Turning a company into a system

This is the introduction to a book that pursues to turn productive companies, companies which produce what they sell, into real systems. Even though it is a ineluctable challenge turning any kind of human Organizations into systems, this work focus for the time being on the entrepreneurial field, more specifically, on productive companies; the commercialization and logistics fields are not in the scope of this work. Productive companies include industry and services, manufacturing and multi-projects.

The term “systemic”, rather rare in the short past, has become more common, almost a fad. However, a fad about a deep subject only touches the surface, often distorted by the former paradigm.

A system is an organized global unity of interrelations and interactions between elements, actions or individuals. The interrelations make up the structure of productive systems. For example, the “routes” in mass production and the “networks” in multi-projects and service environments. The interactions in the time variable between the productive units make up the dynamics of the productive systems. Interrelations are static, interactions are dynamic.

As a system is a global unity of interrelations and dynamic interactions, interrelations and interactions are the essence of the system, not the interrelated and interactive “things”. Things are subjected to uncertainty, interrelations and interactions are not as they make up the system and a basic property of a system is its permanence. «The permanence of a system is not a consequence of the inertia, the bulkiness, the “strength of the things”» (E.Morin, “The nature of Nature”), but the result of an organization with a tautological finality, the permanence of the system.    

Turning a company into a system demands the organization of tis interrelations and interactions. Interrelations are organized as they are defined with the construction of the structure of the system; this is done in any productive environment with a proper level of organization. However, a proper organization of the interactions, that is, of the production dynamics, does not exist in general. The attempt of organization of this dynamics that can be found is scheduling, even supported by sophisticated software. Briefly, scheduling consists of adding tine milestones to productive structures, to routes and networks.

The organization of a system is not just a matter of putting an order in place –this is what the structure provides with- but mainly of defining the interactions that make up its dynamics. As interactions are part of the essence of the system they must not be subjected to uncertainty. It is obvious and well known that no scheduling is immune to uncertainty so it cannot be the organization of any production dynamics. Organizing is not predicting what will be made.

Although the order that the structure provides with is a component of the organization of a system, that order is not applicable to its dynamics, as scheduling assumes. Organization cannot exist without an order, but neither without disorder. Any dynamics is subjected to uncertainty, which can only be managed with a confined disorder.

Regulation is the alternative to scheduling in order to organize the dynamics of a productive system: to define what is made here and now depending on a desired reality and the current reality. Organizing the dynamics of a productive system consists of defining its interactions as regulation loops. Regulation loops are interactions between desired realities, current realities and human behaviours in the way of actions or decisions. Instead of predicting when tasks will be made, it is defined the next one when a task is completed.

Regulation loops confine uncertainty within margins between desired realities and current reality named “buffers”. Buffers are not a protection from uncertainty but a confinement of its effects achieved by management.

Contrarily to scheduling, the dynamics shaped by regulation loops consists of interactions which are not subjected to the uncertainty they confine: if that dynamics is not implemented, another one subjected to uncertainty will be in place.

People behaviours are the active element of regulation, it is people who their way of doing regulate the internal dynamics of the system. It is self-regulation. Local behaviours depend not only on local realities but of other local realities structurally interrelated; and they affect those realities as well

Self-regulation means that behaviours do not depend on anything but the people who produce. Their work needs the support of software providing them with the information for their behaviour, which is not determined by the software. The confined disorder needed for the management of uncertainty is a margin of freedom –freedom is disorder-; margin of freedom confined by the margins between desired and current realities that have been called buffers.

At the system level, the desired reality in a productive system is the improvement of its performance. A necessary condition to turn a productive environment into a system is the existence of one and only one indicator of its performance. As a system is a global unity, there must be one quantified global performance. “Key performance indicators” neither key nor indicators of performance, they just indicate that Operations environments have to be turned into systems as they are not yet.

