Data-driven Decision Support for Manufacturing SMEs

Today, more and more data is automatically collected through the growing amount of IT-systems in manufacturing SMEs. The IT-landscapce is typically heterogeneous and data is mostly stored electronically in different databases and file servers, but is neither smoothly integrated nor used for additional decision support. To address this backdrop, this paper offers a procedure model, how companies could deal with these masses of data and what benefits they might derive from it. It is intended to contribute to the use of data-driven decision support in SMEs, especially in project management of manfucturing SMEs. A four-step procedure model is introduced to support companies to recognize the advantages of data analytics, visual analytics as well as big data to better control their projects and is evaluated during the research project SimCast

Vorgehensmodell zur Datengetriebenen Projektplanung in KMU - Wibke Kusturica
Vorgehensmodell zur Datengetriebenen Projektplanung in KMU - Wibke Kusturica

The amount of stored data in manufacturing companies is growing constantly over the last couple if years (Deloitte, 2014). Data comes from very different origins and is currently mostly stored, but not evaluated for the continuous improvement of the processes, especially in project management. Here, companies mostly rely on the expert knowledge of the responsible project managers. How could companies deal with the available data sets and what benefits they might derive from it? The following article is intended to contribute to the use of data-driven decision support in small and medium sized enterprises (SMEs), because currently strategic, tactical and operational decisions are made mostly manually with the single help of gained expert experience. Particularly in the area of planning and control, it will become increasingly important to analyze historical data in order to be able to derive better decisions (Becker, M., 2011 Wikum, E., o.J.). This is particularly difficult in SMEs of the one-of-a-kind production industry and small-batch manufacturing; they are not able to reuse the existing data 1:1 due to the constantly new, non-repetitive project character of their work. It is therefore necessary to provide companies a guideline that enables them to carry out data analytics and thus generate benefits.

Phase 1: Pre-project Stage

Based on the pre-project, called Quickcheck, it is necessary to define the expectations of all participants and to provide clarification about the next steps. Often SMEs in particular see this time as a lost time, since the costs incurred during this phase are not valued in a value-added way. The companies ignore that "the later changes to a product (or project) become necessary, the more expensive it becomes.

The purpose of an interview is to determine the current availability and utilization of data within the company. The focus is on recording the data available at the time of the actual analysis, the availability of this data, data provision, data processing and the recording of the IT infrastructure. Furthermore, the interview will determine how the company is currently planning project durations and especially their logistical processes and to what extent historical project data has been used to estimate future project durations. This determination is supplemented by a comparison of the data requirements with the company data. This phase includes the company's desired requirements for future data utilization, processing and visualization.

Phase 2: Process Analysis and Mapping

As part of the implementation of the process model, the value-adding processes (production and logistics) should be presented and documented (Pfeffer, M., 2014). Afterwards, a detailed analysis of the processes ascertained takes place according to the top-down approach as well as examination and acceptance of the collected processes. After the process entry follow the documentation by BPMN, each process is examined with regard to data generation, sources and infrastructure.

The central characteristic of a reference model is its intended or actual reuse. It represents a pattern that can be considered as an ideal model for the class of technical or business facts to be modeled (Koch, S., 2011 Becker, J., Delfmann, P., 2007). For this purpose, value-adding processes are summarized and presented in categories. For each production area, a reference model is prepared in preparation for implementing the procedure model. In the beginning, cause-and-effect relationships are recorded and presented in Ishikawa diagrams. The processes are recorded with regard to the estimation parameters for each production process via the production areas assembly, disassembly and production. All estimation parameters are used to schedule the respective production area. The categorization makes it possible to compare individual production areas with each other and to identify possible relationships that affect the process duration already at this point. After the process has been fully recorded, the comparison of the recorded processes with the reference process and the documentation of the value-adding processes are prepared for the creation of the specification sheet required in step three.

Phase 3: Finding Solutions

The design of the solution approaches requires a comprehensive way of viewing and processing. On the basis of the specifications drawn up in the first phase, the consultant examines the customer's requirements and documents solution approaches and detailed requirements. This defines the target concept.

In view of the investigation objective presented at the outset, usable data are filtered out of the stock of the data defined in the requirements. The usable data includes all data which directly or indirectly has an influence on the process duration and thus on the entire project duration. As a result, filtered data sources with corresponding data records are available for further processing for all target processes. Particular attention is paid to the completeness of the data, the data scope and the number of data records considered, which must reach a predefined minimum number. If the quality or quantity of the data is not sufficient, it is necessary to measure relevant data that has not yet been considered, or to record them from the start.

The search for alternatives essentially refers to the selection of other suitable data. The search for such a solution should not be underestimated, despite the high acceptance of the previously preferred solution, since decisive points may have been unconsciously neglected in the previous consideration of finding a solution.

In the following, basic preparations for the implementation and principal procedures of statistical data analysis, which will be applied later, and the associated data mining (Queckbörner, S., o. J. Freistas, A. A., 2013), embedded in the KDD (Fayyad, U., Piatetsky-Shapiro, G., Smyth, P., 1996) process, are presented.

Phase 4: Implementation

After the data analysis has been completed and possible influencing parameters for the duration of value-adding processes have been identified, a mathematical concept will be developed and derived from the findings of (calculation) rules. A solution is provided, which should quickly provide the company with influencing parameters and thus key figures for the duration of value-adding processes in the company. The company is provided with a kind of dashboard that provides important results based on historical data, important parameters for influencing the duration of processes in the case of SMEs, and automatically derived recommendations.

Discussion and contribution

The implementation of the procedure model as described in the previous sections serves to prepare companies for the application of available data in order to extend their knowledge and improve their project management tools; in best-case, they might rely on a implemented demonstrator of a plug-in, that will be implemented during the next phase of the mentioned research project.

Literature

Wikum, E. (o.J.): Making Analytics work trough practical project management. Online: http://analytics-magazine.org/making-analytics-work-through-practical-project-management/

Deloitte und Touche GmbH Wirtschaftsprüfungsgesellschaft (2014): Data Analytics im Mittelstand - Die der Entscheidungsfindung, Aus der Studienserie “Erfolgsfaktoren im Mittelstand“, o.O, 2014

Emmanouilidis, C., Taisch, M., Kiritsis, D. (2012): Advances in Production Managemnt Systems – Competitive Manufacturing for Innovative Products and Services, IFIP WG 5.7 International Conference, APMS 2012 Rhodes Greece, September 24-26, 2012, Revised Selected Papers, Part I, Springer Heidelberg, 2012

Fayyad, U., Piatetsky-Shapiro, G., Smyth, P. (1996): From Data Mining to Knowledge Discovery in Databases, American Association for Artificial Intelligence, in: Al Magazine, Vol. 17, No. 3, S. 41, 1996

Gudehus, T. (1993): Logistik – Grundlagen, Strategien, Anwendungen, 3., neu bearbeitete Auflage, Springer-Verlag Berlin Heidelberg 2004 und 2005, in: Baumgarten, H., Prozessanalyse logistischer Abläufe, Fabrik 43, Jahrgang 6, 1993

Koch, S. (2011): Einführung in das Management von Geschäftsprozessen – Six Sigma, Kaizen und TQM, 1. Auflage, Springer-Verlag Berlin Heidelberg, 2011

Pfeffer, M. (2014): Bewertung von Wertströmen, Kosten-Nutzen-Betrachtung von Optimierungsszenarien, Springer Fachmedien, Wiesbaden, 2014

Queckbörner, S. (o. J.): Was ist Data Mining?, Technische Universität Kaiserslautern, o. J.