top of page
Writer's pictureAlexei Zheglov

Heading West: Scale and Continuous Improvement

My popular longread Kanban and Scale presented a continuous improvement blueprint – a model of five dimensions. The model was essentially a categorization of field experience, patterns observed in companies of all shapes and sizes. The very expression “shapes and sizes” instead of just “sizes” suggests multiple dimensions are present. Scale reveals itself differently in each dimension, with unique opportunities, problems, and approaches. The relative importance (or irrelevance) of some dimensions can vary by company, pointing towards customizable improvement guidance emphasizing what’s important.


I now want to take a different angle and discuss the scientific underpinnings of the phenomenon of scale. What deeper insights and practical conclusions can we derive from the science of scale for the benefit of continuous improvement? As I expected, some findings do validate, invalidate, question, critique the existing work and suggest novel approaches.


First, I need to reduce the context to the field of continuous improvement of companies, the resulting consultant-client relationships, and the related professional communities. Next, within this context, I’ll show that we need to use the very word “scale” more intelligently and less naively. Further, we’ll explore different types of scaling laws, with relatable examples, principles from which these laws derive, and what accounts for the difference. This will lead us to somewhat unexpected conclusions regarding the improvement of our companies, particularly in the knowledge-intensive, intellectual industries of the XXI century. Lastly, some good news: effective practical approaches do exist. And they work because they’re largely in agreement with what the science of scale teaches us.


Context: Continuous Improvement and Consulting

Many companies and organizations in various industries desire some improvement, transformation, better ways of working, and so on. An industry of consultants exists to help them pursue such improvements. It consists of large management consulting firms as well as many smaller consulting practices, specializing in various industries and methodological niches. I have owned and operated such a practice for twelve years. Many large-enough companies have internal units dedicated to continuous improvement that go by endless variety of names, such as “quality management office”, “agility center of excellence” and so on. Many continuous improvement initiatives include some form of consultant-client relationship and several forms of knowledge transfer from consultants to clients, such as consulting per se, coaching and training.


This broad professional community, of course, continues to search for better ways. A casual observer can notice new books, methodologies, management fads, process frameworks, training classes, and certifications. Underneath the obvious, consultants owe it to their clients to be strong in both practice and theory. We should be able not only to discover a novel practice in the field, but also put it later on a solid scientific foundation. We should be able to lean on our foundations to predict something of practical importance, what makes good advice for a client. Outside billable hours, consultants owe it to their clients and to themselves to continue questioning their practices from the scientific angle. We must not to settle for known best practices and “it worked during my last gig.”


I once had to say in response to a client: “I don’t yet know enough specifics of your processes, but I know enough about the physics of flow or work in your industry. I predict that, if you looked up how long it takes your group to deliver stuff to customers, from “we promise” to “here it is”, and turned your data into a bar chart, the chart would look like this.” I sketched the chart on the whiteboard in our meeting room. I had an idealized variant of this chart printed on a card, which I gave to the client. When the actual data matched the chart’s contours, it wasn’t only that we matched practice with some theory. The client understood the meaning of key points on the chart and knew where to focus their next improvement actions.


One of the client employees later confessed to borrowing the card from a colleague’s desk and holding on to it longer than the colleague considered polite. I gave them the PDF file, so they could print as many copies as needed.


Continuing the dialogue in the continuous improvement professional communities, questioning our practices after 2017, we cannot ignore an important book published that year: Scale: The Universal Laws of Life, Growth, and Death in Organisms, Cities, and Companies by Professor Geoffrey West. This book is of great scientific depth, even though some depth is inevitably lost each time a scientist compacts, out of necessity and no matter how carefully, decades of research into 500 pages of non-fiction. The book also includes interdisciplinary breadth, intellectual weight, and literary aesthetics. It deserves a careful read. For professionals in the field of continuous improvement, it offers a much-needed collective look in the mirror. How does what we do stack up against real science?


You Keep Using This Word – Let’s Elevate the Conversation

The word “scale” is used a lot in the improvement context. It is often a proxy for the number of people involved, for example the number of a client company employees included in the scope of a consulting engagement. There are “scaled” methodologies and process frameworks. They embed an assumption: there is a simpler, “unscaled” methodology, which should work fine as long as the number is up to a certain limit. Beyond this limit, you’re supposed to use the methodology in a more elaborate “scaled” variant. Sometimes there are multiple “scaling approaches” to the same methodology, such as the popular software development framework Scrum.