In order to come up with the indicator of the performance the state variables at the system level have to be established: the variables that quantify the state of the system at a time, output, input and response time. The following differentiating concepts have to be remarked:

·         The same magnitude must be used for measuring output and input so that quantification of the only valid efficiency, the one at the system level, the global efficiency, is possible. This magnitude is the work magnitude, man-hours, which fuses two elements intimately linked in productive systems, resources and time. By the way, work in the genuine generator of value in a productive system according to people like A. Smith, D. Ricardo and others who are considered funders of Economic science.

·         As output is a state variable, it cannot be completed output but output in process (OIP) at a time. OIP is the only output whose real cost can be quantified: the input, the man-hours that are available at a time for producing the OIP. By the way, companies that share an OIP can be turned into systems; companies that operate as task forces dedicated to just one work order cannot; for example, while in house aid services not always can work as systems, hospitals can and should be turned into systems.

·         Quantification of output and input requires a time state variable, the response time of the system. A rather ignored variable, response time are generally used at the level of work orders, in production or in projects, not at the system level. The response time of the system is the time that every productive unit has available for producing its share in the OIP. The input of the system is the D man-hours available during the response time of the system rt: D=I.rt, being I the number of persons available for producing the OIP.

These three variables cannot be directly merged into one, although output and input can be fused into the global efficiency as the rate between the man-hours DP in the OIP and the input D: E=DP/D. DP is the quantification of the OIP in man-hours, the “nominal” man-hours in the routes and networks tasks of the productive structure. Defining one indicator of the performance seems to demand a mathematical blend of the efficiency E and the response time rt, same way as output and input have been combined to come up with the efficiency E. However, the efficiency E is an “invented” variable as there is not an a priori relation, independent on us, between output and input. On the contrary, a connection between efficiency and response time does exist we want it or not, there is a trade-off between them: for the same performance, the higher the efficiency the longer the response time, and the other way around. It is so independent on our will that it actually is a natural law.

A mathematical model for the management of a productive system demands the realization of the trade-off between efficiency and response time, as well as its quantification. This quantification consists of an equation connecting efficiency E and response time rt to an indicator of the performance: rt=f(E, B₀), being B₀ the indicator of the performance. This indicator is in fact the parameter that ensembles the unlimited combinations of efficiency and response time that are possible with the same performance. Any point on the curve representing that equation is one way of having the same performance: infinite ways of selling the performance of Operations.[1]

---

[1] The point E=0 does not belong to the curve, it is a singularity.

The existence of the trade-off between efficiency and response time implies that the cost of producing an order depends on the production lead time and hence on delivery times. A derivative of it is that cost is not an attribute of a product but of an order or a project. Ignorance of response time in the calculation of the cost of producing is a crucial flaw that contravenes a physical law: higher speed is not for free. The right calculation of cost as a function of lead times has such an impact on management that it is hard to understand the so long persistence of this conceptual error. Behind it can be found the assumption that the cost of a product is the sum of processes costs, ignoring the waiting times, the more relevant portion of lead times in general, and the cost of reducing them.  

The current reality at system level in the above scheme of generic regulation is its current performance, quantified by B₀. The desired reality is its improvement, global improvement which has to be pursued at local level: all actions are local. The key matter is having right bridges between locals and the global, between the productive units and the performance of the productive system.

The indicator B₀ of the performance is a magnitude related to the buffer needed to produce the OIP. This buffer B is the subtraction of the input D available man-hours minus the DP “nominal” man-hours in the OIP: B=D-DP. The buffer of man-hours B is of course at the system level.

The objective of Operations is to reduce the cost of production of the OIP demanded at a time by the market. One key issue in management is the right quantification of the production cost. Cost accounting calculations are wrong because of its ignorance of response time. However, calculation of the cost of an order or a project is nevertheless needed, calculation that must take into account its lead time. Besides, cost accounting calculations do not have the solid basis of a production cost at global level which is not conventional but true. It is unquestionable that D man-hours is the cost of the OIP at a time. This is a solid basis at global level for the calculations at local level of the cost of orders or projects, first in man-hours and then in monetary terms. Going from a global cost to a local cost requires conventions; right conventions that do not induce wrong behaviours, like the calculation of product cost from the cost at tasks level induce behaviours driven by local efficiencies. The cost of an order or project is certainly conventional, but it is based on a non-conventional cost of the OIP, the only cost whose non-conventional calculation in man-hours is possible. I should hammer in it: conventions are needed to bridge a non-conventional global cost to local costs at the level of orders or projects.

