3.2.2.1. Sensitivity to initial conditions

Sensitivity to initial conditions is a characteristic that all chaotic systems definitely have. However, sensitivity does not automatically lead to chaos. It implies that any arbitrary small change in the initial value of a chain of iteration will more or less rapidly increase as the number of iterations increases.

Minute differences in the input value during an iteration process can generate diverging evolution. These minute differences at each step matter and accumulate as the computing procedure is running. The output of an iteration run is fed back as the input of the next iteration run.

A telling example of sensitivity is shown in Table 3.1. The values of the quadratic dynamic law x + kx (1 − x) are sequentially calculated with a Casio machine according two modes, i.e. x added to kx (1 − x) for one mode and for the other mode kx (1 − x) added to x. As the addition of two numbers is commutative (a + b = b + a), we expect the same results.

The outcome of the exercise is surprising, disturbing and puzzling:

• – surprising because it clearly displays how rapidly a minute difference can be amplified. There is total agreement between the two calculation modes until the 11th iterate. Then, in the 12th iterate, the last three digits differ (734 vs. 724). The 40th iterates in the two calculation modes differ by a factor greater than 10;
• – disturbing and puzzling because what we observe here has to be assigned to the finite accuracy of the computing device. Computers have contributed to the discovery of the chaotic properties of physical phenomena modeled by recursive discrete functions. They are able to carry out thousands of thousands of iteration runs in a short time. One may wonder whether computers could be a reliable means for exploring the chaotic properties of a phenomenon sensitive to initial conditions. The question to address is: is what is observed to be assigned to the very nature of the phenomenon or the finite accuracy of the computing system used? This issue has to be dealt with using great care, and must be sorted out in an attempt to come to terms with the virus of unpredictability.

3.2.2.2. Mixing

The mixing property can be described in the following way. For any two intervals I and J of the variable under consideration, initial values in I can be found so that, when iterated, these values will be directed to points in J. Some initial points will never reach the target interval. This approach is an intuitive way to explore how a small error is amplified in the course of iteration.

3.2.2.3. Periodic and fixed points

fⁿ(X) denotes the composition of f⁰(X) (n − 1) multiplied by itself.

Periodic points are defined by the equation fⁿ(X) = X, where n is the number of iteration runs after which the system comes back to the initial value. When iteration starts from a periodic point, only few intervals are targeted iteration after iteration.

Fixed points are defined by the following equation: f⁰ (X) = X. There are points that are mapped onto themselves by the function. They are not necessarily attractors.

3.2.3.1. Inventory control

Consider a simple model to show how the chosen control parameters and the associated algorithm are a critical matter of interest because inventory incurs cost and is at risk of obsolescence.

Let us assume that a distribution company fulfils customer orders on demand from inventory.

The company tries to maintain inventory levels under full control. The situation can be described as follows:

• – market demand Q_d,t is an unlagged linear function of the market price P_t. Q_d,t is proportional to P_t with a negative slope: as P_t increases, Q_d,t decreases, which conforms to economic theory;
• – the adjustment of price is effected not through market clearance, but through a process of price-setting by the sales department of the distribution company. At the beginning of each time period, the sales department sets a selling price for that period after having taken into account the inventory situation. If, as a result of the preceding-period price, inventory accumulated, then the current-period price is set at a lower level than previously done. However, if depleted instead, then the current price is set higher than the previous value;
• – the change in price made at the beginning of each period is inversely proportional to the difference between the supply quantity and the actual market demand during the previous period, i.e. the on-hand inventory at the end of each period.

Under these assumptions, the inventory control model can be made explicit by the following equations:

P_t+1 = P_t − σ[Q_s,t − Q_d,t], where σ denotes the stock-induced price adjustment.

By substituting the first two equations into the third one, the model can be condensed into a single difference equation

or P_t+1 − (1 − σ β)P_t = σ(α − K)

Its solution is given by

(K − α)β⁻¹ is the equilibrium price P when the market demand matches the fixed supply quantity.

The expression of P_t becomes

The inventory level I_t at the end of the time period t is the inventory I_t−1 at the end of the time period t − 1 plus the quantity K supplied during the time period t minus the market demand Q_d,t during the time period t.

The dynamic stability of P_t as well as the changes in inventory levels at the end of time periods (I_t − I_t−₁) hinges on the parameter 1 − σβ. Table 3.2 presents the situation.

Let us investigate the sensitivity of the initial value P₀ on the explosive oscillations of P_t when σ > 2/β. When P₀ is changed in P₀ + ε, P_t becomes

increases exponentially to ∞ as t →∞. (I_t − I_t−1), the inventory level at the end of each time period t, follows the same behavior.

3.2.3.2. Business monitoring systems

Any business organization is made up of subsystems integrated into a total system. These subsystems are business functions (production, sales, human resources, finance) consisting of human and automatic operators and equipment of various sorts, all of which interact with each other to form the hub of business operations. Data flows between these subsystems secure the appropriate orchestration of operations arranged in a logical sequence for achieving set objectives as efficiently as possible.

Business systems require resources to enable production as well as administrative functions to operate. Finance is considered the major resource because it is an enabling resource for obtaining the other resources. Information is a critical resource essential to the effective coordinated operations of the other resources. It is managed by what are reffered to as “information systems”.

When systems are highly integrated, which is the current situation, they become too complex to understand, not only to the layman but also to their human operators. If one part of the system ceases to function correctly, this may cause the system as a whole to degrade or even to come to a full operational standstill. The trendy integration of systems (Systems of Systems – SoS) does not help to deal with random influences as they occur without too much disruption. Decoupling subsystems is not a viable solution either. The graceful degradation of systems performance has to be taken into account at the design stage and its robustness checked by crash tests to explore the potential systemic failures and establish the appropriate countermeasures to deploy in case of emergency.

In addition to technical problems, the influence of human elements on a complex system must not be forgotten. When management sets objectives, the behavior of people involved in their fulfillment has to be taken into account. Discrepancies may be revealed between organizational objectives and the objectives of the personnel, which form the very fabric of the organization. A great deal depends on the motivational influences and their values to individuals. The behavioral aspects of people at work cannot be programmed. Conflicts of goals can be a serious source of disruption and dysfunction in business environments.