Adding “at scale” at the end of a question, for example, “how does this work at scale?”, signals real or feigned skepticism. Sure, your suggestion is great, for the number you had in mind, but my number is greater than yours. Sometimes the word “scale” really means the number of billable hours in the transformation consulting engagement, for it is roughly proportional (scaling linearly) to the available budget, which is itself proportional to the headcount. Thus “scale” can also be the bragging rights from the biggest procurement win. You keep using this word. It doesn’t always mean what you think it means.


West’s Scale makes a clear impression from the very first pages, that we have to assign the very

word “scale” a more serious meaning than such naïve notions.


Without much complication, in a nutshell, scale is the phenomenon that, if we change one obvious parameter of a system, then what happens to some other important properties of the system? Do they change proportionally (linear scaling) or in some other fashion, following a different scaling law? Answers to such questions may be vital. For example, structural strength of buildings and bridges scales non-linearly to their size, such as the building’s height or the bridge’s span. Dosage of drugs is non-linear to the patient’s weight. Engineers and builders, biologists and doctors have figured out the scaling laws relevant in their professions. They have done better than us.


It is my hope that reading West’s Scale in the continuous improvement and management consulting communities will raise the maturity and the intellectual level of discussion of scale and even the very meaning of the word “scale.” Does the coordination cost scale linearly with the project’s budget? You wish. If it doesn’t, then when will this project collapse under its own overhead? Is the professional development expense per employee in the same industry invariant of the company size? That’s a more difficult question. A large company can give an employee a deeper professional network creating incremental knowledge over time to enhance the employee’s skill set. On the other hand, smaller companies come under greater survival stress. Adaptations to stress can produce the incremental knowledge faster than in larger companies. Do these two forces exactly offset each other? If we don’t understand these forces’ impacts, are we hiring people today who will predictably be dissatisfied with the company in three years?


Reading the book towards the middle and the end opens even more questions and opportunities. I believe it’s important to understand the principles and the main types of scaling laws. We have too much of one type of scaling already and not enough of the other. This already has huge consequences for continuous improvement of companies. My goal is not to retell the contents of West’s Scale or to give you a TLDR version of it. Maintaining the continuous improvement focus, I will now call a few selected examples and takeaways from the book into evidence. They will lead us to some unexpected conclusions.


Animals, Cities, Companies – And Principles

To help us understand different patterns of scale, West presents in his book Scale many phenomena found in animals, particularly mammals, and also in cities and companies.

The basic design of a mammal – a warm-blooded creature, supporting itself with a skeleton, enclosed in skin, having a heart, brain, lungs, circulatory, nervous, and digestive systems, etc. – scales naturally across several orders of magnitude, from whales to shrews. The blue whale, the largest of whales, can weigh up to 200 tons, one hundred million times more than the Etruscan shrew, the smallest mammal weighing only two grams.


Scalability across eight orders of magnitude puts to shame naïve process methodologies whose adherents assume they need a scaling framework when the size of their project group changes from twenty people to one hundred.


Do all characteristics of an animal scale linearly in proportion to its size? Of course, not. Structural strength increases slower than weight. There is a physical limit, beyond which the bones cannot support the animal’s own weight. Elephants, the largest land animals, live near this limit. Marine animals are free from this constraint, but they run into a different one. The space between adjacent capillaries of the animals’ circulatory system increases with size, too, although very slowly. Eventually, when the capillaries are too far apart, the tissue between them cannot get enough oxygen to stay alive. This factor limits the blue whale’s size. At the opposite end of the scale, if you make blood vessels any narrower than the shrew’s, pulsatile blood flow becomes impossible due to blood’s viscosity.


Metabolism, an animal’s ability to convert food into energy to power its life activities, is of particular interest and importance. Elephants are about ten thousand times larger than rats by weight, but their metabolic rate is only one thousand times more. Increasing one parameter by four orders of magnitude translates into an increase of another parameter by only three orders of magnitude. This is called a law of three-fourths in mathematical shorthand. Of all possible laws of three-fourths, the one in biology that pertains to animal metabolism is known as Kleiber’s Law.

Ancient farmers knew this law intuitively. If you need to plow a field, using the largest available domesticated animals, such as oxen or draft horses, is the most efficient.


