Logistics decision-making based on the maturity assessment of imperfect knowledge

The contemporary networked world, characterised by the Internet of Things and Services, is leading to the emergence of smart grids in the field of energy supply, sustainable mobility strategies, such as smart mobility and smart logistics, and smart health in the realm of healthcare (Kagermann, Wahlster and Helbig, 2013). In the area of manufacturing, this trend led to the fourth stage of industrialisation — Industry 4.0, which focuses on creating smart products, procedures and processes. Smart factories are capable of managing complexity, are less susceptible to disruptions and able to manufacture goods more effectively and efficiently. In this environment, human beings, machines and resources communicate with each other in the same way as in a social network. Its interfaces with smart mobility, smart logistics and smart grids will make the smart factory a key component of tomorrow’s smart infrastructures. This drives forward the transformation of conventional value chains and the emergence of new business models. A comprehensive literature analysis of the problems related to the challenges of the so-called the fourth industrial revolution is already discussed in the publication by Ślusarczyk et al. (Ślusarczyk, Haseeb and Hussain, 2019) and will not be duplicated here.

The implementation of Industrie 4.0 requires an interdisciplinary approach and close cooperation with other key areas. Consequently, logistics plays a key role in the mutual coordination of individual processes implemented under these systems. Fig. 1 presents a simplified interrelationship between individual subsystems that compose the “system of systems” of Industry 4.0 (Bukowski, 2016a). Also, it illustrates the symbolic role of logistics management as Logistics 4.0. As the diagram suggests, smart equipment, which is operated using robots and sensors, as well as smart information and communication technologies (ICT based on Big Data and Cloud Computing) are of fundamental importance for the effective implementation of processes under Industry 4.0 and Logistics 4.0. The quality of knowledge acquired from ICT has a particularly significant impact on making the right decisions within these systems. Therefore, an important issue is to develop a method that allows objective quality assessment of the entire process used to obtain and process data and information as well as the knowledge generated on this basis.

Fig. 1

The purpose of this article is to develop a method that allows for such an assessment based on the analysis of knowledge creation and decision-making using expert methods. The work is organised as follows: first, the evolution of the concept and the scope of logistics is characterised, followed by specific features of Logistics 4.0, and, on this basis, a new method for assessing knowledge maturity is proposed. Next, a brief example is presented, depicting the use of this method in the risk-taking decision-making process. Finally, the results are summarised and discussed.

The modern concept of “logistics” was shaped in the military field in the first half of the 19th century and included transport, accommodation and supply of military units, as well as transport, storage and supervision of goods intended for the army (Blaik, 2010; Michlowicz, 2002). In the 1950s, this concept of logistics started being transferred from the sphere of national defence to business. The term “business logistics” emerged, which included transport, warehousing and transshipment processes of goods within one enterprise and between different enterprises (Schönsleben, 1998; Stock and Lambert, 2001).

The turning point in the development of modern logistics was the publication of the article “Note on the Formulation of the Theory of Logistics” by Morgenstern in 1955 (Morgenstern, 1955). In the German-speaking area, the year 1973 is regarded as the date of the birth of civil logistics, which is the year of issue of the book “Business logistics — systems, decisions, methods” (Kirsch et al., 1973). The authors of this book understood logistics as “... shaping, controlling, regulating and implementing energy, information and people flows, in particular materials and products, within a given system and between systems …”. In the next years, the interest in practical aspects of logistics has grown significantly in Europe, which resulted in the creation of a practical definition of logistics in the form of the so-called 7R rules (from the English word “right” or from the German “rich-tig”). The interpretation of this rule means that the basic task of logistics is to provide the right goods, in the right quantity, at the right time, to the right place, with the right quality, at the right costs along with the right information (Lasch, 2014).

At the end of the 1980s, the representative of the German logistics school Reinhard Jünemann generalised this definition, proposing to understand logistics as “... a scientific field of knowledge including planning, control and checking of material, energy and information flows within systems” (Jünemann, 1989). Such understanding of the logistics concept has opened new perspectives for scientists dealing with logistic issues, creating opportunities for the search of theoretical models in the broadly understood area of logistics knowledge. The individual phases and directions of civil logistics development since its foundation to the current state have been described and analysed in detail by many authors (e.g. Pfohl, 1998; Witkowski, 2010; Lasch, 2014; Bukowski, 2019). On this basis, the development of logistics in business can be divided into four stages, whose brief characteristics are presented in Tab. 1.

