Backorders represent purchase orders made to the supplier for products that are already out of stock from a given location being served. Backordering is the process of selling inventory the company doesn’t have on hand. Backordering takes place only when the demand is captured in a formal manner: for example, in a retail store, most customers would simply move on when facing an out-of-shelf situation, without reporting the missing product to the store. Backorders represent specific challenges in terms of inventory optimization, as backordered units are typically associated with a degree of urgency coming from the client.
Overview of backorders
Backordering represents a way for clients to order a unit that is not presently available. This situation often arises in B2B sales contexts. For example, a pool of aircraft parts should be able to serve all the parts requested; and any request that cannot be immediately fulfilled from the stock on hand will trigger a backorder.
Backorders can also take place in B2C contexts, typically in e-commerce. Typically, the product will be flagged as available “within 2 weeks” or any other similar timeframe that represents the e-retailer’s best guess about how much time it will take to have the product delivered at the time the order is being made by the customer. Expensive products are also frequently backordered from the store with the help of a store clerk.
In B2C, the ordered product may not always be delivered to the customer in the end because the costs involved in delivering the product may vastly exceed the benefits (see the discussion on MOQs below). Therefore, when it is indeed possible to have recourse to backorders, it is advised to have a process in place to handle, as gracefully as possible, the situations where the product will not be delivered to the client within the originally advertised timeframe. This process typically includes a proactive refund for the product, but may also include an additional gift voucher as a compensation for the unfulfilled order.
Quantitative modeling of backorders
The backordering process is nearly always linked to clients experiencing extra sensitive to the actual duration of stock-outs. Indeed, with backorders, clients are taking an upfront commitment for purchasing a product that is not readily available, and extended product unavailability is going to be perceived as a lack of good service provided by the distributor.
From inventory control viewpoint, backorders are typically represented as negative values within the available stock. The available stock should not be confused with the stock on hand which represents the quantity of stock physically present on the shelf. By definition, the stock on hand value cannot drop lower than zero; while the stock available can take both positive and negative values.
From an inventory optimization viewpoint, when modeling the impact of backorders using the stock reward function, the economic penalty associated with stock-outs in the specific case of backorders is typically assumed to be quite large, possibly equal or larger than the selling price of the product itself.
MOQs and backorders
When minimum order quantities (MOQs) are present, they typically interfere with backorders: in this case, it is not possible to make a purchase order that exactly matches the backorder quantities since the purchase order needs to satisfy the MOQ constraints as well. When MOQs are large, it is not always a reasonable economic option to fulfill every single backorder because satisfying the MOQ constraint may result in creating a lot of dead stock.
The first step for avoiding this situation consists in refining the calculation that defines the availability of the product taking into account the MOQ constraint. The MOQ constraint is used to compute an advertized shipping delay aligned with the real delay that the customer will most likely have to face if she takes the option of backordering the product.
Furthermore, a prioritized ordering policy needs to be used to correctly model the impact of the extra stock-out penalty generated by the backorders. In fact, order point inventory policies cannot properly handle this type of multi-product constraints since MOQs are typically satisfied not only by ordering the products presently associated with backorders, but also by spreading the quantities over multiple other products that happen to have lower inventory levels as well.
Returns and backorders
Some verticals, like fashion for example, get a steady amount of returns, sometimes representing up to 50% of the original demand. In such situations, a backorder might have a good probability of being fulfilled through the expected returns, and thus, might not even need an actual purchase order to be made to the supplier.
From an inventory optimization viewpoint, this situation is handled by combining both the probabilistic forecast of the demand with the probabilistic forecast of the returns. The two probabilistic forecasts are combined through a convolution which gives the net demand; a distribution of probabilities where negative demand values are possible and represent situations where the returns might temporally outweigh the demand.
Once again, a prioritized ordering policy needs to be used to properly unify the impact of backorders when returns are present at the same time. It also possible to combine returns, MOQs and backorders, but this typically requires the use of a dedicated numerical solver.
Containers and backorders
Some distributors prefer to blur the line between backorders and regular orders when selling inventory while it is still being shipped by sea. Indeed, overseas imports typically involve very long lead times, up to 10 weeks or more. However, a precise analysis of the incoming stream of containers offers the possibility to sell products with substantially lower lead times, while not selling products that are effectively in stock.
Selling inventory that is still in transit stage represents significant advantages for distributors from a cash flow perspective, and also contributes to decreasing the overall inventory risk by reducing a company’s overall commitment relating to unsold inventory quantities.
In practice, selling inventory in transit, such as when containers are still at sea, requires reliable probabilistic lead time forecasts, because the calculation of the shipping times advertised to customers requires a fine-tuned modelization of the economic risks associated with an incorrect estimation of the supplier lead times.