Time lags in data transmission are an important source of untimely and thus inappropriate decisions. Instructions applied to a subsystem with a delay, after it has evolved from the state when sensed, are prone to trigger behaviors difficult to bring under control. We shall focus on this feature in the following narrative.

The control entity is assumed to be employing a battery of state sensors to control a variety of targets. We may think of it as monitoring the behavior of planned objectives, adjusting the action instruments, sensing the effects of the instruments on the targets and readjusting the action instruments until the sensed effects are in line with the set objectives. This closed-loop process of observation and action may be continuous or trigged at predetermined time intervals. Regardless of the type of closed-loop process, when data are collected from different sources, they have to be aggregated and processed to be made exploitable by the control unit.

The time lags may be thought of as comprising three parts: the deployment lag, the data collection lag and the reporting lag, as portrayed in Figure 3.2.

Regardless of the nature of time lags and their duration, a time-lag structure is a whole array of quantities occurring with a range of delays. It is outside the scope of this book to go into detailed quantitative technicalities of delays, but some simple examples can illustrate what time lags mean in terms of business control. They are shown in Figure 3.3.

Assume that the planned target is subject to a step increase OA at time t₁ (part 1 of Figure 3.3). What is the response time for the planning unit to receive data signaling that the order has been fulfilled? Other parts of Figure 3.3 show how the step response can converge to a definite value. Parts 2 and 3 portray two time-phased profiles of contributions to reach the increase OA at time intervals t₁ + 1, t₁ + 2, t₁ + 3, and so on. Part 4 represents a combination of the dynamic effects shown in parts 2 and 3.

In this example, we have assumed that the increase converges along a smooth path to the desired value. Many other types of behavior (cyclical path temporarily diverging but ultimately converging, or explosive oscillations) can be observed when the state of the controlled processes is taken into account. A critical factor is the difference between the state of the controlled processes when the decision was taken at the planning level and the state when the decision has been factually applied. The corrective course of action is engineered to remedy a situation which no longer exists. This is described as the time lag between physical events and information flows. Let us take a simple example to describe such circumstances.

If a production department is reported to the central planning unit to have failed to meet the set target, corrective action could be taken to increase the output level. If the adjustment order is transmitted with delay, then it is possible that the local production manager has already taken the steps to modify the production schedule. The closed-loop system can amplify the disequilibrium between the desired output and its actual level. Alternative positive and negative feedbacks intended to increase or decrease the output level and reduce the stock-out situations, when implemented at the wrong time, will contribute to create explosive oscillations when iterated. Deviations are amplified instead of being damped. Figure 3.4 portrays the effect of time lags on an oscillating system. Data relative to Figure 3.4 (production per week, delayed information flows, corrective actions taken) are summarized in Table 3.3

Business activities are always subjected to internal and external random perturbations and, as a result, hardly ever achieve steady operational states. They often exhibit oscillating results, which is perceived as a dysfunctional chaotic pattern. Feedback mechanisms are engineered to hunt after the target state. Figures 3.5 and 3.6 visualize the effects of positive and negative feedback actions, respectively.

3.4.1.1. Setting the picture

A mantra heard everywhere in the business realm and disseminated by all sorts of media is digitalization. What does this tenet mean? All economic agents, corporations as well as individuals, are embedded in technical networks providing instant communicative interactions between them via telecommunication signals at the speed of light. This refers not only to interactions between corporations and their suppliers and clients but also between individuals. This situation results in overloading all the economic agents with heaps of both structured, but mainly unstructured data, which have to be processed to either discard or integrate their relevant content, if any, in decision-makers’ assessments for taking action.

The impact of interconnectedness will be enhanced in unexpected ways by application of AI (artificial intelligence) and machine learning not only in the realm of financial markets and institutions but also in all socio-economic ecosystems. Institutions’ and organizations’ ability to make use of big data from new sources may lead to greater dependencies on previously unrelated macroeconomic entities and financial market prices, including from various non-financial corporate sectors (e-commerce, sharing economy, etc.). As institutions and organizations find algorithms that generate uncorrelated profits or returns, if these will be exploited on a sufficiently wide scale, there is a risk that correlations will actually increase. These potentially unforeseen interconnections will only become clear as technologies are actually adopted.

More generally, in a global economy, greater interconnectedness in the financial system may help to share risks and act as a shock absorber up to a point. Yet the same factors could spread the impact of extreme shocks. If a critical segment of financial institutions rely on the same data sources and algorithmic strategies, then under certain market conditions, a shock to those data sources – or a new strategy exploiting a widely adopted algorithmic strategy – could affect that segment as if it were a single node. This may occur even if, on the surface, the segment is made up of tens, hundreds or even thousands of legally independent financial institutions. As a result, collective adoption of AI and machine learning tools may introduce new risks.

As individuals, what are our main feeds of data deluge? Persuasive psychology principles are used to grab your attention and keep it. The alarming thing is that several dozen designers living in California and working at just a few companies are affecting the flows of data received by more than a billion people around the planet. Furthermore, spending increasing amounts of time scrolling through our feeds from different channels is not the best support for our ability to make decisions on the ground of sound pieces of reliable information, nor for our well-being or performance. Think about our own habits in the past year or two and how any changes in behavior, and specifically spending increasing amounts of time with our feeds, have influenced our beliefs and attitudes.

There is also another element of feeds that affects our behavior. We design our own feeds, although heavily influenced by those several dozen designers who are, in fact, more or less our ghost tutors. We connect with friends and colleagues; we like and follow companies and public figures that we admire. And the source of this attraction and connection often stems from some similarity that we see in them in terms of shared values and opinions.

As a result of this, our feeds are giving us a world view that is anything but worldly. It is segmented and we are blindsided to the opinions, values, preferences and affiliations of those whom we interact with in a virtual world out of sight, and often without any implement to foster trust.

