Is there a Science behind Supply Chain Management?

Does SCM have a Science?

Disciplines such as Mechanical, Electrical and Civil Engineering are based upon Physics but the practice of Supply Chain Management (SCM) appears not to have any foundational science. Instead SCM is subject to various fashions/trends/fads/acronyms that, at various times, come to dominate its application and practice. Examples include JIT/Lean (1980s to 2000s), TOC (also 1980s to 2000s), S&OP (1970s to 1990s), IBP (2000s to present), VMI (1980s), CPFR (1990s to 2000s), MRP/MRPII (1960s to 1980s), ERP (1990s), APS (2000s to 2010s), Digitalisation/AI 2020s........).

Does this matter?

If Electrical, Mechanical and Civil Engineers didn't have a foundational science to guide them in their work, our lives, as users of their 'products', would be in serious danger. The consequence of incorrect decision making by Supply Chain leaders is rather less serious.......just bad investment decisions (eg. re IT expenditure), poor customer service, poor cash flow, excessive costs and wasted resources.

How would a foundational science make a difference for those working in Supply Chain Management? Certainly it would help them to cut through the confusing avalanche of management acronyms and discern the real value of advice from consultants/gurus/Gartner/thought leaders/influencers; more importantly it would provide an unambiguous framework for guiding performance improvement.

The Science behind Supply Chain Management

You maybe surprised to learn that SCM does have a foundational science but its little known about.

The issue is addressed by Industrial Engineering academics Professors Wallace Hopp and Mark Spearman (authors of Factory Physics) in their 2021 paper The Case for a Unified Science of Operations. In it they define an operation as

'....an activity that utilizes resources to transform one or more attributes of an entity or set of entities into some good or service that is required to satisfy some external demand'.

and go on to say that:

"...because we know from empirical observation that there is always variability in demand and transformation, it is virtually never the case that demand and transformation occur at precisely the same time. If transformation occurs before demand, we create inventory and if transformation occurs after demand, there will be some amount of wait time between demand occurrence and satisfaction"

The following graphic of an elementary operation helps illustrates these fundamentals:

"If a demand arrives (from the right) to find a suitable item in stock it is satisfied immediately. If not, the demand enters the queue on the left and awaits the transformation activity which uses resource capacity to satisfy demands in the queue".

Based upon this basic model/system:

".......a central challenge is understanding the relationships between demand and transformation and how these are affected by variability (because) variability disrupts the synchronization between demand and transformation, which degrades system performance.

How variability affects performance of an operations system depends on how it is buffered. In the above elementary operation, if we do not do anything proactive, the lack of synchronization between demand and transformation caused by variability will cause stocks to build up when the transformation rate exceeds the demand rate and backorders and/or lost demand to accumulate when the demand rate outstrips the transformation rate. We can reduce the amount of customer waiting and/or lost demand by increasing the inventory buffer by setting higher stock targets. Conversely, if we reduce the stock targets too much, and hence the inventory, the time buffer in the form of waiting and/or lost demand will grow. Finally, we can reduce both inventory and time buffers by increasing transformation capacity to enable us to keep up with demand surges or capacity contractions".

The value of an operations science is its ability to predict the consequences of changes to an operation/system and thereby serve as a decision guide in their design, control, and management. Achievement of such predictive power requires an understanding of the fundamental relationships between demand, transformation and variability.

The key relationships are described below - some are statements of the obvious and others less so, they all stem from the principles of queueing theory which is the study of flow/waiting lines and, from the perspective of Operations and SCM, are covered comprehensively in Factory Physics and Supply Chain Science:

Little’s Tautology: over the long-term, average inventory, throughput, and lead time for any stable process are related according to: Inventory = Throughput × Lead-time and, conversely, Lead-time = Inventory / Throughput

Supply Chain Variability: occurs when the rate of transformation differs from the rate of demand. There are two sources of variability 1. fluctuations in the rate of material arrivals at a work-centre, relative to demand, and 2. fluctuations in the rate of transformation by the work-centre, relative to demand

Law of Variability: increasing variability degrades the performance of a production system because variability in a supply chain system will always be buffered by some combination of inventory, capacity and time. (for more on variability see SC Variability - what it is, why its bad & how it can be minimised)

Bottleneck Definition: 1. the bottleneck in a production or service network is the resource (node) with the highest long-term utilization. With a 100% loading its lead-time grows without limit due to the impact of variability 2. accumulation of inventory is not necessarily an indication of a bottleneck (it could, for instance, be due to upstream batching policies)

Capacity Principle: the output of a system cannot exceed its capacity and, due to variability, it is impossible to achieve the theoretical maximum throughput.

Laws of Queuing: 1. if a work centre increases utilisation without making any other changes average lead-time and inventory will increase together in a highly non-linear fashion 2. at any given level of work centre utilisation, average lead-time and inventory will increase together in direct proportion with any increase in variability. These relationships are described by this formula and are shown in the below graphic:

Variability Pooling: combining sources of variability so that they can share a common buffer reduces the total amount of buffering required to achieve a given level of performance.

Buffer-Flexibility Law: flexibility reduces the amount of variability buffering required in a production system

Key Lessons from SCM Science

Some insights from the science are more obvious than others. For instance the Capacity Law demonstrates why S&OP / IBP capacity planning should focus upon bottleneck work centres and why TOC places so much emphasize upon bottleneck management. The Buffer-Flexibility Law explains the value of postponement strategies and use of standardised components to reduce buffer through variability pooling; it also explains the value of cross training labour resources across multiple work centres to provide capacity resources that can synchronise supply with demand variability.

Queuing theory also provides an interesting perspective on big batches, they are actually a source of arrivals variability generating wait time and inventory buffers. By contrast, if the ideal of a batch size of one that always arrived in line with demand were to be achieved (and without any transformation variability), each piece would move through processing/transformation without any static inventory and waiting/queuing time (ie. perfect flow). Less obvious perhaps is the impact of variability upon supply chain efficiency/performance. For instance, attempts to plan work centre activity to very high utilisation rates is a mistake because of the consequent rapid growth in lead-time and inventory buffers.

A further illustration of the importance of variability to SCM is that Hopp & Spearman define Lean as being (1):

"....fundamentally about minimising the costs of buffering variability", those costs being lead-time, inventory & capacity.

Through this perspective we can begin to understand why the tools that come under the Lean umbrella are so very effective - TPM, TQM, Standard Work, Poke Yoke, 5S are all about helping materials to flow reliably through work centres without processing/transformation variability; SMED enables batch size reduction without loss of capacity that reduces arrivals variability; Lean Pull is critical to minimising variability / delivering flow because it is the only replenishment methodology that attempts to synchronise material transformation with demand (it is also worth noting that the founders of the Toyota Production System knew all about variability elimination - though muda means 'waste', its components: muri and mura, translate as 'overburden' and 'un-eveness' respectively).

And it is perhaps ignorance of supply chain science that has contributed to the prolific use of forecast-driven DRP/MRP/ERP/APS systems across industry. It is maybe intuitive to believe that we should plan, or 'push', replenishment with master production schedules using our best guess of what demand will be in the future - and use safety stock to buffer the error. Unfortunately this methodology actively degrades supply chain performance because Planners are also encouraged to forecast backorders, using those same inaccurate forecasts (usually with the help of MRP exception reports), and implement backorder averting expedites. These generate variability buffers because the interruptions delay processing (time), cause stock congestion, waste capacity and, because they're also increasing lead-times, encourage further backorder averting interruptions thereby generating a self-perpetuating vicious circle of behaviour that causes all 3 buffers to be excessively high and ever present. You can see how schedule interventions generate these buffers at Are all Supply Planners driving red cars? and you can estimate the amount of excess inventory buffer you hold in your own supply chain by comparing the ￡/$ value of your actual inventory with the theoretical calculated from aggregating, for all sku's, their ￡/$ Safety Stock+1/2 Ave. Order Quantity (for a more thorough explanation see Factory flow is non-linear so don't use master production schedules and/or The SC replenishment problem (and how to solve it)

Going Forward

Supply Chain Management's foundational science supports the application of Lean and TOC and also explains why their performance improvement potential has been inhibited by the widespread implementation of forecast-push DRP/MRP/ERP/APS.

With greater and more widespread knowledge of the science behind material flow, the mistakes of the past would be learnt from and future supply chain performance improvement activity guided more accurately. Variability minimisation is key and this is particularly important now that AI and digitalisation have become so prominent. Lora Cecere provides some very pertinent advice (see Please Don't AI This) warning that "applying AI to our current broken processes has the risk of only helping us to make bad decisions faster.......there is a reason why the most used technology by planners is a spreadsheet".

On a more positive note, more application of operations science would significantly improve the performance of global supply chains and, for those working in the area, improve their chances of reaching the C suite based upon unequivocal performance improvement. For science based advice re how your next SCM IT investment can help deliver material flow (with all its attendant service, inventory, lead-time and cost benefits) and forward planning capability using Enterprise-wide Pull and without making your current ERP redundant, see Neither water or supply chains need 'Big Tech' to tell them how to flow

(1) - See Hopp & Spearman's To pull or not to pull: what is the question? and for an excellent in-depth review of all aspects of Lean from the perspective of operations science by the same authors, see The lenses of lean: Visioning the science and practice of efficiency.

See also Factory Physics was once all the rage but I don't hear about it any more - was it wrong? by Michael Balle, (author of The Gold Mine) in which he admits that "Lean in the 2020s has probably gone so overboard with coaching, mentoring and so on that we’ve forgotten to teach the basic mental models people need to understand about what goes on around them – how the system behaves 'mechanically'.........thinking of stocks, flows, instructions, and variability. Time to revisit 'factory physics,' I guess!"