While this design is scalable, it is not scale-free or scale-independent. For example, a tiger is a scale-up of your house cat, weighing 100 times as much. But most of the tiger’s characteristics aren’t 100 times greater. For one, it needs only 32 times as much food (Kleiber’s Law of three-fourths). Its heart beats three times less frequently (the law of inverse one-fourth), pumping blood through the aorta about 5 and a half times wider in diameter (the law of three-eights). And its blood pressure is about the same – scale-invariant.


Larger animals typically enjoy longer life spans, have lower heart rates, and can be described as living at a slower pace. Lumbering elephants, moose often stopping to think, mice and squirrels running constantly in seemingly random directions. Let’s keep these images of animals in motion in our minds for a short while. They will become important shortly.


As fascinating as these facts about animal anatomy and physiology may be, Prof West is not content with observing them. His journey and identity as a scientist were key in his quest for the underlying theory. He earned his doctorate and made a career first as a theoretical physicist studying elementary particles. Then he transitioned to interdisciplinary studies, including his time as the director of Santa Fe Institute. Immersing himself in biology, one of the “inter” disciplines, he was surprised to find it different from physics: skewed towards experimentation and lacking a strong theoretical component.


Physics has for a long time had two strong branches, theoretical and experimental. It is quite easy for a layperson, very far from any science, to imagine a theoretical physicist solving complicated equations in their study, predicting a phenomenon discovered later in nature or in a lab experiment. The most recent popular example of this image could be J Robert Oppenheimer, portrayed by actor Cillian Murphy, writing formulas on a chalkboard in the eponymous Oscar-winning movie. While there are empirically discovered laws of planetary motion, these laws also have a rigorous mathematical derivation from very few first principles, provided by Sir Isaac Newton. The theoretical-experimental duality of physics and the dialog between the two branches existed since at least the XVII century. This is different from how Prof West found the late XX century biology. He saw the lack of a rigorous theoretical foundation as an opportunity.


My friends in the field of bioinformatics may object to this notion. They work in offices resembling a software company more than a biology lab. Large multi-monitor setups on work desks. The primary type of vessel where they can grow bacteria in their workplace isn’t the Petri dish, it’s the coffee mug. Their computer programs solve complicated equations, for example, to predict properties of molecules to be validated later in labs, part of the long journey to discover a new life-saving drug formula. But it would also be cool to describe them as XXI century biologists.


 

So, Prof West and his team searched for foundational principles, from which the scaling laws, such as Kleiber’s Law for animal metabolic rates, the law of three-fourths, could be derived. They settled on a set of three principles. These will be important to our conclusions about the implications of scale for continuous improvement of companies.


The principles are:

·       Space filling – the life-sustaining networks inside animals’ bodies aim to reach every organ and every living cell, to carry blood, oxygen, nutrients, and chemical signals. Similarly, the physical infrastructure of a city, regardless of its flavour of urbanism, aims to reach every building with its transportation networks, such as highways, streets, walkways, bus routes, bike lanes, and its utility networks, such as water pipes, electric lines, Internet cables, and so on. Companies meeting their customers in physical spaces have networks covering the geographical areas where their customers live, such as gas and charging stations, bank branches, and retail stores. Even high-technology companies operating almost entirely in invisible virtual spaces lay out their communication networks to reach every employee and every point of access for their customers.

·       Invariance of terminal units – in plain language, things at the edge of the network, should be of comparable size to each other and independent of the animal’s size. Elephants have thousands of times more cells in their bodies than cats, but the cells themselves are about the same. Electrical outlets and bus stops aren’t hundreds of times bigger in New York City than in small towns. Cafes in the Starbucks’ network occupy roughly the same real estate footprints as independent chains’ coffee shops.

·       Optimization – the network seeks to carry out its mission with the least amount of energy. For animals, this means minimizing the energy expended by their beating hearts to pump blood and deliver the vital oxygen and nutrients with it to every remote cell in their body. For cities and companies, it’s sustaining the movement of people, goods, and services with the least expenditure of constrained resources. The optimization can be evolutionary. Of multiple possible designs the more optimal ones survive. For animals, this means natural selection based on the Darwinian test for survival in the wild. For companies and cities, the societies and competitive markets provide the fitness tests, with the possible addition of artificial selection based on government policies.


Prof West and his colleagues showed it’s possible to apply these principles to mammals and derive the scaling law of three fourths, from these three principles. (The book cites four scientific papers published in physics and biology journals between 1997 and 2005.) The theoretical result is n/(n+1) for an n-dimensional space, so for n=3 the scaling exponent works out to 0.75.