Analytics of incoming data flows in a business or administrative framework has become a need and a challenge, not only because of the very high volume and variety of data, but also because of the required filtering of high-rate data flows in terms of relevance over time from the point of view of decision-making for activity planning. Three V’s can be used to describe data deluge: volume, variety and velocity.

3.4.1.2. Data flows and management practice

Analyzing the contents of data flows received through various channels by decision-makers turns out to be a daunting challenge. Ceteris paribus the response time left to decide upon a course of action to take is a critical factor in an evolving environment where information-laden data travel at the speed of light.

Generally speaking, it has become more difficult to set priorities because the human brain has a limited capability to deal with the input of a large quantity of interdependent variables. In some cases, by the time the incoming data have been crunched into understandable management numbers and made available to the controllers in charge, it may be too late to react because these data reflect a situation that has changed. Collected data always reflect what happened in the past: are they still relevant when they reach the decision-maker involved?

Regardless of the type of industry cycles (long or short) considered, it is vital to get a full understanding of market behaviors. How are early signals of a changing world recognized? Products manufactured, services delivered, are they going to be sellable tomorrow or do they have to be changed? What is the new competitive environment? How does the chain of customers change?

To bring an answer to these questions, critical information systems have to be developed for securing relevant data feeds to the “traditional” management information systems implemented in the past decades. Why critical?

Managing a company is still analogue and not digital because human beings are analogue, and the way you manage a company is by dealing with human beings. The value chain of human resources needs to be kept intact.

3.4.1.3. Data overload

Information is essential to make relevant decisions, but more often than not it overwhelms us in today’s data-rich environment. Is a systematic framework a viable approach to make better-informed decisions?

The data-aggregation complex that has developed around political events has failed again in a spectacular fashion. As big data and analytics have become all the rage in the professional and corporate worlds, a host of pollsters, aggregators and number crunchers have assumed a central and outsized role in our prognostication of contemporary events. For example, the odds of a Clinton victory in the 2016 US Presidential election were posted at more or less 70%.

The difference between the signals relevant to the object studied and the associated noise is in tune with the zeitgeist. We are deluded by software that is perceived as black boxes. We are most often not aware of algorithm-supported models that are translated in computer programs. Any model has been evolved on the basis of premises, conditions and hypotheses that impose domains of validity to deliver trustable outputs.

Thanks to cheap and powerful computers, we can quite easily construct, test, feed and manipulate models. The fact they deal in odds and probability lends an air of humility to the project and can induce confidence of a sort. To a degree, it is the lack of human touch that makes this approach so appealing: data supposedly do not lie.

Stories gleaned from reporting and conversations may be telling but are not determinative. The plural of anecdote is not data, as the saying goes.

In our age of data analytics, competitive advantage accrues to those organizations that can combine different sets of data in a “smart” way, that understand the correlations between wishful thinking and real-world behavior, and that have a granular view of the make-up of the target market.

Marrying the available information with companies’ own experiences and insights – hopes and biases – appears to be the best way to draw conclusions. Purely data-driven approaches are not as evidence-based or infallible as its advocates like to think:

• a) in our age of mass personalization and near-knowledge about consumer behavior, polls and surveys offer a false sense of certainty. Polls ask people what they think they will do and not what they actually do. There is a big chasm between the two;
• b) poll-takers must decide what questions to ask, how to phrase them and how to weight their samples (demographic, gender, age group). It is difficult to blow a snap shot into an accurate, large picture;
• c) there is a strong human element in summarizing and presenting the findings. What weight to give to historical data or early signals? How to assess model-embedded uncertainty?

Predicting an event is not simply a matter of gathering and analyzing data according to predetermined algorithms and patterns. Emotions, desires, incentives and fears of millions of people cannot be captured without complexity of a sort.

The Brexit campaign in the United Kingdom in 2016 was a telling illustration of how, in an age of endless information, “smart” algorithms and the relentless aggregation of polling and sentiment indicators, it is possible for many data-driven people to be easily manipulated. It is a cautionary tale of how instinct and the desire to transform ambiguity can steamroll probability, especially when predicting the outcome of profoundly important emotionally charged public events.

When any type of public campaign is staged, beliefs and desires of the targeted public change along campaign progress. The analysis carried upstream to launching a campaign is prone to deliver wrong pieces of advice because evolving beliefs and desires of targeted people are difficult to assess dynamically.

3.4.1.4. Forecasting and analytics

Decisions in the fields of economics and management have to be made in the context of forecasts about the future state of the economy or market. A great deal of attention has been paid to the question of how best to forecast variables and occurrences of interest. The meaning of forecasting is straightforward: a forecast is a prediction or an opinion given beforehand.

In fact, several phases can be distinguished in the forecasting process.

3.4.1.4.4. What should I do?

In the last step, if the context is stable and foreseeable enough, prescriptive analytics can be envisioned. It relies on a combination of data, mathematical models and business rules with a variety of different sets of assumptions to identify and explore what possible actions to take.

It stretches the reach of predictive analytics to organize courses of action that can take advantage of predictions.

Figure 3.11 shows the phased forecasting process in terms of value added.

There are several distinct types of forecasting situations including event timing, event outcome and time-series forecasts. Event timing refers to the question of when, if ever, some specific event will occur. Event outcome forecasts try to estimate the outcome of some uncertain event. Forecasting such events is generally attempted by the use of leading indicators. Time-series are derived from data gathered from past records. We focus our attention here on time series.

A time series x_i is a sequence of values generally collected at regular intervals of time (hourly, daily, weekly, monthly, etc.). The question raised is: when at time n (now), and considering the set of past values available now I_n, what is the distribution of x_n+h over a time horizon h?

The answer to such a question is generally engineered by the use of an underlying parametric model. Models rely on assumptions that give an oversimplified perception of daily life. Forecasting activities to establish planning follow this iron rule. The future is derived from past measurements recorded along a longitudinal time range and presented in time series as data points. Model parameters are adjusted from experimental data. Several methods have been worked out and depend mainly on the time horizon considered and the volatility of the situation under study. It is beyond the scope of this book to elaborate on the quantitative methods that have been developed to deal with this issue. Two seminal references are given by Box and Jenkins (1970). Let us mention some models: logistic model, Gompertz model, ARMA models (autoregressive moving average) and exponential smoothing models.