Logistics 1.0 covered the sixties and seventies of the twentieth century and was characterised by marketing orientation. The main task of logistics was defined as “providing raw materials, semi-finished products and finished products in accordance with the principle of 7R”. The dominant approach at this stage of logistics development was functionally targeted, which means that the optimisation of logistic operations was seen as a part of separate functions and not as a part of the whole delivery process. This orientation often resulted in problems at the interface of individual functional areas (e.g. bottlenecks in flows and weak links in systems) and hindered the comprehensive optimisation of entire logistics processes (Coyle et al., 2010; Krawczyk, 2011).

In the eighties of the twentieth century, a new concept — Logistics 2.0 — was introduced. It emphasised the coordinating role of logistics in managing the flows of all goods between the place of their origin (e.g. by a producer) and the destination (e.g. a recipient). It was connected with an increase in the complexity of logistics systems and the dynamic development of the new role of logistics, namely, the organisation of returns of used materials, goods and packaging (reverse logistics). The functional approach gradually gave way to a regulatory approach, in which management decisions were usually reactions to changes taking place over time. The function of the logistic objective was subordinated to the idea of “lean thinking”, which concerned the manufacturing (lean manufacturing), organisational and management (lean structures and lean management) areas. The next phase of Logistics 2.0, at the turn of the 1980s and 1990s, was largely a response to significant market changes, namely, limiting mass production, allowing the full use of economies of scale for flexible adaptation to the needs of the unpredictable customer (logistics systems and processes, regardless of the ownership structure of individual parts of the supply chain, and the economy of scope instead of the economy of scale). Such a change in the functioning of the industry required the use of a new dynamic approach, considering the environmental variability in time. The predominant problem in this phase was the conflict between the struggle for full flexibility while adapting to the changing market requirements (responsiveness) and rising costs (Harrison and van Hoek, 2009).

The next step — Logistics 3.0 — can be characterised by the concept of the supply chain as a sequence of processes related to the flow of goods across the borders of individual organisations, ranging from the acquisition of raw materials to the final user of goods (E2E concept). Such a change in perspective related to logistics issues enabled a comprehensive approach to the optimisation of transition from flexibility in supply management to its higher level, i.e. agility. The integration of logistic systems and processes became the foundation for further stages of modern logistics development (Blanchard, 2015).

The fourth stage — Logistics 4.0 — started with the beginning of the 21st century and was driven by the ever-expanding economic globalisation. Optimisation methods based on the concept of Porter’s value creation chain to improve the efficiency of logistics systems and a dynamic approach in modelling and simulation of logistics processes have been used on a larger scale (Blaik and Matwiejczuk, 2008). At the same time, the virtualisation of logistics networks occurred within the so-called third IT platform, which is often described by the acronym SMAC (Social, Mobile, Analytics, Cloud). This trend, commonly known as e-logistics or smart logistics (Adamczewski, 2016), fully fits in the currently developed model of the so-called economics of the moment (also known as “now economy”), and implemented mostly in the real-time organisations (RTE — Real-Time Enterprise). The goals of modern logistic systems are multiple, namely, effectiveness, economic efficiency, leanness and, at the same time, agility (so-called “leagility”) and resilience to disruptions (Gudehus and Kotzab, 2009).

In practice, Logistics 4.0 was created as an integral part of the Industry 4.0 concept (Kagermann et al., 2013; Wieland, Handfield and Durach, 2016). The dynamic development of manufacturing according to the concept of Industry 4.0 is a result of such processes as globalisation, IT development (e.g. Internet of Things) and international cooperation. To complete these tasks, the production must be supported with modern logistics, i.e. Logistics 4.0. Therefore, the next section will discuss the most important features of Logistics 4.0 that distinguish it from earlier versions of logistics.

The origins of Logistics 4.0 is the concept of the fourth industrial revolution defined usually by machine-to-machine communication and autonomous processes (Strandhagen et al., 2017). This section will try to answer the following question: what are the distinct characteristics that define the logic behind Logistics 4.0? Five most important features that characterise the modern version of Logistics 4.0 age given below.

