The Lightbulb Moment: Moving beyond trying to predict variability

My recent article “DDMRP: A practitioner’s perspective” enabled me to articulate my perspective on the limitations of conventional planning methodologies, namely Material Requirements Planning (MRP) and Master Production Scheduling (MPS). Both of these methodologies are built on the premise of forecast driven supply order generation coupled to and dependent upon forecast accuracy. Alternatively, Demand Driven Material Requirement Planning (DDMRP) through decoupling and creating independence between supply order generation and forecasts results in improved supply chain performance.

The concepts that underpin DDMRP are built upon the foundations and principles that have been long promoted by Lean, formal planning (MRP/MPS), and Theory of Constraints. The innovation that makes DDMRP special is that these known and accepted principles have been brought together into a single integrated methodology that provides planning professionals with an effective and pragmatic method to better manage Planning and Execution decisions for Manufactured, Purchased, and Distributed parts.

‘The lightbulb moment’ to truly understand and embrace what makes DDMRP unique is the hurdle many people struggle to negotiate. Conventional planning methods, and everything that both I and others were traditionally taught as best practice, were based on the premise that the more accurate the demand signal, the better we can plan, and subsequently the better we can improve both service levels and inventory performance. This concept is perfectly rational and logical based upon the rules and tools that have proliferated industry since the 1970’s.

To explain this, let me boil a long and protracted journey of Supply Chain planning down to just a few points:

• Starting about 50 years ago, MRP systems utilized early computing power to take a Bills of Material, offset lead times, and do gross to net calculations to tell us what we needed to make/purchase and when. MRP was industries first attempt to utilize technology to drive improved supply chain performance.
• Master Production Scheduling was the next innovation as we moved into the MRPII days as industry first attempted to mitigate system nervousness, also known as the bullwhip effect, by decoupling a firm supply plan (MPS) from Demand (forecast), so as to reduce supply chain variability.
• Since the MPS was established technology has proliferated, much of it focused on how to improve forecast accuracy as without an accurate forecast companies are forced to compensate for forecast error with Safety Stock. This is where we now are in most of industry; we are still at the premise and understanding that came about over approximately 50 years of MRP that the better we can forecast, the better that we can manage our supply chain.

As forecasts are almost never precisely right, at best they are loosely indicative, the point we invariably get stuck at is that we need to better predict variability. The elimination of variability has long been known as the catalyst of supply chain performance: “All benefits will be directly related to the speed and flow of materials and information”, George Plossl, 1994.

I am sure that the majority of practitioners reading this article will relate to the seemingly repeated and never changing discussions they have on demand variation or forecast error.

The lightbulb moment where DDMRP advocates experience their “aha” moment is the premise that variability is not going anywhere, nor can we predict it, we need to accept that it occurs, and that better predicting it will bring us incremental improvements at best. This requires us to move past the deep-rooted belief, brought about by the history and development of formal planning systems, that forced us to try and better predict variability in order to achieve the desired service and inventory outcomes.

Remember, as explained above, the conventional wisdom that came about from the design of planning systems drove us to this point. By coupling and making all supply order generation through the supply chain dependent on a forecast, the seemingly logical and rational outcome was to focus on how can we make forecasts more accurate, and/or ‘optimize’ inventory by using statistical variation models constructed on the premise of mitigating forecast error.

Moving beyond trying to predict variability

Before proceeding further it is important to not create the impression that DDMRP is anti-forecasting. This couldn’t be further from the truth. Forecasting is a necessary and important business process that feeds resource planning decisions, budgeting processes, financial outlooks, etc., and therefore we do this in the different relevant ranges with the appropriate and necessary level of detail to support the effective running of the business. However, what we don’t need to do is directly tie supply order generation to forecasts. Committing company resources to forecasts is what drives sub-optimal outcomes in the operational relevant range.

Utilizing a forecast in DDMRP involves evaluating buffer profile adjustments that may be necessary to support different demand profiles, projecting critical resource loading to ensure the capacity available will be sufficient, and to subsequently act as an input to necessary Financial forecasts of revenue, costs, and cash. However, the key difference to MRP, that is certainly not trivial, is that the forecast does not drive supply order generation, only as actual REAL demand is realized does DDMRP drive supply order generation.

Rather than chasing the incremental gains that come from attempting to better predict variability in conventional planning approaches, DDMRP moves beyond this to accept that the variability will occur, decouple to not drive the dependencies of conventional MRP, and instead just use forecasts in a manner that is appropriate for something that we know will be wrong, that being, as just an input to our planning model, not the thing driving of it.

The DDMRP methodology and its detailed calculation methods are well articulated in published literature, white papers, case studies, as well as being supported by a growing list of technology providers. I encourage all to investigate these details further if you are not already familiar.

Concluding Comments

The beliefs and methodologies that have proliferated industry since the MRP crusade are inbuilt into our tools and belief structures to the point that embracing DDMRP is a challenge for many Supply Chain professionals. Advanced planning and scheduling tools (APS) and Forecasting systems may continue to add value to companies if implemented under the correct conditions, with all the required skills and prerequisites in place, however, there are many cases where these will not add the desired value and thus not provide the intended ROI.

Meanwhile new methods such as Probabalistic Forecasting may also create interest for planning professionals, and could offer some further advancements to DDMRP buffer design, particularly for environments with ‘long tails’ consisting of products with highly sporadic demand profiles.

What is clear however is that DDMRP breaks down the conventional belief that improving forecast accuracy is the key driver of improved supply chain outcomes. By moving beyond trying to better predict variability, and instead accept that it will occur, decouple, and not pass on variability in the form of coupled dependencies that occur in conventional planning, we can better align replenishment decisions and inventory profiles to actual customer demand, and not to forecast error that is then compensated for with Safety Stock.

Good luck with your lightbulb moment to move beyond trying to predict variability.

Matthew Gardner