“Without the right analytical method, more data give a more precise estimate of the wrong thing”: The Economist – Technology Quarterly, 7 March 2015.

Hence, a three-stage approach to analysis is suggested:

• – first stage, choice of a small subset of models;
• – second stage, estimation of model parameters;
• – third stage, comparison of the different models in terms of adequate fitting of past data.

Bear in mind that the future is not an extension of the past. Model parameters that fit past data well can change over time. The very model can become obsolete because the situation has changed.

Fractals are patterns that are similar across different scales. They are created by repeating a simple process over and over again. The fractal business model creates a company, out of smaller companies, which in turn is made out of smaller companies and so the pattern continues. What is in common with all these companies is they are all subject to the same rules and constraints. A fractal is a form of organization occurring in nature, which replicates itself at each level of complexity. When applied to organizations, this characteristic ensures that at each level, the needs of the lowest level are repeated (contained) in the purpose of the organization.

The fractal theory has brought about a new mind set for data analysis. At first sight, a time series of a single variable appears to provide a rather limited amount of information about the phenomenon investigated. In fact, a time series contains a large number of interdependent variables that bear the marks of the dynamics, allowing us to identify some key features of the underlying system without relying on any modeling. The issue is to extract patterns from an irregular mishmash of data.

A central issue is addressed from experimental data: can the reconstruction of the dynamics of a complex system derived from time series data?

It can be unfolded in several sub-issues:

1) is it possible to identify whether a time series derives from deterministic dynamics or contains an irreducible stochastic element?

The key parameter of a time series is its dimension. Dimension provides us with valuable information about the system’s dynamics. The mechanism used to derive the dimension of a time series is presented in appendix 2.

In general, fractal dimensions help show how scaling changes a model or modeled object. Take a very complex shape, graphed to a scale, and then reduce the scale. The data points converge and become fewer. This kind of transformation can be measured and judged with fractal dimensions. When a length (dimension 1) or area (dimension 2) is reduced by a factor of 2, the effect is well known; namely the length is divided by 2 and the area by 2². When a d-dimensional object is reduced by a factor 2, its measurement is divided by 2^d;

2) assuming that the time series converges to what is called an attractor (a point or a stable behavioral pattern), what is its dimensionality d? d = 1 reveals that the time series refers to self-sustained periodic oscillations. If d = 2, we are dealing with quasi-periodic oscillations of two incommensurate frequencies. If d is not an integer and larger than 2, the underlying system is expected to exhibit a chaotic oscillation featuring a great sensitivity to initial conditions, as well as an intrinsic unpredictability;

3) what is the minimal dimensionality n of the phase space within which the attractor, if any, is embedded? This defines the number of variables that must be considered.

3.4.1.5. Market demand sensing

“Real time” demand sensing between the stakeholders of short or long supply chains in consumer industries is a common practice to fulfill customers’ orders “just-in-time”. Can this data source be used to elaborate long-term forecast? The answer is clearly no. Long term must be appreciated against the manufacturing cycle time of manufactured products and their lifetimes. The time to manufacture a car is much shorter than that for a jet engine or a railway train. Even in “low-volume” industries such as aerospace or railway train manufacturing, where the lifetimes of products exceed several decades, demand sensing is becoming more relevant because of its applicability not only to the provision of repair services and spare parts, but also to assess the needs of new products with innovated features. Aircraft engine manufacturers routinely have live streams of data coming from their products during flight, for example, allowing the manufacturers to monitor conditions in real time, carry out tune-ups and set spare-part inventory. Software packages embarked on railway trains fulfill their operational monitoring and even their collaborative interactions without any central control left.

When it comes to implementing demand sensing, companies divide into two camps: those that tend to build their solutions in-house, using open-source or proprietary algorithms, and those that use a range of software solutions, fully or partly tailor-made, provided by software houses.

However, the bottom line is that all that is required to make demand sensing work is the willingness to spend some time looking for potential demand signals, putting them into an analytics engine and integrating the results into supply chain planning and execution.

Four broad areas of data come into play here, which are:

• 1) structured internal data, such as that from a PoS (Point of Sales) system, e-commerce sales and consumer service;

• 2) unstructured internal data, for example, from marketing campaigns, in-store devices and apps;
• 3) structured external data, which includes macroeconomic indicators, weather patterns and even birth rates;
• 4) unstructured external data, such as information from connected devices, digital personal assistants and social media.

3.4.2.1. Change management: a short introduction

Many articles have been written by academics and consultants on what we call “change management” over the past hundred years. They focus on critical priorities for an organization’s change effort through learning about its own structure and operational practice, adapting its workforce skills to its new technological environments and capitalizing on acquired knowledge thereof. Change and learning are organically intertwined.

Consultants aim to help companies that find themselves facing new technologies or a change in customer expectations. Since the beginning of this century, many CEOs have experienced disruptions to their businesses and the trend can be anticipated to keep going globally. In spite of advice provided by consultants in the form of guiding principles that are supposed to apply to every company, the human impact of business reorganizations imposes a high levy on employees.

Every employee looks at the organizational change from the standpoint of how he or she will personally be affected. Selfpreservation becomes a major concern. Change management often, not to say always, affects the workflows associated with the operational procedures people are used to. Furthermore, in many cases, the change in the chain of procedures is implemented using software packages, the deployment of which is embodied in a project called “MIS” (management information system).

This situation is prone to developing dysfunction in business operations and, as a result, a loss of control spiraling into disorder and chaos. Until personal career issues have been resolved satisfactorily, employees are too preoccupied with their own situations to effectively focus on their work and develop resistance. Even if they are given a good explanation of the rationale for the changes and of the possible alternatives and trade-offs, they have to come to terms with a new working environment. In addition, as a business organization operates as a set of functional subsystems which influence each other, it reveals that it is difficult to trace back the cause of a long-range effect within the system. This operation often resembles looking for a pin in a hay stack.