The Newtonian feat of such derivation is thus within reach of the mathematical apparatus of modern theoretical physicists. But we aren’t talking about the proverbial spherical horses in vacuum. These are real three-dimensional animals, our beloved polar bears, giraffes, koalas, and sea otters, who perform incredible feats of fitness in the wild daily, and whom we admire for their striking appearance and graceful movements.


Cities and Companies Are Different Animals (pun intended)

With the understanding of principles, scaling laws, and examples how they apply to animal life-sustaining networks, it is easier to move on and consider cities and companies.


 

West presents data on scaling of cities in his book, considering first various types of physical infrastructure networks as functions of the city population: filling stations, roads, and various pipes and cables. It turns out, all such networks scale very similar to the animal metabolic networks, except the scaling exponent is 0.85 rather than 0.75. What’s even more interesting, the 0.85 exponent remains constant across cities in various countries and cultures.


Both metabolism in animals and physical infrastructure in cities are examples of sublinear scaling. (The exponent, 0.75 or 0.85, is less than 1; linear scaling implies exactly 1.) Both give economy of scale. Larger mammals are more efficient pound for pound. Larger cities have, on per capita basis, less metal and concrete to keep the flow in their transportation and utility networks.


Besides the infrastructure, cities also have characteristics that scale very differently, super-linearly, with the exponent 1.15. Wealth, wages, production of goods, services, and ideas is greater in bigger cities, not only in absolute terms but on per capita basis as well. Cities are also social networks, where interactions of people produce a nonlinear network effect: doubling the city size more than doubles what its average citizen produces and consumes. Such opportunities, not only economical but also cultural, continue to attract people to cities, despite the bad stuff, such as crime and disease, also benefitting from the network effect. This scaling pattern is the opposite of economy of scale. It’s about the out of proportion, sky-is-the-limit outputs and outcomes given the fixed inputs, such as the sunk cost of the city’s existing infrastructure and the number of people who are already here.


I recall my impressions of visiting one of the largest cities on Earth, Sao Paulo. From almost anywhere within the city, all you can see is maybe the skyline of the next city district. It may seem to foreigners that Sao Paulo is located on Brazil’s Atlantic coast, but it is actually away from it, on a broad plateau at about 700 metres of elevation, and it takes about 2-3 hours’ drive and long descents along mountain roads to reach the coast. There aren’t many opportunities to meditate in front of a dramatic view like the surroundings of San Francisco’s Golden Gate bridge. The locals seem to always turn their eyes to the faces of fellow citizens, to engage in some social networking, which can be about any thinkable human endeavour: education, art, fashion, medicine, food, technology, and so on. And the exponent 1.15 can go a very long way when applied to Sao Paulo’s scale of population.


On scaling of companies, West presents data, various company metrics as a function of the number of employees, showing how companies fit somewhere between animals and cities. The dominant pattern is linear scaling, with the exponent of 1, such as the scaling of sales. But the picture is not as clear-cut as in animals and cities. For example, expenses scale sub-linearly up to a point and then flatten out to linear scaling. There’s also sub-linear or quasi-linear scaling of various other metrics, with exponents between 0.79 and 0.96 cited in the book. Interestingly, the data set on Chinese companies shows sub-linear scaling of sales (exponent 0.83), potentially revealing important structural differences between Chinese and Western economies – ideas for a future study. One thing is clear: there is no widespread, universal super-linear scaling in companies like what we saw in cities.


The quasi-linear scaling of companies leads to some sobering realizations for continuous improvement consultants. Companies – their clients – have already figured out how to grow to their current actual scale and how to operate at it. You have to meet the client where they are at. If your understanding of scale tells you that you need to scale your improvement method, methodology, process framework, toolkit, or playbook, then you are already behind the very client you’re trying to advise. This defeats the purpose and value of consulting. The net flow of knowledge should be towards the client.


An intelligent discussion of scaling laws reveals another popular fallacy: introducing animal-like sublinear scaling – and settling for the slower pace of life – into business environments where the superior linear or quasi-linear scaling is already prevalent. Companies can do better than merely substitute for agile squirrels working in one- to two-week sprints with elephants, “agile” at their “scale” and lumbering in three-month timeboxes.


Sub and Super, Arteries and Capillaries

One last piece of the puzzle I need to extract from West’s Scale today is the difference maker between the sub- and super-linear scaling. What is it that makes cities scale differently than animals?


The short answer: look where the flow is.


