The Demand Driven Supply Chain. Part 1 – What is Demand?

Unless you’ve been living under a proverbial rock for the last eight years, you must have heard of the Demand Driven Supply Chain. But if you ask ten different supply chain practitioners what it means to be “demand-driven”, you will get fourteen different answers. Twenty-four, if any of those questioned were consultants.

Some would argue that it’s a buzzword. Old wine in a new bottle. Haven’t we always been demand driven? Isn’t good old MRP demand driven? You start with a forecast and you blow through a bill of distribution and bill of material. Net any inventory you have or will have, and viola – you have a demand driven supply chain.

Others would say that MRP is not demand driven at all since it uses a ‘push’ mechanism; and instead the Japanese had it right when they introduced the concept of ‘Just-In-Time’ (JIT). Can’t get any more demand driven then a pure ‘pull’ model.

Then there is the new (relatively) kid on the block. Demand Driven MRP or DDMRP. It has demand driven right in its name to drive home the point. It uses a decoupling strategy to reduce lead times with strategic buffer positioning along with a prioritization mechanism to drive replenishment of buffers.

Discussions around the demand driven supply chain topic can be passionate, with supporters and critics each making their points. I will not be rehashing those debates today. Instead I want to see if we can unpack the term ‘driven-driven’. I aim not to provide clear-cut answers but instead to provoke questions so we can approach the topic with critical reasoning. This is a big and complex topic, so this will be the first of a series of articles.

What is Demand

In this first part, we start with the basics – What is Demand? Each of the supply chain methodologies listed above have their own definition.

MRP considers both actual sales orders and forecast as ‘demand’, using forecast consumption strategies to avoid double counting. It then propagates this as dependent demand through the network and bills of material.

JIT consider Kanban signals (an empty bin or a Kanban card) from the next work center as an indicator of demand. In a JIT environment, forecasting may still be carried out but it is used typically for longer-term strategic planning (and not for operational planning). There is a significant effort to stabilize customer demand in the short and medium term by utilizing pricing or other incentives. Sometimes a short- term forecast may be used as a basis to adjust the number of Kanban cards.

DDMRP considers only sales orders as true demand. And its relationship to forecasts is best described as, ‘It’s complicated’! Forecasts are one input used to adjust historical average daily usage calculations and make it forward looking. DDMRP then takes this ‘projected’ average daily usage and sets buffer levels at strategically determined decoupling points on a periodic basis. But once buffer levels are determined, replenishment planning (the DDMRP run) does not directly consider forecasts as source of demand. This has the potential to reduce the bullwhip effect downstream caused by inaccuracy in forecasts.

As you can see, each of the above philosophies can make a plausible claim to being demand driven in their own definition of the term.

When is an Order not an Order (or a Forecast not a Forecast)?

The question of what constitutes demand is not as straightforward as it first seems. Take the traditional distinction between a customer (sales) order and a forecast. Every ERP on the planet follows the same strict distinction between the two; and this has been a fixture of supply chain applications forever.

But take the example of industries that sell make-to-order or configure-to-order products. During their long sales cycles, they typically have opportunities and quotes that are tracked in some kind of a CRM system. Some of these quotes and opportunities have a high probability of conversion and these fall into a gray zone. They are not, strictly speaking, sales orders (yet) since they don’t have a customer purchase orders associated with them. But for all practical purposes, these high likelihood opportunities are treated as confirmed orders often with material or capacity reservations made against them.

Sometimes there is the converse scenario, where sellers may get a purchase order from their customer - but that order is more of long-term contract or a blanket order that will specify pricing and other commercial terms. The delivery dates/ quantities and detailed configurations may be part of a separate delivery schedule or call-off orders. It may not make much sense to treat these contracts/ blanket orders as customer orders from a supply chain perspective.

Or take the example of sales orders on some kind of hold. Say a credit hold or other hold for commercial reasons (export documentation for example). Is it an order for supply chain purposes? Kind of. Maybe. Sometimes.

Customer orders are traditionally considered ‘firm’ demand whereas forecasts are treated as fickle. The reality is that customer orders change and get cancelled all the time. Delivery dates, quantities, SKUs or configurations – everything that can change, will change. The longer the delivery lead times, the more the changes. Even with all those legal sounding terms and conditions in fine print at the bottom of every contract, changes are inevitable and frequent. Just ask Boeing.

On closer analysis, the clear distinction between orders and forecasts does not even hold even for CPG companies – retailers typically enforce a ‘fill-or-kill’ policy, which means any backorder quantities are automatically canceled. The demand represented by the cancelled backorder quantity does not just disappear – it is potential future demand (forecast) or is absorbed into a new customer order.

As you can see, an order is not always an order and a forecast isn’t always a forecast!

A New Definition

Maybe then in this digital age it is time to go past the traditional definition of sales orders vs forecasts and understand what digital natives think of demand.

Do you think Amazon ignores what is in your cart but not yet ordered? Do you think they ignore what’s in your ‘saved for later’ or what’s in your ‘wish list’ or even what’s in your search history? After all, you haven’t ordered any of that yet. Of course not! Amazon is able to do what they do because they ignore nothing. Every click you make is tracked. Every input is attuned, anonymized, adjusted and aggregated based on the probable impact on the demand signal. Nothing short of this can provide the kind of extreme lead times that Amazon is known for. [Tons of inventory can help in the short term, but it’s not very sustainable!].

This gives us a hint as to what demand really is from a supply chain perspective – it’s a signal. Nothing more, nothing less. For every supply chain node (say product-customer or product-location), it’s the best signal that you can identify based on various inputs that are available for that node.

Forecasts and orders are just proxies for something much more complex - for a product sold to a customer, the demand signal is an amalgamation of a myriad set of inputs - historical sales, open orders, lost demand due to stockouts or hard allocations, cannibalization, quotes/opportunities, promotions, VMI forecasts, customer forecasts, new product forecasts, etc. Not to ignore Point-of-Sale (POS) data, syndicated data (including competitor data) and other drivers such as demographic, weather, social media sentiment, etc. Phew! Quite a long list there.

At some point the law of diminishing returns will catch up with you and adding inputs will not materially impact the demand signal - but most organizations are very far away from that. It is easy to give up and say that forecasts are always wrong. But today’s supply chain battles are fought on the basis of reducing lead times with minimal cost-to-serve, and getting the external demand signal right is a critical first step.

As the number of inputs increases it becomes increasingly difficult or even impossible for demand planners to keep up. We are talking about thousands of SKUs and millions of data points. This is where technology enabled tools such as touchless demand planning can help augment planner capabilities so that humans can focus on true exceptions and on improving the process.

But Wait, there is More!

We saw how approaches such as MRP, DDMRP and JIT differ in their treatment of the definition of demand. But they also differ in an even more fundamental way – their philosophy on how to respond to demand and how to propagate demand upstream across bills of materials and networks.

Unfortunately, I am coming up against the word limit for LinkedIn articles (and more importantly, readers attention spans!). So that will have to wait for subsequent parts of this article.

What do you think of the Demand Driven Supply Chain? Let me know in the comments below.

#SupplyChain #DemandDriven #Signal #TouchlessPlanning #AutonomousPlanning #DigitalTransformation

Disclaimer: This publication does not represent the thoughts or opinions of my employer. It is solely based on my personal views and as such, should not be a substitute for professional advice.