It is very important that there are no intentional buffers in the DP nominal man-hours of the OIP. Buffers at tasks level are not desirable at all, not even buffers at the level of orders or projects. No buffers should be consented in the routes of orders or the networks of projects. The right buffers are at global level and at productive unit level, and both have to be connected to the OIP. 

The cost of production in man-hours can and should be split into two: the structural cost, which is the nominal man-hours of the OIP, and the man-hours in the buffer B. The buffer B is the margin between the desired reality D=DP and the current reality D=DP+B. Performance should be improved at constant structure, by reducing the buffer B needed for the production of the OIP. Performance is not improved by acting on the processes; process improvement is an arena, the improvement of the structure of the productive system, different from the improvement of the performance. In the recursive loop structure improvementDdynamics improvement, dynamics improvement is improvement of the performance, while the improvement of the structure opens new windows for the improvement of performance by removing “constraints” blocking it.

In order to really catch the rationale behind the improvement of the performance by freeing up man-hours in the buffer B, it is necessary to understand how uncertainty acts.

Uncertainty acts in two ways:

·         At local level, on specific processes. It is the source of uncertainty: local fluctuations and disturbances. It is a structural (statistical) variable not connected to the performance.

·         At global level. The processes are interrelated in the productive structure. These interrelations happen to be channels of propagation of the uncertainty. Propagation of uncertainty has an impact on the system significantly higher, and even much higher in most scenarios, than the uncertainty itself on the processes. Performance improvement must focus on the propagation of uncertainty, not on the uncertainty itself that processes improvement pursue to reduce.

The reduction of the global buffer that performance improvement consists of is achieved by a dynamics of regulation that blocks the propagation of uncertainty. The buffer B is not protection from uncertainty; it is the margin wherein the impact of uncertainty is pursued to be confined a priori by the blockage of its propagation throughout the system that the regulation of the system´s dynamics involves.   

The reduction of the buffer B requires “freeing up” man-hours in it as they have not been used. If fB is the freed man-hours as they have not been needed, a rate fB/B of improvement of the performance is achieved.

Man-hours in the buffer B have to be freed up by the productive units: the “Skill Groups” (SGs). A SG is a group of people that share a skill on their work on the OIP at a time. The structure based on SGs is a very flexible alternative to the classic departmental structures: by definition, people do not belong to SGs.

Every SG has its share in the OIP, sDP nominal man-hours to produce; the SG has sD man-hours available during the response time for producing sDP, and therefore a buffer available sB=sD-sDP. A SG frees up sfB man-hours in its buffer sB if those man-hours have not been actually used. Herein the rate sfB/sB=sM is called the “margin” of a SG at a time.

The local-global bridge, between the SGs and the global performance, is based on the “law of the weakest node”: in an interactive productive system, wherein parts interact, one part determines at a time the global objective of the system. This part is the “weakest node” at that time. In the net of interactions the weakest node at a time is the SG with the lowest margin. This SG is the “limiting SG” (SGL), as its margin sM_L=(sfB/sB)_L determines the performance improvement at a time: fB/B=sM_L.

The difference between sM_L and the margin of any other SG is useless, it is just wasted man-hours. The more balanced the SGs margins the better the system´s performance. So, SGs should use their margins to raise the margin sM_L of the SGL, included the SGL itself. Interactions between the SGs, regulated by the SGs regulation loop, pursue balancing upwards the SGs margins. SGs can contribute to it in two ways: prioritizing the tasks that push downstream flow of work towards the SGL is one way; the other one is using their versatility to transfer man-hours to the SGL.

At SGs level, the desired reality in the regulation loop is balanced higher margins. The pursuit of balanced margins is the way of blocking propagation of uncertainty, the way to confine the effects of that propagation formerly mentioned.

As the performance depends on the level of balance of the SGs margins, the better the performance the more instable the weakest node. Pursuing the stability of the weakest node implies giving up further performance improvement. 

A quite generic statement: in a system with interactive parts, like Operations, the “global optimum” does not depend on a sum or conjunction of local optima: this is “analytical” thinking, linear relations between the parts and the whole, prevalence of the parts on the whole. But the global optimum is neither achieved through the “exploitation” of a specific node and the subordination of the others to it: this is “holistic” thinking, the prevalence of the whole over the parts.  

Both analytical and holistic thinking are simplifications of reality. «Simplification is barbaric thinking», claims E. Morin. Moreover, according to Morin, «the paradigm of the holistic simplification leads to a neo-totalitarian functionalism which fits the modern forms of totalitarianism». Holism is based on a contradiction in terms, the localization of the global at a specific node which the system is supposed to subordinate to. That subordination to a localized supposed global optimum turns it into a dominant node: this is the basis of totalitarianism.

The right paradigm is neither “analytical” nor “holistic”, is the complexity paradigm. In a system, the parts do not make up the whole neither the whole determines the parts; the parts and the whole are interactive complementary antagonisms in a complex relation essential in any real system. That complex relation can be named “cohesion”, “harmony” or whatever metaphor is better liked. In a productive system, cohesion takes the form of upwards balance of the margins of the parts –the SGs-. This statement can be widely generalized.   

A very relevant implication of behaviours driven by the pursuit of balance of margins by the SGs is that there is no right local measurement for the SGs. In a more generic way, there is no right local measurement for the parts in an interactive system. Right behaviours of the parts cannot be driven by any quantitative variable but by the pursuit of the complex relation between the parts and the whole in a system –cohesion, harmony…-, a relation which cannot be encapsulated in one figure. That complex relation in productive systems is eased by the feasibility of the quantification of the SGs regulation loop. Other systems –like social non-productive systems- demands another kind of regulations based on the same principle of balancing other kinds of individual margins. 

The reader who has reached up to here might wonder about the little room dedicated to the issue more usually cared in Operations: the management of orders or projects individually considered. There certainly is a contrast between the focus of classic management on orders and projects as compared to the focus on resources in an Operations environment that has been turned into a system. For example, if a multi-projects environment has to be turned into a system, the system cannot be the set of projects in process actually managed. Current multi-projects environments are not systems but sets of projects.

There are orders/projects and resources in a productive system. What is actually managed, orders/projects or resources? The usual answer to that question is “both”, which is not fully wrong. However, if the focus is on orders or projects, which is the common one, then resources are allocated to orders/projects depending on their “progresses”. As a matter of fact, this focus causes conflicts between the people responsible for orders/projects demanding resources for “their” orders/projects. On the contrary, if the focus is on the resources, tasks of orders/projects are assigned to SGs. In a productive system resources are not assigned to OTs but OTs assigned to resources. This is what regulation loops are based on. It is a real cultural change. For example, in multi-projects systems people do not feel ownership on “their” projects, so that they do not identify with the reality of projects rather than with the reality of their system. People responsible for projects are not project managers but responsible for the relations between Operations and the external entities in relation with the OTs (vendors and customers).

The SGs regulation loop is an internal self-regulation. Additionally, it is needed to care of the interrelations between Operations and the external environment, specially the market, through Sales or directly. The OIP is the result of releasing new orders/projects as well as the completion of others that have been part of the OIP. Release of new orders/projects is not an interaction between Operation and the market through Sales or just directly, but an interrelation which imposes some conditions to both sides. The Sales commitments regarding delivery times, for example, must respect the condition of being according to the current performance of Operations. Operations must produce nothing else that what has been demanded by the market, in the conditions committed with the clients.