The consultancy literature often treats the topics of “change management”, “top-down management” and “bottom-up management” differently. Each of these is often engineered in a complex and changing corporate environment without providing the top management with a clear picture of the issues at stake. The key success factor is aligning all the working forces so that they turn responsive (change-management), informed (bottom-up) and properly executing (top-down). Although this statement is simplistic, and itself requires further insight, the point that there is no one best approach for all issues should be kept in mind. The bottom-up approach will be poorly executed if the top-down approach does not provide oversight. The top-down approach will be poorly designed and will often miss intended targets if the bottom-up approach is not leveraged as an asset. The change management methodology will not be effective unless it similarly consults with all top-down tiers of management (to identify priorities) as well as the bottom-up tiers of employees (to identify risks, barriers, and threats).

The bottom-up approach in change management would typically be viewed as preventative in a business model where the external environment is greatly emphasized over the internal environment (of the company). One could often say that change management seeks to prevent threats to the company’s advantages. This is in contrast to the traditional model, wherein change management seeks to leverage and take advantage of existing opportunities in an effort to increase the company’s competitive advantages. Ignoring the internal environment during the change management process is a risk that does not clearly address the following two questions:

• – purpose: what is the change management for?;

– action: what are managers responsible for to secure the success of the change management project? The top-down, bottom-up, framework provides the multi-directional feedback that is critical for an accurately designed change management methodology that is executed effectively. Failures in change management can frequently be linked to providing partial answers to incomplete questions. Roughly speaking, consultants can be said to deal with “shop floor” realities, whereas academics are supposed to elaborate on theories, principles and paradigms underpinning the approaches engineered by consultants. In the next section, we scrutinize what academics have elaborated on the subject matter “change management”.

Change management can be linked to the notion of bifurcation found in the science world. Any organization can be modeled as a mechanism converting inputs into outputs. As long as the inputs are coherent with the outputs along with the inner transforming mechanisms, the system is stable. As soon as the outputs change, for instance, the products or services offered to the demand side, either the inner transforming mechanisms or the inputs, or both, have to be adjusted. Another situation can be imagined. In order to keep abreast with competitors, technological changes have to be introduced, which can entail a drastic upheaval in organizational patterns and skills in the labor force. When a breaking point in terms of output versus input is hit, several routes to a new organizational pattern can be followed according to plans or, more likely, contingent circumstances, which are always difficult to forecast. The breaking point can be interpreted as the elasticity limit of the inner transforming mechanisms to secure the delivery of outputs of a certain sort from available inputs. This analogy with respect to the notion of bifurcation is depicted in Figure 3.12.

3.4.2.2. Change management: the academic approach

In section 2.7, we explained that the many aspects of what is called the organization theory defy easy classification. We then categorized different approaches into structural and functional theories on one side and cognitive and functional theories on the other side. This classification was proposed from an enterprise point of view.

Organizational structures are found in many environments other than the enterprise. Other criteria have been considered to classify them according to their functions or goals, nature of technology employed, ways to achieve compliance or the beneficiary of the organization. D. Katz and R. Kahn (1978) use functions performed or goals sought as the criteria to categorize organizations. According to them, production or economic organizations (manufacturing, logistics, distribution, retailing) exist to provide goods and services for society, whereas pattern maintenance organizations (education systems) prepare people to smoothly and effectively integrate other organizations. The adaptive organization (research and development) creates knowledge and tests theories, while the managerial or political organization (regulatory agencies) attempts to control resources use and authority.

In order to keep track of the baseline ideas of management thought in the profuse variety of literature produced about organizations, it helps to distinguish between organization behavior and organization theory when organizational change, adjustment or reconfiguration is considered.

Organization behavior primarily deals with individual and group levels. It covers motivation, perception and decision-making at the individual level and group functions, processes, performance, leadership and team building at the group level.

Organization theory deals with organization level such as organization change and growth, planning, development and strategy. Management of conflicts between groups and/or organizational units is dealt with at the intersection of organization behavior and organization theory.

From the point of view of change, the metaphor of organizations as open systems subject to external pressures facing them has become an accepted mindset for guiding management thought. Several corpora of thought that have direct or indirect connections with the reasons why organizations have to change have developed. Let us review the main ones.

Contingency “theorists” strive to prescribe organizational designs and managerial actions most appropriate for specific situations. It is clearly a perspective as an offshoot of systems theory applied to the study of organizations for organizational analysis.

“Contingency” theorists have tended to examine organizational designs and managerial actions most appropriate for specific situations. They view the environment as a source of change in open systems. Simultaneously, they highlight the interdependence of size, environment, technology and managerial structure, and their compelling congruence for business success. P. Lawrence and J. Lorsch (1967) who pioneered this approach argued:

“Underlying this new approach is the idea that the internal functioning of organizations must be consistent with the demands of the organization’s tasks, technology or external environment and the needs of its members if the organization is to be effective. Rather than searching for the panacea of the one best way to organize under all conditions, investigators have more and more tended to examine the functioning of organizations in relation to the needs of their particular members and the external pressures facing them. Basically this approach seems to be leading to the development of a ‘contingency’ theory of organization with the appropriate internal states and processes of the organization upon external requirements and member needs”.

Another group of theorists viewed environments as all-powerful in determining the fate of organizations. Drawing on the work of Charles Darwin, this group presented an “ecology” model that portrays successful organizations as the fittest in a competitive arena, where rules are imposed by the environment. H. Aldrich (1979), in an influential book, argued:

“The population ecology model differs from traditional explanations of organizational change […] it focuses on the nature and distribution of resources in organizations’ environments as the central force in change, rather than on internal leadership or participation in decisionmaking”.

Organizations as open systems, like living organisms, depend on their environment for critical resources, and a key managerial task is the appropriate provision of human-socio-cultural and political, technical, functional and informational resources.

All the arguments developed by different theorists converge to the point that organizational changes are induced or compelled by forces originating from organizations’ environments. The digital technology being introduced in all sectors of our societal context is a confirming and convincing example of this standpoint. It is a driving force for triggering deep reconfigurations of many an economic sector, beyond sheer adaptations. The interplay between the dynamical environment and current internal structures is a source of complexity to bring about a controlled evolution and to avoid disruptive situations running out of control.

