since 1997, running his own consultancy since 2001. He develops Metrica, a Warehouse Optimization System.