In the sub-linearly scaling network that pumps blood throughout our body, the greatest flow is through its central part, the aorta, next to the heart. The aorta branches into several major arteries, each carrying only part of that flow. Larger arteries branch into smaller arteries, into progressively narrower blood vessels, and so on, until the blood reaches capillaries where the flow is the least.


Infrastructural networks of a city work in a similar way. City planners don’t call major roads “arteries” for no reason. The Union Station in Toronto’s downtown is the aorta of the city’s public transit network. Many parallel train tracks extend from the station in the west and east directions. They branch into lines reaching into every sector of the city and its suburban area. Subway, streetcar, and bus lines also connect at this station. Toronto’s motorists would of course reserve the term “aorta” for Highway 401. Going west-east across Toronto’s uptown areas about 12 kilometres north of downtown, 20 lanes at the widest point, it is North America’s highest-throughput road, at about 500,000 vehicles per day. It, too, branches into “smaller” arterial highways, parkways, boulevards, and, ultimately, calm neighborhood streets, the “capillaries” of this transportation network.


Social networks are the opposite. The greatest flow is near the network’s edge, its terminal units. In Internet social networks, which we have all become familiar with in the last 15 years, the terminal unit is an account associated with a person. For most of us, the greatest communication volume comes from posts, likes, comments, notifications from the nearest circle of friends, connections, and followers.


In our face-to-face social networks, the greatest interaction amount is with the immediate family. Somewhat less intense are interactions with slightly wider circles of nearest relatives and closest friends. As we consider gradually widening circles of relatives, colleagues and acquaintances, the interaction intensity weakens accordingly. British scientist Robin Dunbar discovered this relationship between intensity of interaction and closeness to a social network’s edge.


Prof Dunbar defies easy categorization by field of study: biologist, anthropologist, psychologist?  West describes him in Scale as an “evolutionary psychologist.” Dunbar’s Wikipedia page says, “biological anthropologist.” Popular culture collapses Dunbar’s research into a single number, 150, also known as Dunbar’s number, with a recommendation: this is how many friends you should have, because your brain cannot handle any more. A much deeper and yet very accessible insight is Dunbar’s hierarchy of social circles each of us has and the unmistakable inverse relationship between the circle’s size and the interaction intensity. Dunbar further shows how we perceive a qualitative change in our interactions between any two circles that differ by a factor of three to five. (That’s where the number 150 represents, under certain conditions, an important qualitative threshold.)


A modern professional with 500+ LinkedIn connections (which is almost everyone nowadays) can take a hard-and-fast rule of 150 with a grain of salt. But the more nuanced insight into the expanding social circles and diminishing signals makes perfect sense. I maintained my LinkedIn network over the years to include people I’ve had at least some meaningful exchange on a professional topic, yet I sometimes struggle to recall the circumstances in which one of its almost 2,000 members and I met. An interesting phenomenon of open networking emerged in the early days of LinkedIn, despite the network’s initially restrictive approach to making connections. Open networkers or LIONs declared the policy of accepting all invitations and accumulated tens of thousands of connections. They accepted the tradeoff of weaker links to most of their network as optimal for their career or business. Going in the opposite direction, if LinkedIn forced me to reduce my number of connections to, say, 500, the strength of the remaining connections and the amount of exchange would no doubt increase, turning gradually into an information firehose with the top 150, 50, and 15.


The modern city is thus a combination of two very different types of networks. One is the sub-linearly scaling infrastructural network, with physical terminal nodes, such as power outlets and bus stops, with greater flow though the arterial lines and less at the edge, providing economies of scale. The other is a super-linearly scaling human network, where the terminal node is the social animal, where the flow is greater at the edge, and which creates the excess outputs, wealth, innovation, and cultural outcomes. Prof West concludes in Scale that the two networks are mutually dependent. The infrastructural network sustains life and facilitates human exchange and movement with economies of scale, while the excess economic productivity of the human network pays for the infrastructure.


Shouldn’t We Just…?

A reader can finish this sentence easily. Why don’t we design our creative, intellectual companies of the future as cities and social networks?


We already have the sub-linearly scaling corporate infrastructure. The company’s budget is its “aorta”, which splits into arteries – business units, departments and so on, until we eventually reach the terminal node: the employee position. The position has a budgeted salary, a fraction of what has flown down the branching arteries. It has a job description. The person hired for this position is expected to perform the described function, which is probably at best a subset of their human potential.