In the second half of the twentieth century, a model of “the pyramid of knowledge hierarchy” was created by Ackoff. This model became the foundation of a dynamically developing concept of knowledge management, and it has undergone a number of interpretative modifications. The pyramid of knowledge is based on data that usually occur in the form of symbols and represent individual observations of real-world states. Data represent raw facts, events, or statements without reference to other things. It does not have a meaning in itself. Usually, data is understood as atomistic tiny packets that have no inherent structure. Data can be measurable or not, analogue or discrete, as well as considered statically (e.g. data record) or dynamically (e.g. data stream). It can be obtained from various sources (Ackoff, 1989).

Information is data that are processed to be useful, and it means data that has been given meaning (e.g. relational connection), relevance and purpose. Information is partly subjective because it depends not only on the data but also on the process of their interpretation, which is based on the knowledge held by the knowledge interpreter at that time. Thus, it can be assumed that information is a collection of selected data, processed and presented in a form that can be useful to the recipient (e.g. the decision-maker). The basic condition for the usefulness of information is its ability to be interpreted in a specific context (Skytter, 2008).

Knowledge is created by integrating new information with existing knowledge (so-called background knowledge) about a particular area of interest. It requires the ability to evaluate available information and understand the reality in light of this information, in accordance with the current state of knowledge. Knowledge relies on the ability to use information effectively and efficiently to find answers to more complex questions, such as “how?”. Knowledge is emergent in relation to information — it is possible to generate new knowledge by applying processes of systematisation and structuring of information (Klir, 1991).

In 1957, Bergmann published “Philosophy of Science”, in which he made a distinction between “perfect” and “imperfect” knowledge. Perfect knowledge is based on complete theories, deterministic nature laws (such as Newtonian mechanics); it concerns closed systems, in which the initial conditions are known accurately, and the only method used in the inference is the deductive method (Bergmann, 1957).

Whereas, in applied sciences, such as engineering, social sciences, economics or management, knowledge is based on incomplete theories and indeterministic laws. It concerns open systems, in which the initial conditions are known inaccurately, and the method used in the inference is usually the inductive method. It is, therefore, imperfect knowledge. The data, on which the entire pyramid is built, come from observations or measurements whose accuracy is limited and, therefore, burdened with errors. If it is possible to repeat the same observations or measurements, the use of statistical methods minimises the impact of these errors on our knowledge. But in many cases, it is not possible, and then, the knowledge has a high degree of uncertainty (Magruk, 2017; Bukowski, 2019).

The problem of knowledge imperfection has become particularly important as a result of the increased use of vast data amounts, called “Big Data” (Corrigan, 2013). For these reasons, one of the key problems in the effective use of smart systems based on data streams, such as Logistics 4.0, is the objective assessment of the reliability of this data and the quality of knowledge based on it (Weerdmeester, Pocaterra and Hefke, 2003). For this purpose, a general framework was developed based on the imperfect knowledge concept, the diagram of which is presented in Fig. 2.

Fig. 2

The entire process of knowledge preparation for managerial decision-making can be divided into eight steps, namely: raw data acquisition, data processing to Big Data format, data veracity appraisement, information forming, information utility estimation, creation of new knowledge, knowledge processing quality evaluation and knowledge maturity rating. The basis of the procedure is the idea to combine both quantitative and qualitative methodologies in logistics research (Mangan et al., 2004). Next, the main stages of this process will be discussed in detail.

Big Data includes data sets with sizes beyond the ability of software tools commonly used to capture, manage, and process data within an acceptable period, and can be described by the following characteristics (so-called “10Vs” model):

amounts of dubious data (data in doubt), which constitutes database imperfections.

Data veracity has decisive influence on the degree of quality of information. It depends on the imperfection of the data due to (Klir, 1991) incorrectness, incredibility, implausibility, inconsistency, incoherency, vagueness, ambiguity, incompleteness, imprecision and inaccuracy. The proposal is to use the following definition of data veracity based on Bukowski (2016):

Data Veracity (DV) is the property of the data described by the five-dimensional vector of its attributes, namely:

This definition can be illustrated by the following relation

To use this model for the assessment of data veracity, the expert method is proposed. Each veracity attribute Ai is rated on a 3-point scale as:

and on this basis, the Data Veracity Value (DVV) is calculated as the product of all five attributes A_i.

The result of the assessment is to assign them to one of five classes, namely:

Data, for which Data Veracity Value is very low (DVV = 0), should be fully eliminated and must not be used to create information.

