In conversations with prospects and customers, these questions often come up: Should we not forecast based on true demand, meaning what the customers want from us? Does our shipment history represent true demand, if it captures negotiated quantities, dates, and/or simply supply issues?
It is difficult to argue against this idea of using true demand as the basis for forecasting. If one is going to forecast demand into the future, it would make sense to get as true a picture as possible. For that, starting with historical demand (and not sales or shipments or invoices) would be the obvious choice. Choosing this data as the basis for forecasting would ensure the best possible projection in the future.
If the true demand data is available, one should of course use it as the basis for forecasting. I have no quarrel with that. However, in 95+% of the applications that I have seen, either order history or shipment history is used as the base data for forecasting. Why do companies not use the true demand? For the most part, no one really knows the true demand. Why so? Well, let us look at some of the reasons.
Negotiation During Order Taking
Suppose a customer calls to place an order, and the Customer Service Representative (CSR) knows that the product is not available in the right amount before the date requested. There is typically a quick back and forth about what might be an acceptable compromise. At the end of it, a different quantity and/or date is agreed to. At this point, what gets recorded in the ERP system as the quantity and the date? My experience working with many different companies suggests it is the negotiated values. As an extension of this, if the product is simply not available, many times the order does not even get recorded. The equivalent of this in the retail space is when there is no product on the shelf. How does the store know how many people wanted to buy it right then and in what amount?
In fact, it does not end there. Sometimes, as the due date approaches, the product is still not available. In such a situation, the CSR must call the customer to say that the delivery would be late. Is this recorded as a separate renegotiated quantity and/or date? Hardly, if ever. What I see is that the original values are overwritten.
In both these cases, what we see recorded in the ERP is very often not the true demand but instead the negotiated demand.
The remedy for this situation is rather simple: Use different fields in the ERP to capture ordered, negotiated, and re-negotiated quantities and/or dates.
This happens a lot in the chemical industry. In principle, it is like Vendor Managed Inventory (VMI).
In VMI, the vendor or the supplier oversees managing the inventory at the customer site. In exchange, the customer provides the data on the latest sales and current stock to the vendor.
Consignment is like VMI but there are significant differences. The inventory at the customer site is owned by the customer, for example, a rail car worth of supplies. Customer ‘taps’ the inventory as needed. However, the information flows back to the vendor on a periodic basis as opposed to in real-time. Once a month is quite common. This leads to a real gap in when the actual demand occurred and when the demand was reported.
Here is how Peter Edwards suggests that one handle this situation:
On consignment, I’ve always looked at this as vendor inventory at the customer site. In determining the appetite to enter into consignment, I ask:
- What is our strategic position at the account? (ideally an important customer)
- What is our visibility to inventory and demand? (ideally real-time or daily)
- If we need the inventory, will the customer load our truck? (ideally yes)
One should enter into consignment only when the answers to the above questions are favorable to doing so.
Customer Ordering Too Much
Let us say that a particular supplier is often short of product and negotiates the quantities down. Customers might learn from this and modify their behavior and begin to order more than what they need. The hope there would be to get an amount closer to the actual need post-negotiation.
A similar situation might happen in the marketplace when there is a panic situation going on. (This has been a reality for many products in the last 18 months or so because of COVID). There might be real supply disruptions leading to allocation, or bullwhip effect, or just unfounded panic. In all these cases, customers might modify their ordering behavior and order more than what they really need.
In these situations, even the ordered quantity does not represent true demand. It represents a higher number from where the negotiation will begin.
This type of pain can be self-inflicted. A business might decide to give out volume discounts to the customers. One way this is done is through blanket orders, where the customer places an order large enough to avail the volume discount. However, the actual delivery is done with multiple small orders which are deducted from the blanket order. These small orders are moved around multiple times before the delivery is taken. How does one determine the true demand in this scenario?
I am not aware of any remedies (simple or otherwise) for the last two cases.
What to Do
If true demand data is available, one should always use it. However, one cannot allow perfect to be the enemy of the good. If the true demand data is simply not available because of above mentioned or any reasons, one could look at what is available and make the best use of that data. Order history, shipment history and delivery history can all be useful, though with their unique flaws.
What do you use as a basis for forecasting? Please let us know in the comments.