3.4.2.3. Change management and social networks

3.4.2.4. Change management, data deluge and information systems

Functional subsystems influence each other, but the informational subsystem plays a special role because all organizational changes are reflected in MIS (management information system), which provides communication and coordination between all functional subsystems through the control of resources.

Figure 3.13 depicts the interplay between resources and primary processes in a manufacturing enterprise. Information is singled out at the same time as a resource, which mirrors the functioning of business models and secures collaboration and cooperation between resource management units.

Hereafter, we focus on the features of an informational subsystem, which have to receive special attention within the framework of a change management project.

Transmission of information-laden data takes place through both formal and informal channels.

The formal channel refers to what is called the MIS. Collecting, gathering, processing and transmitting data according to set procedures embodied in software programs and mirroring management tools are the functional capabilities of information systems.

The way business agents or groups of agents make use of information systems can deviate significantly from the intentions of their designers. Many reasons explained their attitudes: misunderstanding of new procedures, refusal to change existing practice, tighter control of activities by eavesdropping, breakdown of informal channels of communication and so on.

Formal elements of information about structures and operations are attractive to company management because they are tangible. They can be easily defined and measured. However, they are only half the story. Many informal communication channels parallel or short-circuit the “official” patterns of communication flows. They are based on trust between the stakeholders involved. In case of reorganizations, many companies reassign decision rights, rework the organization chart or set up knowledge-sharing systems – yet they do not see the results they expect. That is because they have ignored the more informal, intangible building blocks. Norms, commitments, mind sets and networks are essential in getting things done. They represent (and influence) the ways people think, feel, communicate and behave. When these intangibles become desynchronized with one another or with the more tangible building blocks, the organization falters.

The informational subsystem must be identified as a subsystem in itself because advances in technology have made it the hub of coordination between all other business functional subsystems and business partners (suppliers, clients).

The critical issues at stake are delays in data transmission as already described to shun untimely decisions and their consequences, and accuracy of disseminated data (the destructive effect of fake or biased news does not need to be proved anymore). If these issues are not brought under full control, the whole system may run out of control because counterproductive decisions are taken untimely, as shown in the section “emergent behaviors” with unintended consequences (hunting and oscillating).

In the next sections, we focus on information systems and their architecture challenge in the era of data deluge.

3.4.2.4.2. Traditional layer architecture of information systems

Layer architecture is based on layers of functional operations and strongly contrasts with a hierarchical architecture developed on management levels (strategy, tactics and operations).

In a layered architecture, the underlying layers relate to the implementation, that is, to the different types of resources, while the top layers relate to the business. Figure 3.6 shows a schematic example of a layered architecture. A business system may well have more or fewer layers than the ones shown in Figure 3.14, and it can also have layers in several dimensions, so that one layer can itself be layered.

The basic principle of any layered architecture is the delivery of services to a layer by underlying layers. When standardization is looked for, formats of services provided are defined, not the very mechanisms engineered to supply services under consideration.

The major benefit derived from layer modeling is the versatility it has for focusing on “relevant” views of a complex system and to trace back, when required, relations with detailed or aggregated descriptions.

• – The resource layer contains the different concrete resources needed for realizing the organization. Here, human resources with different kinds of skills – education, management, experience and so on – and different types of equipment such as information systems and transmission and switching equipment are found.
• – The infrastructure layer consists of functions generic to the type of business being modeled. Here, functions to support network and services management processes, accounting processes, personnel administration, office administration, and so on are found.

• – The common business process layer contains internal processes generic to the type of business modeled. These internal processes can be used by many different business processes within the company being modeled. Some of these processes can be described as cooperative services allowing for a structured workflow between the various activities of a single business process or different business processes.
• – Finally, the business system layer consists of the different business processes. These processes must be chosen to provide the top management with synthesized views of the enterprise. These views should be described in terms of performance indicators able to trigger leverage on relevant operations.

In systems with a layered architecture, it is important to define the interfaces between the various layers. Object-oriented technology provides an elegant method of doing this. This approach will be reproduced by substituting object for activity.

Every layer is a composite set consisting only of its constituent activities. The interface to a layer is made up of the interfaces to its public activities. By public, it is meant here to be accessible from outside. How a layer can be used by layers higher up in the structure is an important architectural decision. There are several such types of layer relations to choose from. Here, we present two of them:

• – the use of activities in the lower layer is encapsulated in the activities of the higher layer. The design of the higher layers uses the public interfaces of the basic activity layer as primitives; this is because the interface is encapsulated in the overlying layer’s activities and can only be seen in these activities’ implementation parts. The interface should not be used in the documentation, which describes the interaction between activities in the higher layers. A larger business system consists of thousands of activities that communicate with one another in an extremely intensive way; by concealing the communication with underlying activities inside the activities in their own layer, all documentation will be simplified.

Since they are designed to be extensively reused, basic command activities should be managed in a very different way compared to business activities: they should be carefully tested before being released; they should not be allowed to change until all business activities dependent on them have also been changed and so on;

• – activities in the lower layer need not only be encapsulated in activities of the higher layer but can be used in the design documentation of the higher layer; however, they cannot be changed by the designers of the higher layer. The common business process layer grows and matures during the development of several different business subsystems. It should contain the common activities of an entire business, and it should, therefore, become relatively stable after a number of releases. The layer should be managed in a way different from the business system layer, but similar to the infrastructure layer. An important difference, compared to the objects in the infrastructure layer, is that it is necessary to show the activities in the common business subsystems. When designing a business subsystem, you may show the common business processes you are using, but you are not allowed to make any changes in them whatsoever. You are only allowed to make changes in your own activities, that is, activities to realize your business subsystem. If a common business activity is to be changed, all business subsystems using that activity will need to be changed at the same time.

Every layer can be divided into subsystems. The subsystems in the highest layer, if any, cover large business areas and are supposed to reflect the core business processes as viewed by the top management. They can be dependent on each other. In this case, they are interrelated through an activity or a process of the underlying layer.

Business system areas and layering are complementary. You can apply both techniques to describe a complex organization. First, you use the business system area technique to find the different business system areas and their associated processes, and then you develop underlying support processes within the framework of a layered architecture. The layers below the highest will be used for all of the business system areas.

Figure 3.15 describes the concrete example of a layered information system architecture pertaining to a manufacturing company. This architectural pattern can be easily adapted to other types of organizations.

The specific features of a layered architecture are best understood by contrasting it with hierarchical information system architecture for a manufacturing company as depicted in Figure 3.16.

3.4.2.4.3. Prospective overview of the future structures of information systems

It is common saying that we are in an era of rapid digital transformation. All sections of society, individuals as well as businesses and public administrations are embarked on a stream of change in the ways we interact and communicate. Virtual environments, namely server platforms, will be the hubs of faceless human social life in a global economic competition field.

Businesses are already embedded in agent networks of a sort, and in the past decades, they had to adjust to their evolving environments. Increasingly more relations with their suppliers and clients have become paperless and what is called dematerialized. The data sources they feed them will drastically increase in the forecast future altogether by their number and variety (CRM, Web portals, emails, social networks, collaboration systems, HR portals). In particular, IoT sensors monitoring real-time manufactured products will generate large volumes of remote data inflows and will bring unstructured data to businesses in order to be adequately processed. New policies of data governance will have to be promoted and deployed for reaping the benefits of the availability of “data lakes”.

The upper layer (customer care) and the lower layer (resources management) of the I S architecture depicted in Figure 3.15 are expected to be the most impacted. Their configuration will go through a drastic disruption to meet the processing requirements appropriate for interacting electronically with clients (customer care) and suppliers (resources management) via portals, e-shops, EDI systems, IoT sensors and so on. Each agent of these layers (sales, order handling, problem handling, purchasing, inventory, etc.) will receive or send managed transferred files from/to other agents, which are either other in-house agents or the front office, the enterprise’s trading partners, clients or suppliers.

IT systems are purpose-built for supporting efficient collaboration between all the in-shore and off-shore stakeholders of the business entity. Three types of systems can be identified:

System of records (SoR)

They are the traditional EDP (electronic data processing) systems that manage structured data related to business operations (financials, payroll, CRM – customer relationship management (inventory, order entry, HR – human resources), product life cycle management).

They have been designed on the basis of management concepts such as MRP, ERP and CRM (Briffaut 2015). They include data storage and access systems. They are the authoritative source of enterprise data.

System of engagement (SoE)

This system consists of the front offices of a business with its clients on one side and its suppliers on the other side. It is device-centric and associated with a variety of communication channels ranging from unstructured data from call centers, mobile devices, IoT sensors and so on to structured data.

System of insight (SoI)

This system is designed to convert unstructured and heterogeneous data sets received from various sources into structured data that can be fed into the system of records to support decision-making.

IBM (2015) has designed an SoI building block model for offering an enterprise architecture to deal with incoming data flows. It is represented in Figure 3.17.

The consume and collect blocks refer to extracting data from the transactional system that will be used in the decision-making process and to collecting and aggregating it in view of further processing. The analyze and report blocks are intended to make sense of incoming data streams and prepare them for feeding the detect and decide blocks. The detect capability refers to discovering the existence of an event or a situation that requires one to make decisions. The capability of the decide block is to engineer the appropriate steps for delivering a decision.

The same model can be used to architecture the interaction of an e-enabled enterprise with its suppliers. In this case, data are generated by in-house transactions and are transformed in prepared data to be sent to suppliers. It is clear that some in-house data sources correspond specifically to external data sources.

The SoR, SoE and SoI systems are subject to constant adaptation in order to be kept in line with the enterprise evolution. In order to keep their adaptation under control, they are generally configured within a framework that is called an enterprise architecture (EA). A wide variety of definitions have been given to this concept. Gartner in the IT glossary of enterprise architecture (gartner.com, retrieved July 29, 2013) defines it in the following terms:

“EA is a discipline for proactively and holistically leading enterprise responses to disruptive forces by identifying and analyzing the execution of change toward desired business vision and outcomes. EA delivers value by presenting business and IT leaders with signatureready recommendations for adjusting policies and projects to achieve target business outcomes that capitalize on relevant business disruption. EA is used to steer decision-making toward the evolution of the future state architecture”.

Traditionally, the role of the IS system within the framework of a business approach is visualized as shown in Figure 3.18.

When a business unit is described as a system, the purpose is controlling its business operations. Three entities have to be identified: the controlled system, the controlling system and the information system. The controlled system, often called the transformation system, because it converts inputs into outputs, is generally modeled as a process. The relationships between these three entities are explicit from Figure 3.18. (for more details, see Briffaut (2015)).

It is noteworthy to elaborate Figure 3.18 to understand the features of the system approach to business description and especially the meaning of direct and indirect control. Direct control refers to direct action on the controlled process to maintain or change its state. Indirect control resorts to some entity external to the system to influence the state of the controlled process by means of inputs.

Let us take an example in a manufacturing context to explain how the messages exchanged between the involved entities are connected and how their contents trigger decisions. The controlled process is assumed to be a manufacturing process made of storage and production activities. A message coming from the market place (environment data) is captured and processed by the information system (IS). The message content says that a market slump is forecast. It is directed to the production scheduler in an appropriate format (control data). As a result, the scheduler decides to reduce the production level by releasing orders to the manufacturing shops (direct control) on the basis of inventory levels (process data) and to send orders to suppliers to decrease the number of deliveries (indirect control).

What has to be changed to cope with an e-enabled context of digital interaction with clients and suppliers and of big data flows? The answer is straightforward: indirect control messages and environment data flows have to go through a layered system of SoE, SoI and SoR, as sketched in Figure 3.19.

3.4.2.4.5. Specific features of manufacturing information systems

Manufacturing corporations are faced with specific situations. A buzzword found in the technical press and bruited at conferences organized by a wide spectrum of stakeholders revolving about this industrial sector is digitalizing manufacturing. Examples of this broad notion are 3D printing, the Internet of Things (IoT), mass customization and last but not least big data or data deluge. In fact, this paradigm aims to marry information and communication technology to the process by which highly complex products are designed, manufactured and delivered to the final customer. Data flows add a massive amount of leverage in advanced manufacturing.