Information uncertainty may be due to objective reasons caused by data imperfections and subjective errors caused by errors in their interpretation. In practice, the uncertainty of information is largely determined by the quality of the available data. Thus, their interpretation, and the elimination of false and contradictory data, as well as the clarification of their definitions in the case of data ambiguity, are of great significance in the process of information acquisition. This crucial as the degree of information uncertainty determines its quality and usefulness for creating new knowledge.

The following characteristics of information quality are proposed (Bukowski, 2019):

In reality, the quality of information depends on the context, in which it is used. Therefore, in the decision-making process, it is essential to evaluate the usefulness of information in the context of its specific purpose. Based on the literature (Kulikowski, 2014), it is proposed to use the term information utility as an equivalent of the usefulness of information, and the following definition of this term (based on Bukowski, 2019):

Information Utility (IU) is the property of the information described by the five-dimensional vector of its attributes:

This definition can be exemplified by the model, which represents the following relation

To use this model to assess the Information Utility, the proposal is to use the following expert method. Each IU attribute Bi is rated by experts on a 3-point scale as:

and on this basis, the Information Utility Value (IUV) is calculated as the product of all five attributes B_i.

The result of the assessment is to assign them to one of five classes, namely:

The imperfection of knowledge is due to the uncertainty of the obtained information and the imperfection of the process of integrating this information with the existing knowledge in the field (Albjoren, and Haldorson, 2002). The general model of the knowledge creation process is shown in Fig. 3. This entire process can be presented as a chain of operations and flows that are subjected to disturbances.

Fig. 3

The raw data stored in the “Data Acquisition” block is sent to the “Data Processing” block and subjected to the preparation process. The processed data flow to next block of “Data Verification”, where they are verified for their veracity (VER) and then sent to the “Data Forming” block for information to be generated using the procedures of Data Mining and Fusion of Data. Raw information flows to the “Information Evaluating” block, where they are evaluated for their utility value (IUV) and classified on this basis. The selected information is used in next block of “Building Patterns” to seek relationships between individual pieces of information and build logical patterns. On this basis, new knowledge is generated in the “New Knowledge Creation” block. The last stage of the chain is to enrich the existing knowledge base with new knowledge and to create the background knowledge for further acquisition of new data.

Aiming to ensure the quality of information, the process of data processing is of utmost importance; therefore, it will be described in more detail. In this step, four main tasks may be differentiated: data cleaning, data integration, data transformation, and data reduction. The description of typical procedures involved in these tasks is shown below (based on Al Shalabi, Shaaban, and Kasasbeh, 2006).

The data cleaning task involves three main operations, which can be supported by numerical methods:

Knowledge extraction relies on denoting the data in the form of a two-dimensional table (matrix) because the column-row structure (e.g. a calculation sheet) is the most convenient. At this stage, the main operations are performed with following procedures:

Data reduction consists of the attribute selection with the expert opinion and experience pertaining to the data under analysis. Those attributes that do not have any influence on the modelled features should be removed from the data set. The general aim of the data reduction techniques is to significantly reduce the amount of data. Data reduction stage includes four operations:

The generalised model of imperfect knowledge management is shown in Fig. 4. The complete process is presented as a sequence of operations and flows that are subjected to different kind of disturbances. The knowledge stored in the “Knowledge Base” block is sent to the “Model Building” block and forms the basis, on which the exact type of a decision model is selected. The information about the type of the chosen model flows to the “Parameter Evaluation” block, in which parameters of the chosen model are determined, and it is then sent to the next block, namely, “Modelling and Simulation”. In this block, suitable investigations are carried out on the chosen parametric model to find the best of all possible decisions, and then, the made decision flows to the “Decision Communication” block. In the next step, the information about the decision flows to the “Decision Realization” block, and after, flows into the “Effects Verification” block. In this block, the results of this decision are tested, conclusions for the future are proposed, and a part of new knowledge in New Knowledge Creation is generated. The last stage of the process is to improve the existing knowledge base with new knowledge and to create the updated initial knowledge for further decision-making.

Fig. 4

Based on the above concept, the proposal is to describe the knowledge processing quality (KPQ) by using the five-dimensional vector of its attributes as follows:

This concept can be illustrated by the model, which represents the following formula:

To use this formula for the assessment of knowledge processing quality (KPQ), the following expert method is proposed.

Each KPQ attribute C_i is rated by experts on a 3-point scale as:

On this basis, the value of the Knowledge Processing Quality (KPQ) is calculated as the product of all attributes C_i.