The chronic problem of the reliability of committed delivery times remains because of persisting in the ignorance of a response time at the system level. As this time rt is a variable of the system, a reliable delivery time of a new order or project has to be calculated by using a simple mathematical relation between rt and the production lead times of the orders/projects in the OIP: DP/rt=SDP_i/lt_i, being DP_i and lt_i the nominal man-hours and lead time of order/project i. Both sides of this equation express the flow of work through the productive system.

Sales –or perhaps Operations itself- has to understand that besides products or services they are supposed to sell performance. Selling performance is selling with every order/project one point (E, rt) in the diagram of the trade-off between efficiency and response time. It is a condition for the release of a new order/project that the position of the system after the release is not below the performance curve currently offered to the market; otherwise, we would be demanding to the system a performance that has not been reached yet.

The new order/project must not have to be processed by a SG with zero or negative margin sM, that is, a SG in alarm. This is a second condition for a release. A third one has to do with the availability of external inputs: materials, information, subcontracting, etc.

Releasing new orders/projects is one of the roles of the SG in Operations that is herein called “External Relations” (ER). This SG is not in the productive system as it does not work on the OIP. People in ER are interlocutors with clients and suppliers regarding specific OTs; interlocution which starts with releasing of new OTs and goes on with the follow up of the OTs in the OIP.

Right understanding of the ER role requires going a bit deeper about the relations between systems. Different systems can be interrelated but should not interact. The interactions within a system make up the system´s self-regulated dynamics. Different systems have different dynamics –though based on the same principle of self-regulation-, which cannot be merged into one dynamics: if they could they would not be different systems but just one.

Possible interrelations between systems:

·         Functional interrelations between subsystems within a system. For example, Operations is a subsystem in a company, the Value Generation System (VGS) a company should be. Operations is interrelated with other functions in the VGS, but there are no interactions between them.

·         “Symbiotic” interrelations: win-win interrelations between systems, subsystems in a system or just systems by their own. For example, different systems within Operations may help each other with quantitative loans of capacity in certain occasions demanding flexibility in the market. By “quantitative loans” I mean transferences of people with skills already owned: skills are not loaned; only capacity is, as changing the skills in a system would imply changing its structure, which is together with its dynamics, the essence of the system.

The role of ER in a system is managing the interrelations with outside systems –or at least entities even though they are not real systems-. That role includes a very relevant subject: preventing from interactions with interrelated external entities. This prevention consists of respect to the restraints involved by those interrelations. The restraints formerly mentioned to the releasing of new orders/projects are an example of it; changes about orders/projects in the OIP should be subjected to conditions as well. If interactions between systems are not blocked, self-regulations in any one of them would be disturbed by external uncertainties that would interact with the internal ones, a perfect recipe to get closer to chaos. Uncertainties in different systems are heterogeneous uncertainties that must never interact. For example, the release of an order that has to be processed by a SG subjected to an alarm implies disturbing the self-regulation of a productive system with an external event.

Synopsis as a kind of systemic decalogue.   

1.  The distinction between interrelations and interactions; in a system, interrelated units perform, interactive units do not.

2.  Our interactions based on self-regulation make up the dynamics of a genuine productive system.

3.  Regulation of actions here and now instead of scheduling.

4.  One only indicator of the performance in Operations.

5.  The trade-off between the efficiency of a productive system and the response time of the system; the cost of producing depends on delivery times.

6.  The impact of the propagation of uncertainty is more important than uncertainty itself; buffers are needed for confining the propagation of uncertainty, not for protection from an uncontrolled uncertainty.

7.  In a system with interactive parts the law of the weakest node rules its dynamics.

8.  Neither analytical nor holistic, the right paradigm is the systemic paradigm: the global optimum is neither pursued through the aggregation of local optima nor through the subordination to a localization of the global optimum, a contradiction in terms; cohesion among the parts is the systemic alternative.  

9.  Management focused on resources, rather than on orders or projects.

10. The distinction between selling products and selling performance.

I invite professionals of management to read the Introduction to this book. At the end of this Introduction it is shown as a kind of synopsis a list of concepts presented in the book. I suggest that the ones who might be interested participate in a discussion forum about them.