The trend of moving towards a multi-node network and for brands to move towards an omnichannel model has been on the horizon for a while. The COVID-19 pandemic pressed the fast-forward button on the need for companies to adapt their strategies as the world was put under governmentally mandated lockdowns, shops closed but consumer demand for goods continued. Customers moved their shopping online and with it came the demand for a better-organised fulfilment strategy.
Even with shops open again, consumer behaviour has not returned to pre-COVID levels. Norms of only 3-4% on average of sales going through the e-Commerce (e-Comm) channels are no more. Post-COVID lockdowns we are now seeing businesses reporting around 40-50% of sales being done on their e-Comm platforms. Alongside now choosing to shop online, consumers still expect their delivery times to be short (with companies such as Amazon promising 2-day delivery times globally). Consumers want their products to be delivered fast meanwhile, brands need to fulfil customer expectations while also managing the costs of their deliveries.
EMBARKING ON THE PATH TO CHANGE
To allow companies to be able to fulfil customer orders effectively, timely and be cost-effective - using multi-node networks, within an omnichannel set-up, is the way forward. Consequently, the problem that needs solving becomes:
“How do companies segment inventory and allocation within their nodes to best meet all their targets?”
Companies need to decide their priorities - Customer Promise, Fast delivery times, Cost savings, Quality, Sustainability initiatives etc. Companies then need to decide their physical location set-up, based on the country or countries they will distribute to, including the fixed costs of these locations, as well as if there should be an Inbound Sortation Centre to distribute inventory and how they will handle returns.
Let us take a simplified version as the basis for further discussion (see picture below). Assume we have one Inbound Sortation Centre (which is close to a port for ease of deliveries from overseas). Then we have 4 Distribution Centres (DCs) randomly spaced around the country with one sitting near the centre. The Inbound Sortation Centre will be where all products are sent from suppliers, the Inbound Sortation Centre will then distribute inventory based on allocation rules to the DCs. The DCs will then ship products directly to customers. Which DCs fulfils each order will also depend on fulfilment rules set up in advance.
What needs to be considered for initial inventory allocation? How do we segment our inventory?
Inventory analysis and segmentation will need to be run on the product classifications when they arrive in the Inbound Sortation Centre. Product segmentation is based on expected sales (including previous similar attributed products); consideration should also be taken around the specific needs of the regions near the DC. For example, holding less inventory of snow boots if the region has a generally warm climate. For a more in-depth understanding of how segmentation can be set up - keep an eye out for future articles.
The most popular forecasted products from this segmentation should be distributed to all DCs. The items which may be less popular in one area versus another should be more heavily balanced towards the DC closest to the popular area. Any less popular products may only need to be stored in the central DC. The purpose of the central DC will be to stock all products, and this should also be the location where returns are received (due to all products being stocked there). Individual companies need to create their own rules to further allocate inventory once the initial allocation is completed. Inventory allocation rules will also need to include checking current DC stock levels to balance out the inventory across the network.
Even with allocating inventory across the nodes to optimise distribution, it is impossible to get the inventory levels ‘right’. Although a forecast may be 80% accurate, it will never fully predict consumer behaviour or the external factors which may impact consumer behaviour.
A new question arises, therefore, whether we should consider using transhipments across nodes as a way to rebalance the network. Again, this will depend on individual business requirements. Often rebalancing the nodes with transhipments will not be the optimal solution. However, there may be instances where it is. Often, transhipments cost additional money to the business in the form of transportation cost which cannot be passed directly to the consumer. While products are in the process of being moved, the inventory becomes unavailable for customer orders, reducing the overall availability of these products to consumers. These two factors, in addition to the fact it is impossible to predict where the inventory will be needed, implies it is most often best to keep inventory in its originally allocated location.
How will DCs be chosen to fulfil inventory for a given order?
Each DC will have multiple radii (see picture below - circular for simplification). The levels of radii will represent different costs of transportation from each node (or possibly a balance of cost vs CO2 emission calculations). The closest areas are the cheapest to ship to, getting more expensive with each area. The outer grey area would symbolise the area which is only fulfilled from this node if no other node has inventory available, and therefore this node is the only option (even if it is not cost-effective).
Based on the system being multi-node this means that radii will inevitably overlap, causing a Venn-diagram effect (see image below). The image shows that the cheapest regions will rarely overlap, meaning it is almost always best to fulfil orders from the closest node. However, when radii overlap, additional calculations will need to be taken into consideration to optimise which node will fulfil a given order.
What additional constraints need to be considered when setting up the allocation of orders algorithm?
When node's radii of cost overlap then these are some of the additional factors which should be considered in the algorithm which allocates orders to DCs:
Protecting inventory in the central node – when is it better to protect this inventory? If a new range is coming out, it might be better to use this inventory to utilise space for new inventory to be stored in the central node.
Is it sometimes better to fulfil a product from a more expensive node, to protect inventory from the cheaper node?
When are new deliveries arriving at the DC and what inventory is it? If we are restocking the current inventory, we wouldn’t need to protect inventory as it wouldn’t become out-of-stock before the new inventory arrives.
Are there one-off promotional offers coming up, if so, how is this forecasted to be sold by location? Do we need to protect those nodes in anticipation of the upcoming event?
How do we reroute orders if a node goes offline? How does this differ if the node going offline is planned compared to a one-off event e.g., planned maintenance work vs a power cut?
Do the same principles apply to rerouting if a node is running at reduced capacity? Should a company tell customers their deliveries will take longer, or should companies take the hit of a higher transportation cost to ensure customer satisfaction?
It is important to note that all these considerations need to be reviewed by the business at regular intervals, as well as having the need for override capabilities for when one-off changes are needed.
As we can see, there are a lot of considerations needed when a multi-node network is set up, and these considerations need to be continually reviewed and analysed to ensure optimisation for both the business and the customer. It requires a lot of planning and a lot of business-level decisions to be effective. Regardless, the rapid increase in the e- Comm space over the last few years means multi-node networks are now a tool that businesses urgently need to set up, to keep up with demand.
At ebp Global, we are currently working with clients to help them set up effective multi-node networks. If your business is also interested in learning more about this, or how we can help, please reach out to us at email@example.com.