The result of the assessment is to assign them to one of five KPQ classes, namely:

The last step of the knowledge management process is the assessment of the knowledge maturity level. Maturity models describe how an entity develops through the levels over time until it reaches the fully satisfactory level. Maturity models have the following general properties (Klimko, 2001; Weerd-meester et al., 2003; Schumacher, Erol and Sihn, 2016):

The proposal is to use the concept of maturity to assess knowledge that is available when making decisions based on Big Data. The assessment of knowledge maturity (KM) should be based on three main criteria, namely, the data veracity value, the information utility value and the knowledge processing quality. Methods for determining these criteria are described above. This concept can be illustrated by the model, which represents the following formula:

Each knowledge maturity attribute Di is evaluated on a 5-point scale, and on this basis, the value of the Knowledge Maturity Level (KML) is calculated as the product of all attributes D_i.

The maturity of knowledge can also be assessed on five levels, e.g.:

A very low (VL) maturity level can be identified by a lack of knowledge, while the level L— by weak, imperfect knowledge, the level M — by medium imperfect knowledge, the level H— by strong imperfect knowledge, and the level VH — by complete, perfect knowledge.

The example demonstrates how to apply the method, which is proposed in Section 3, in the logistics decision-making process. The problem concerned the assessment of disruption risk in a new distribution channel, in which the modern smart technologies, based on Big Data flow streams, were provided to monitor the supply continuity. To assess the disruption risk for a given risky scenario, the triplet of attributes of this scenario was proposed. This model, developed by Aven (2015), and modified by Bukowski (2019) can be described using the following formula:

where:

R(S_i) — the risk measure for a risky scenario S_i

C'i — the consequence measure for a risky scenario S_i (e.g. severity of the scenario impact),

P'i — the uncertainty measure for a risky scenario S_i (e.g. likelihood of the scenario occurrence),

K'i — the knowledge measure for a risky scenario S_i (e.g. knowledge maturity level for the scenario and its attributes).

In practice, the first two attributes are considered together and presented in the form of so-called risk matrix (ISO, 2008). An example of such a matrix that was used to assess the level of disruption risk for the most adverse scenarios is shown in Fig. 5. In the analysed case, experts concluded that for the most unfavourable scenario, the level of risk was “moderate”, which consists of “moderate” impact and “possible” likelihood. However, the final decision depends on the assumed risk acceptability criterion, i.e. determining the division of risk levels into acceptable and unacceptable. The criterion for this division, according to the risk model (9), depends on the level of knowledge maturity that is available to the decision-maker.

Fig. 5

The following classification rules have been proposed:

Based on the above calculations, the level of knowledge maturity is insufficient to meet the required condition. To meet it, the levels of both the Data Veracity Value and the Information Utility Value have to increase to the “high” level. This was done by improving the ICT system, which has been achieved by better “Accuracy”, “Traceability” and “Realisation compliance”. As a result of these treatments, the knowledge maturity rank increased to 64, which corresponds to the “high” level. After these changes, the decision was made to accept the planned expansion of the supply network for implementation.

Knowledge about applied sciences, such as economics, engineering, management or social sciences, is based on incomplete theories and indeterministic laws. It applies to open systems, in which the initial conditions are not known exactly, and the induction method is usually used. The data, from which information is obtained, comes from observations or measurements, the accuracy of which is limited and often biased. Therefore, knowledge built on this basis cannot be perfect. Consequently, it is necessary to assess the level of knowledge maturity so that rational decisions can be made under conditions of risk and uncertainty.

The paper presents the historical development of logistics concepts from Logistics 1.0 to 4.0. On this basis, the discussion covered dominant challenges for users of the newest logistic systems as well as the most important features that characterise Logistics 4.0 and distinguish it from previous versions. The most important characteristics of Logistics 4.0 are full integration with Industry 4.0, the implementation of Internet of Things, the use of smart systems and solutions, visibility across the entire supply chain, and Big Data Analytics. The basic condition for its fulfilment is access to advanced and reliable IT systems. These systems are based on intangible assets in the form of Big Data. Therefore, the main part of the work concerns the provision and management of knowledge based on Big Data veracity as a prerequisite for the functioning of systems based on the Logistics 4.0 concept. A new model for the assessment of knowledge maturity as a result of Big Data processing was proposed. This framework allows assessing the usefulness of acquired Big Data to make the correct logistics decisions in a risky environment.

The practical example of using the method presented in the paper allows to formulate the following conclusions: