feature – the release of christoph roser’s new book inspires john shook to discuss the origins and true meaning of “pull” and why it is incorrect to blame jit for the shortcomings of global supply chains.
words: john shook, chairman, lean global network
one thing covid-19 has unequivocally proved is that global supply chains in every industry are broken. when the pandemic began, we saw this in the empty shelves of supermarkets around the world. today, we see this in the painfully slow vaccine rollout experienced by many countries as a result of supply issues. the fact that global supply chains are broken is not new (although current conditions seem to be even more dire than usual), but the fact that now everyone knows they are broken is. also not new is the fact that the method that could dramatically improve the situation is getting blamed for the failure: the pull system devised by toyota more than half a century ago, known as just-in-time. so it is timely indeed that christoph roser has launched his new book, all about pull production.
what is a pull system? here’s a few words of context.
roser’s book dives deeply into the ins and outs of “pull”, a method of matching supply with demand that is widely referenced but poorly understood. the commonly accepted academic definition of pull has been offered by wally hopp and mark spearman, who claim in their influential 2004 paper (manufacturing and service operations vol. 6, no. 2, spring 2004, pp. 133–148) that the defining characteristic of a pull system is the presence of inventory control limits: “a pull production system is one that explicitly limits the amount of work in process that can be in the system.”
that is an attractively simple definition; however, it is incorrect. toyota developed “pull” and supply chain design through real-world practice at a time when it had no access to automotive suppliers and therefore needed to innovate. inventory was not the main problem they were trying to solve (nor is it today). the critical supply chain (or, as lean thinkers prefer to call it, a value stream) problem to solve is that each point in the chain needs to know at any given moment exactly what to make and what to move. this does indeed entail the problem of inventory and inventory leads to many other associated problems (including carrying costs, among others). for this reason, pundits, academicians and even practitioners tend to focus on the piles of stuff we call inventory without looking further to understand the deeper causes. the piles are there. they continue to grow. they block our view.
a pull system, such as toyota’s broader system of just-in-time, approaches the problem of supply chain design not merely from the perspective of inventory control. how can we organize all the work, all the material, and all of the time entailed in a long supply chain so that it operates as effectively and efficiently as possible? how can we have each step producing precisely what the next step needs when it needs it, in the amount needed? is a top-down planning process the best way, informing each step in a large supply system what to make and when? such an approach might work, if everything went according to said plan. but, alas, in the real world, everything going according to plan is an increasingly rare occurrence. things go wrong. machines stop running. trucks break down. humans make mistakes. and, anyway, in today’s world, customers make last-minute decisions. people change their minds. so, how can we organize all this complexity when the beginning of the supply chain is half a world and half a year away from the customer?
these are the problems that pull systems tackle from an unconventional set of assumptions. minimum inventory is desirable, to be sure, but (sorry hopp and spearman) setting inventory limits is not the defining characteristic of a pull system. the amount of inventory needed is the result of actions taken and decisions geared to address multiple problems – mainly, how to ensure that what is needed gets to where it is needed when it is needed, neither too late nor too soon.
here is the definition of pull system from the : “a method of production control in which downstream activities signal their needs to upstream activities. pull production strives to eliminate overproduction and is one of the three major components of a complete just-in-time production system”. and note that, while minimal or even “zero inventory” might be “ideal”, in the real world inventories can be used strategically – utilized in six distinctly different ways (see ) – to make the overall system work with aim for greater “total system effectiveness and efficiency”, not “local or point efficiency”.
toyota discovered early on (as has, more recently, tesla) that the auto industry presented a multitude of deeply complex problems beyond the obvious ones of designing, building, and marketing the product. there are employees to hire and train, parts and materials to source, products to sell and service, regulatory concerns to contend with. and there is the thorny issue of managing the thousands of parts and materials that go into a car. as straightforward as it may sound, corralling and herding the endless stream of materials can be as difficult as transforming them into the product!
toyota auto company founder kiichiro toyoda came up with the concept – and even the curious name – of “just-in-time” in the 1930s on a reconnaissance mission to the uk. he missed a train connection, providing him a lesson he would never forget in the value of promptness and inspiring the insight that one should never fail to arrive where one needs to be just in time. as his company back home was figuring out how to actually build the automobiles that he designed, the factory was facing the problem of constantly running out of parts and materials. as a startup, the fledgling company didn’t have the capital to purchase large stockpiles (which the fledgling suppliers struggled to produce in quantity anyway). the answer was to bring in parts and materials daily, just in time, to be fabricated or assembled.
a decade later, kiichiro’s machine shop manager taiichi ohno began experimenting with simple pull systems on the plant floor. he found that workers at downstream processes were frequently out of the materials they needed to do their work, while upstream processes were busy producing the wrong thing. he decided to tackle the problem of “overproduction” (producing too much or, and this second point is overlooked by many, too soon) by reversing the direction of the flow of information – via kanban cards or simply the empty bins themselves – that instructed each operation what to produce and move (ohno didn’t invent pull; he developed the definitive, sustainably successful, pull system). (a note to lean geeks: value stream maps, aka material & information flow analysis, track the control or instruction information – not all the information that travels about in any work system. this is an especially critical distinction to make when “mapping” – to analyze and improve – a system in which the product of the system is a transformed form of information.)
in the end, two simple rules prevailed: 1) never make or move too much or 2) too little. simple indeed. but to actualize these simple rules in the real world requires turning things upside down: pull, don’t push. start by connecting individual processes. make production flow, ideally one-piece-at-a-time wherever possible. wherever it isn’t possible to flow from value creating step to value creating step, connect disparate processes through some sort of mechanism (such as kanban) to enable downstream processes to pull from upstream processes what they need when they need it. with those connections made, and single-piece flow established wherever possible, focus on shortening lead times between each step, resulting in a high-velocity and highly responsive value stream. simple! and brilliant!
but simple doesn’t mean easy and brilliant doesn’t necessarily mean intuitive.
my first exposure to toyota’s pull system was in the first half of 1984 on the stamping shop floor of toyota’s takaoka assembly plant. toyota was preparing for the beginning of operations at nummi (new united motor manufacturing inc.), the company’s joint venture with general motors in fremont, california. my aha moment – “oh, that’s what jit means!” – came after, not during, my plant training, which was when i was supposed to learn all about tps.
though i was still new to the process, i was already in the position of explaining tps to visitors, mostly from general motors. while i had learned enough to be able to walk through a basic explanation of kanban and “pull vs push”, i didn’t really get it. then one day, as i was being backed into a corner to explain, really, what pull was and why it was so important, i became frustrated with one particular production manager from gm as he was frustrated with me.
a practical mid-level manager with decades of experience, he pressured me: “i don’t get it. so what? i hear what you’re saying but i don’t see the big deal. why should i care about ‘pull instead of push’? i don’t see the advantage.” i was as frustrated as he, annoyed that he couldn’t understand the same explanations that had seemed to suffice for others. our frustrations rose as we egged each other on. i recall my face reddening as i realized that i couldn’t explain it well because i myself didn’t really know what the big deal was. i understood the rudiments in terms of the actions on the plant floor. and i understood to some extent the big-picture argument for what pull was and how it was unconventional, in abstract terms. but i didn’t know how to connect the dots and hadn’t fully grasped what was so revolutionary about jit and pull.
i think my frustration – first toward my guest and then gradually but steadily toward myself – was the spark that led to insight and my aha moment. i was suddenly, finally, struck by the power of all these quick cycle replenishment loops that were repeating seemingly endlessly throughout the entire supply chain. the small loop of material and information flow that we were observing was fractal. what we were witnessing at the end of a big press line as stamped steel was taken away every few minutes to a store located just a short walk away, which was then connected in a similar way to the process that followed (the body welding shop), was taking place all up and down the supply chain. the next day, we visited some outside suppliers connected with the assembly plant by kanban. those kanban were cycling between the two factories as frequently as eight times per day. it is a system with feedback loops, self-correcting and, indeed, revolutionary. systems within systems within a system.
within a few months, nummi was up and running and the world outside toyota was getting its first close look at a successful pull system in operation. still, what observers can see on the surface – such as low inventory – is only the result. the work behind the scenes to execute a system that looks so simple is in fact deceptively simple. toyota had established an entire department to innovate every aspect of the design and operation of its supply system. to this day, it is the rare company (especially outside the auto industry) that is willing to invest in developing the organizational capability required to successfully transform to a pull system. (toyota views conversion from push to pull as a defining characteristic of the higher-level functioning of its system of just-in-time – which is in turn a major component of its toyota production system – but pull is just one piece of jit, along with the concepts and practices of creating continuous flow with a defined takt time. a complete jit supply system – the right item in the right place in the right amount at the right time with no shortages and no overproduction – requires many additional pieces beyond pull (which is after all simply a technique), including the location and scale of capacity installation, supplier-oem relations, and, first and foremost, accurate and deep understanding of the needs of customers! (see toyota’s official definition of jit .)
misunderstandings, fallacies, confusion and damage done
it is unfortunate indeed that the most viable way forward to fix what ails supply chains is frequently blamed for their failure. academicians (such as those referenced above) do not have a monopoly on spreading misunderstandings about supply chain dynamics. periodically, even the most highly esteemed journals publish articles blaming supply problems on just in time or lean manufacturing. shortages of goods, such as those that followed 9/11 (jim womack and dan jones responded at that time ) or the from the massive earthquake and tsunami that struck japan in 2011 or the covid-19 pandemic (remarkably predicted with this from womack in 2006), invariably lead to nonsense such as in the wall street journal in late 2020: why are there still not enough paper towels? in it, the author writes: “blame lean manufacturing. a decades long effort to eke out more profit by keeping inventory low left many manufacturers unprepared when covid-19 struck. and production is unlikely to ramp up significantly any time soon.” (lei responded to this via a .)
now, in early 2021, the latest global supply chain crisis – progressing to items perhaps more challenging than toilet paper – entails semiconductors and geopolitics. policies that led to three decades of chasing lowest piece prices in china are being reversed. the pendulum swings again. through it all, there is a better way.
a way forward
enter christoph roser’s timely exposé all about pull. with his latest book, roser makes no pretense at giving us a sound bite for casual readers of 1,500-word simplifications. the content in these pages will require you practitioners to do some work.
roser introduces many techniques and tools that will help you to reconfigure supply chain operations. but more importantly, by exploring numerous approaches and methods, he challenges readers to rethink their strategy, to reconsider what their supply system needs to accomplish and why. how can you get each item to be in the right place at the right time? how can the thousands of people and processes of an extended value stream do the right work at the right time?
the true nature of the lean enterprise is a holistic business system. in it, everything is connected. therefore, the discipline of point optimization – the decades-long focus of supply chain research and a now-accepted dogma – typically creates waste everywhere in the system except the point being optimized. until supply chain leaders train their vision on the entire value stream, cost savings will be illusory and eye-candy improvements unsustainable.
the power of lean is realized at the gemba through the way in which activities are connected and uncoupling those connections – be they on the shop floor or in the extended value stream – destroys the key dynamic of a lean system: the ability to learn and adapt. break supply chain management into disconnected points of lowest-piece price locations or a set of black-box optimization algorithms and you will lose the ability to build a rational, practical, adaptable supply chain configured as a living learning system. ultimately, this challenges managers to deeply understand how their value streams currently work and to design improved versions of them that are manageable and adaptable to rapidly changing real-world conditions.
system designs that ignore the human side of decision-making and hope to control supply chains through technology-only or technology-mainly solutions will not solve our problems. centralized decisions (whether made by humans in command & control style or by machines using ai in black-box control style) are doomed for two reasons: 1) in today’s world they cannot match the dynamic adaptability of a distributed decision-making system, and 2) any time we divorce decision-making and ownership from the humans who do the work, we take away from those humans their sense of ownership, their engagement, and increase the likelihood that errors and even sabotage will occur as a result of alienation. conversely, pull via kanban or any other means – along with the other elements of a just-in-time production system, and, indeed, all the tools and methods of a lean system – cascades ownership to wherever the work is taking place. ohno spoke of “kanban democracy”, an elegant system with responsibility and authorization distributed dynamically based on need and transparent rules of engagement.
supply system designers have no end of problems to solve and approaches to tackle them. the shift to pull is a technical challenge but it also requires a fundamental shift in the mindsets of not only supply chain professionals but also of chief executives. and it requires teamwork on a massive level. there is no denying the need for effective supply systems that meet the rapidly evolving complexities of the 21st century. learning the principles and practices of pull is a great place to start.
christoph roser’s new book all about pull production is available for purchase .
john shook is the chairman of the lean global network.