This may just be the optimal design for the industrial era. It provides economies of scale. Market forces and government regulation, where appropriate, provide the fitness test that selects companies that implement this design more optimally. If less fit as in underbuilt, the network fails to provide economies of scale and loses to competitors. If overbuilt, the network gets too expensive and unprofitable. Management consulting rooted in the industrial paradigm can help companies rooted in the same paradigm be fitter, by helping them build the optimal networks, whether they are bank branches, retail outlets or transportation solutions. Traditional continuous improvement can help further optimize inputs, outputs, and processes in between.


It isn’t news to many that while we live in the new, post-industrial, information era, much of how we set up, run, and improve our companies still carries many industrial-era habits. In the two-sided picture of sub- and super-linear, industrial and creative, infrastructural and social, the continuous improvement thought has still done more on the side of the former, and less for the latter.


So, for the information era companies, why don’t we take our sub-linearly scaling corporate infrastructure, offices, desks, computers, coffee machines, watercoolers, Internet connectivity, Zoom, cloud storage – and augment it, like a city, with the super-linearly scaling human social network? Shouldn’t we just do that?


Yes and.


What is the minimum we need to do to prevent this bright idea from becoming another stupid slogan painted at an elevator landing in your office building? What is that we must do absolutely to ensure it becomes more than an impractical platitude?


Back to the principles.


Terminal units. If we are to design our company like a city, a super-linearly scaling network, we need to figure this out. What would be the terminal units of this network? What is the design of the unit? How does it connect to and interact with other units to form a network?


This is by no means an exhaustive list of questions, but I’m asking for the bare minimum. We are indeed talking social networks, but in a more serious sense than sharing and liking on Facebook or upvoting on Stack Exchange. We would expect an electrical engineer designing an electrical network to know their power outlets, distribution panels and substations. A logistics expert would know their ports, distribution centres, warehouses, stores, ships, and trucks. What have we got as building blocks of our network?


The good news is that there are several workable approaches. Proven in practice and continuing to develop, they are certainly promising. I will now describe two of them briefly.


Team As a Verb

A Toronto-based agile expert and consultant Jeff Anderson has published an important book recently: Organizing Toward Agility: Design, Grow, and Sustain Self-Organizing Structure at Scale (2023). The work of the firm Agile by Design Jeff founded about a decade ago and still runs provided much of the practical material in the book. This work fits the popular Agile theme of the last 15 years but has deeper ideas on the inside.


Jeff Anderson points out the modern knowledge work was never meant to be done by isolated individuals. It was always meant to be done collaboratively. The author explores collaboration patterns: pairing, of two similar or different workers, swarming, mobbing, again of people of the same specialty or a mix of specialties, connecting to other teams, and so on.


The author effectively positions individuals and teams as the terminal units of the network. He shows it is wrong – industrial-era thinking – to view them as mere structural units of the company. He encourages readers to think of “team” not as a noun, but as a verb. Teaming, the ability to quickly form various necessary work relationships with other individuals and teams, is an essential skill, at least equally important to specialized, “technical” skills. It is therefore an essential feature of a modern company to sustain and develop such skills in its people.


In contrast to “structural” Agile frameworks and the dogma of “stable three-pizza teams”, Jeff acknowledges the variety of specialist skills and the resulting explosive diversity of dynamic work relationships forming in large modern companies. The initial menu of teaming is only a teaser – the author finds more patterns as he explores Dunbar circles, their identities and interaction intensities gradually weakening with size. His emphasis on the network edge and the thorough design of terminal units are both correct if we are to trust the scientific framework of scale presented by Prof West.


Services Everywhere

The Kanban method, originated by David J Anderson (no relation) in 2004-07, effectively positions services as terminal units. A modern post-industrial enterprise is a network of services, says the service-orientation agenda, one of the central parts of the method.


My goal here is of course not to explain Kanban, but to recast its common written and tacit knowledge, what its users know and do in practice, in terms of the design of the social network, particularly focusing on the crucial part, the terminal units and connections between them.

Kanban’s design of a service as a terminal unit has several important attributes.


