The (customer order) decoupling point (CODP) or order penetration point is one of the most well-known logistical concepts. It indicates the inventory location in the value stream up to which the customer order penetrates. The CODP is one of the key design decisions in structuring the value stream and its management. For instance, the CODP separates the part of the value stream that is order- driven (downstream of the CODP), from the part that is forecast-driven (upstream of the CODP). But is that necessarily true? Doesn’t kanban, in fact, represent a good example of managing the upstream part of the value stream based upon consumption rather than forecast? And aren’t a lot of companies still clustering their customer orders downstream of the CODP or working with lot sizes, resulting in products manufactured too early, too late or just too much? This blog post nuances the existing control dogma in relation to the CODP.
The Customer Order Decoupling Point
Hoekstra and Romme introduced the (Customer Order) Decoupling Point in 1985 — closely related to Shingo’s earlier D:P ratio — as a concept to structure the value streams of different product-market combinations with regards to the role of stock points in the value stream and the use of control principles. It thereby also instantly became a concrete element in designing supply chains and value streams. One of the aspects related to the CODP that was introduced at the same time, was that the CODP also indicated the control principle to be used before (upstream) and after (downstream) of this main stock point. Hoekstra and Romme thereby state that activities upstream of the CODP are forecast-driven, and those that are downstream of the CODP are order-driven.
Since the end of the 1980’s these ideas have constituted the bases of supply chain thinking in the Netherlands. There is no self-respecting supply chain or logistics book that doesn’t refer to the concept in one way or another and there’s no good supply chain education that doesn’t pay attention to it. And rightfully so, I think, with regards to the design and configuration of supply chains. But with the increasing attention for, and adoption of, Lean principles, I feel it is time to revisit the statements concerning the control structure up- and downstream of the CODP.
Control Downstream of the CODP
Downstream of the CODP, Hoekstra and Romme -and many after them- state, the activities in the goods flow are planned and controlled based upon actual customer orders. And they should be. But some also tend to speak of order-driven being equivalent to “by definition working in pull flow”. This isn’t necessarily true, however. I see many examples in practice where order-driven activities in fact aren’t truly order-driven, nor are they driven by a true pull signal. The consequences are excess or even anonymous stocks downstream of the CODP and variable lead times with poor reliability towards customers. Three examples:
Clustering of customer orders
In order to create production-efficient lot sizes, customer orders for certain are being accumulated and then clustered by product type. In this way, changeover time is being economized, but lead times towards the market start varying and late deliveries or excess stocks result, depending upon either the production wheel in place or the pattern of incoming customer orders containing certain product types.
Minimum lot sizes
Another often witnessed practice is that production indeed is being triggered by a customer order for a certain product type, but that a full production lot of that type is subsequently produced. This then of course leads to anonymous surplus and complex order promising and scheduling the moment new customer orders involving this product will be received. Maybe the initial trigger for the lot makes you believe you’re producing-to-order, but you’re not of course.
Customer order “push”
Not contrary to the control concept introduced by Hoekstra and Romme, but still a undesirable form of “push” downstream of the CODP is the situation whereby indeed individual customer orders are processed, but that the related production triggers for the products are being released based upon a pre-determined schedule. As schedules (contrary to sequences) do not consider the actual state of the shop floor this often leads to early release of work to the shop floor, congestion downstream and excess stocks with even increased risks related to order cancellations.
In order to able to really work “to-order”, a lot size of one is a condition. Furthermore, in order to speak of “pull flow”, the sequence of order-related products needs to be released based upon the actual situation on the shop floor related to actual WIP and the truly available capacity downstream of the CODP. This sequenced and pulled flow can be concretely realized by making use of so-called “kanban squares”, generic kanban signals or POLCA-cards for instance.
Control Upstream of the CODP
Where I feel we really need to re-calibrate the statements of Hoekstra and Romme on control concepts, is upstream of the CODP. As said, according to Hoekstra and Romme and many after them, goods flow control upstream of the CODP is forecast-driven. This is a form of “push” where the release of work is governed by forecasts and assumptions about the status of the shop floor, and not considering actual consumption and availability of capacity on the shop floor. Strangely enough, most examples of “pull”-production exactly are examples of the use of product-specific kanban signals or “kanban squares” for anonymous production (so for which no customer order exists yet), so exactly in that part of the value stream that is upstream of the CODP.
Forecasts, however, are still necessary for managing the capacity both downstream and upstream of the CODP, but certainly not the only (or even desirable) method to control the actual goods flow and production of parts and products (or even component supplies and raw materials for that matter).
It surprises me that many still postulate that production and goods flow control upstream of the CODP takes place based upon forecasts, whereas there are so many companies that meanwhile have proven that “pull” may be a better strategy, resulting in lower inventories, more stable and controlled lead times, better delivery reliability and reduced managerial complexity with less overhead.
The above indicates that we should be more careful with plainly stating that upstream of the CODP we work forecast-driven, and that downstream we — by definition — work to order and “pull”. It may well be, as we have seen, that downstream of the CODP we are working in a “push”-mode with the negative consequences that go with this mode, and that upstream we extensively make use of “pull” in the management of the goods flow. Creating a truly Lean production system therefore demands a critical review of existing knowledge and assumptions that may exist in the organization or even beyond. You should carefully evaluate what moving towards “flow” and “pull” actually will mean for your value streams and your organization.
Great article to examine the assumptions around us in pull & push debate. I concur with you about your conclusion but at least I appreciate Hoekstra and Romme way of thinking. I was not aware of them.
In example 2 ‘Minimum lot sizes’, I assume that in your example above is that it is not batch process
by its nature. If so, why ‘Minimum lot sizes’ mean? why do it? what factors does trigger this behavior?
I can’t imagine that a downstream process after CODP that is not batch process and still commit to min minimum lot size approach.