Information technology in advanced manufacturing

Large technology companies operate what can be called a digital factory where the real world, hardware and software merged with the virtual world, where simulation software elements allow us to assess whether the designed product meets the operational requirements set by the designers and can be manufactured with the equipment units and competencies available in the company.

The concept of CIM (computer-integrated manufacturing) was introduced by IBM in the seventies of the past century. The basic idea was to memorize in one database all the specifications of products in terms of materials contents and manufacturing operations associated with equipment units. Figure 3.20 describes the main features of this concept.

Let us explain the main functions and their relationships. The mission of the product development function is to design products in terms of components, materials required to manufacture final products either manufactured in-house or obtained from third-party suppliers (BM – Bill of Materials) and manufacturing operations expressed as programs for numerically controlled (NC) tools according to a routing appropriate to the job shop layout.

Customer orders are collected by the sales department, and their fulfillment is scheduled for making cost-efficient use of manufacturing resources. The implementation of schedules triggers the materials requirement planning function so that manufacturing and procurement orders are released to meet the delivery due dates of production schedules of finished products.

A software package called PLM (Product Lifecycle Manufacturing) allows us to simulate production processes and robotics ahead of time for optimizing engineering, procedures, quality, load time and uptime in order to make production more flexible and reliable in terms of delivery lead times. Before a manufacturing line is built and operated, simulation allows one to not only reduce and bring under control the time to put a new product to the market (time to market), but also to operate the manufacturing line with limited downtimes.

Sensors collect real-time data to software packages about the states of production lines. These data are used for predictive maintenance. In order to fulfill customer orders on time with the level of quality committed, it is critical to rely on fully operative equipment as scheduled by planning. Conventional approaches to equipment maintenance establishes a schedule for inspecting and maintaining all the components of the equipment units. This schedule is fixed and is based on the mean time before failure for each of the components. In a connected multi-factory environment, analytics of maintenance data can be shared with other sections of the enterprise manufacturing facilities that use the same type of tool. Taking one step further, the data from an individual tool can be sent to the equipment vendor. This piece of information can initiate discussions about improvements and feed into the supplier’s delivery schedule.

Further, the concept of proactive maintenance does not stop at the factory wall. The challenge is to make meaningful analytics out of these data for the benefit of customers. Many people do not yet understand why customers would pay for information making use of the products they have bought, easier, better, less costly or more valuable when IoT is taken into account for interacting with vendors.

Sensors embedded in products deliver data that make sense in terms of optimized operations, utilization rate, predictive maintenance and better support to the user. This set of data can become the holy grail of innovation with the proviso that data analytics is carefully carried out by manufacturers. In fact, weak signals can be hidden in the main stream of data flows, and each type of weak signal can be relevant to different business stakeholders.

Data analytics gives a company a lot of information it can use to optimize and shorten the value chain. This value chain consists of the suppliers of your suppliers (supply network), your company, your customers and their customers (distribution network). The purpose is to make your products faster, more cost-efficient, more flexible in small lot sizes while providing customers with enhanced support.

The manufacturing company can cut out different links of its value chain because they provide the least value in the company’s value chain. Hence, it needs to be understood. Where is the manufacturing company in the global value chain? How can it remain a key link by providing more value than anyone else in the chain?

Customers do not look any longer for products but for the services delivered by products. This is a paradigm shift. Many manufacturing companies claim that they sell solutions, applications and comprehensive systems versus products, namely value to customers instead of function capabilities. Car manufacturers are a telling example with the Internet of Things for cars. A connected car is a digitized vehicle with a built-in wireless network enabling advanced information and entertainment systems, real-time location and routing services, vehicle-to-vehicle and vehicle-to-infrastructure communication systems and remote diagnosis and update applications. It is easy to figure out that this approach to market, especially remote diagnosis and update applications, not only entails complex organizational methods and the deployment of appropriate information systems for car manufacturers, but also generates flows of data that have to be processed and analyzed.

A state of mind different from B2C and B2B offerings is emerging, namely “business to society” enterprise (B2S). The B2S concept means to contribute to civil society’s development in the world through providing people with reliable, safe services and making life in cities more bearable. This may reveal a start for changing relationships between industry and civil society.

Company employees are a special part of society and as such have to be made aware of their roles with respect to the social community they belong to. As company members, they deliver products/services to meet the requirements of their community fellows, add value to the company inputs, allowing them to receive salaries and to make other economic actors active by fulfilling their own needs. The final purpose of any company is durability through sustainable courses of action and societal responsibility.

Industry 4.0

What is meant by Industry 4.0? Industry 4.0, as Industry 3.0 draws on telecommunication services based on the Internet protocol of communication between message emitters and receivers. Industry 3.0 refers to the automation of machines and processes monitored by a single computer platform associated with network facilities. Industry 4.0 covers the concept of systems run on interconnected platforms combining engineering and manufacturing that can deal with global networked organizations.

For the sake of clarity, access to a computer server to retrieve data-laden messages on requests with the Internet protocol was called Internet 1.0. When an additional service became available, namely the capability from a remote terminal to feed data into a computer server, this service was called Internet 2.0.

The manufacturing flow is organized to make small lot sizes to meet customer orders on demand. The software steers the manufacturing process so that small lot sizes can be delivered cost-efficiently, at the due date and with the quality level committed to the customer.

The manufacturing process is monitored by machines talking to machines. Production is stopped when a defect is detected so that finished products do not have to be re-worked in case failures are detected at the final stage: “do it right the first time” is the paradigm underlying the manufacturing procedures and layout.

Industry 4.0 is intended to take the cost of scale to zero: regardless of the lot size, the unit cost price is about the same. This technical situation offers the possibility to convert standardized products into customized products, allowing for a competitive advantage. A product whose specifications are sent via the Internet will be built to order. Specific labor skills are required. Cost-efficient 3D printing makes possible lot size of one, the dream of the just-in-time endeavor launched by the Japanese industry in the 1970s.