First, the service has customers. Here on the left is a customer with some need or request, over there on the right is the customer receiving the result. Some process, perhaps worthy of a continuous improvement effort, is in the middle of this picture. Kanban is non-prescriptive about services’ interfaces to their customers, instead acknowledging the variety of interfaces existing in real-world businesses. It obviously accommodates clear-cut interfaces, where customers explicitly place their service requests, what most people associate with the word “service.” But it is also, less obviously, inclusive of very fuzzy interfaces, where customers cannot recognize let alone articulate their needs, but businesses nevertheless have abilities to sense those and to turn them into work that ultimately becomes the continuous stream of exciting products, their variants, updates, and so on. Kanban’s concept of service thus includes the work of innovative product companies. If the economics of customer relationships involves large intake or output batches, Kanban systems reflect this choice with the appropriately sized buffers. Services thus include project work, too.


Second, the service has a manager. Because a service has customers, who may become happy or unhappy, and exists in the context of a business that may succeed or fail, there’s accountability for the differences within this range of outcomes. Accountability for delivering the service, for dealing with customers, for running, understanding, and improving the service’s process is the essence of the Service Delivery Manager (SDM) in Kanban. Specific implementations of the SDM vary widely in practice, from very formal to informal, individual or by committee, all dependent on the organization’s culture and the local context. A common element in all this variety is the realization that services don’t deliver themselves, it is not enough for people to just perform their functions, not enough to work on the content, someone’s mind must be on the service’s process, and this someone must be ready to take acts of leadership.


Third, the service has a notion of capacity, the ability to serve incoming demand concurrently. The amount of concurrent work cannot be arbitrary or out of control, while the service magically maintains consistent time-in-process, quality, and other criteria customers care about. Kanban dispenses with such wishful thinking. Instead, it makes key features of a service to sense its present concurrent work (using, for example, the practice of visualization) and to balance capability and demand using feedback loops, explicit policies, and WIP controls.


Fourth, Kanban employs a number of quantitative measures to describe the performance of a service as a terminal unit. It is possible to quantify the service’s demand and throughput, time-in-process, inventory (work-in-process, WIP), other relevant indicators, see as trends and patterns in them, validate improvements, and so on.


Kanban gets even more interesting when it comes to the terminal unit-service design features that enable it not to merely work in isolation, but to interact with other services-terminal units and improve as a network.


First, demand analysis. As SDMs of numerous services have discovered, a real-world service gets its demand from multiple directions. At a high level, there’s value and failure demand and internal improvement. Within the preferably dominant value demand, there are deliverables differing by the nature of their content, multiple customer segments, having different time and quality expectations, volume, timing patterns, and so on. A realistic demand picture is heterogeneous, even for one service. Demand analysis as part of the Kanban system design is about capturing this variety and presenting it in an understandable form. Some of the demand comes from internal customers – other services within the network. For shared services, all of their demand comes from other services. Demand analysis thus reveals and establishes important connection points between the services-terminal nodes.


Second, workflow mapping. In a shallow interpretation, this may be a mere sequence of steps to perform to get the result delivered to the customer. Kanban encourages tracing the process of knowledge discovery, often resulting in simpler maps revealing the locations in the process where the service interacts, depends, and acts as a customer to other services. Effective Kanban workflow visualizations thus reveal the outlines and the chokepoints of the company’s social network that goes to work to discover the necessary knowledge.


Demand analysis and workflow mapping are two of the steps of STATIK, the so-called systems thinking approach to introducing Kanban, one of the cornerstones of the method. They are among the essential skills in Kanban system design.


Third, there is the system of Kanban cadences, enabling the services-terminal units to close organizational feedback loops, to propagate the necessary information, to synchronize when necessary, and to pursue improvements as a network.


Kanban’s service orientation agenda, the essential skills of Kanban system design with STATIK, the system of cadences, and the key improvement strategies turning into actions in between the cadences have all been part of the Kanban Management Professional (KMP) training since the establishment of the KMP credential in 2014-15. The KMP has thus been the proven training component of the knowledge transfer necessary to develop SDMs to run the terminal nodes of the organizational network in ways that enable super-linear scaling.


Without going further and over the entire method, it is clear that Kanban offers serious practical answers to the challenges of super-linear scaling in the XXI century post-industrial, creative, intellectual enterprise. Kanban users can accomplish this by seeing their company as a network of services, focusing on the edge of the network, and applying Kanban guidance on service design as a terminal unit plugged into the network of similar units. This is of course a necessary minimum rather than a sufficient condition. It goes almost without saying that a successful solution will take leadership, at the service-unit level to sustain and improve, and across the network to shift the culture away from industrial-era habits and towards the new paradigm, emphasizing flow at the edge. But cultural shifts happen through many practical actions taken frequently rather than dramatic slogans. Finding a service and designing a Kanban system around it is a step in the right direction. So is finding several interconnected services and seeing them as a network.


Five Dimensions Revisited

My earlier article on Kanban and Scale presented a categorization model of five dimensions. We can now take another look at this model from the scientific angles established in this text.


In the Width dimension, some companies model their processes as a sequence of activities with handoffs between functional units in between. This isn’t incorrect. They have a right to see it that way. But in more mature Kanban applications, Kanban method users model their workflow as a knowledge discovery process, Kanban board columns represent collaborative knowledge discovery activities, and the boundaries between them show changes in the activity and the collaboration pattern. With each new activity, a different “subnet” of the company’s social network may go to work to discover the knowledge. I showed show this approach works in my short speech at the Lean Kanban North America 2014 conference in San Francisco (about 10 years ago almost to the day).


In the Height dimension, companies set up hierarchies of boards and work items. Do the links from children to parents trace the company’s org chart? If so, this isn’t incorrect, but it’s the same old-era industrial thinking. In more mature implementations, we again see the knowledge-centric approach. The gaps between parent and child work item types instead represent the information gained through decomposition.


The structural approaches to the Width and Height dimensions aren’t incorrect. But their inherent limitations should now be obvious to a reader understanding the scientific framework of scale. The knowledge-centric approaches may be our ticket to the super-linear world.


The Scale-free assumption is the linear scaling law in the more scientific terminology.


The Depth and Knowledge dimensions are where we aim for super-linear scaling. On Depth, I have just discussed Kanban’s approach to networks of services, focus on the edge, and the design of a service as the network’s terminal unit. The Knowledge dimension has a similar social-network approach, with managers (SDMs) and improvement coaches serving as network nodes.


It becomes clear now that the discoveries of STATIK (the systems thinking approach to introducing Kanban) and the Kanban Maturity Model (KMM) as Kanban’s “hidden gems” bridging gaps in the Knowledge dimension weren’t accidental. Both serve as languages facilitating exchange between the two types of nodes in the knowledge networks. Proficiency in each, signified by Kanban’s KMP and KCP credentials respectively, is a key feature of the network node design.


Finally, the science of scale informs the design of the network of Kanban experts who sustained and advanced the state-of-the-art of the method over the years. As this topic is of interest to a very narrow audience, I will save it for a separate memo.


Conclusions/Summary

Continuous improvement professionals use the term “scale” often as they deal with companies and pursue consulting engagements of various sizes.


Continuous improvement community remains highly focused on practice, yet it should also be vigilant and check its guidance against established scientific knowledge from time to time.


Professor Geoffrey West’s 2017 book Scale: The Universal Laws of Life, Growth, and Death in Organisms, Cities, and Companies gives such scientific knowledge in an accessible non-fiction form. As we continue our dialog about scale in the continuous improvement communities, this book is impossible to ignore.


Scale provides a somewhat sobering collective look in the mirror. It does validate some practical approaches and helps put them on a more solid foundation. At the same time, it exposes the naivety of the ongoing dialog about scale and of various methodologies and process frameworks.


Exploring ideas presented in Scale (the book) can help elevate the level of discussion and possibly the quality of consultants’ improvement advice to their client companies.


It is important to grasp the basic math of scaling laws, linear, sub-, and super-linear scaling. These patterns are well illustrated in relatable examples, such as animals, cities, and companies.


Scaling laws derive from the first principles: optimization, space filling, and similarity of networks’ terminal units.


Animal metabolism and city infrastructure networks are examples of sub-linearly scaling networks. This scaling pattern is appropriate for industrial-era companies.


The social networks of cities give an example of super-linear scaling. This is where we want our future creative, intellectual companies to be. But we aren’t there yet as industrial-era habits still dominate much of how modern companies are run and improved.


The key difference between sub-and super-linearly scaling networks is where the greatest flow is. With sub-linear scaling, it’s the networks’ centres, such as animals’ aortas and cities’ arterial roads. With super-linear scaling, the greatest flow is at the network’s edge, the terminal units and their immediate connections.


A credible approach to scale in modern knowledge-work enterprises has to focus on the network’s edge and address seriously the design of the network nodes and communications in their vicinity.


Practical continuous improvement approaches meeting this test do exist. They are each proven by more than a decade of practice. Their success is not accidental. It owes not only to the enthusiasm and hard work of the adherents, but also to the solid scientific foundation.

 

181 views0 comments

Comments


bottom of page