From performance analysis to efficiency improvement strategy: A data-driven approach

Data has become indispensable in today’s society, likewise in business and industry. Utilising data has become more and more important in the decision-making process. In particular, accessing data in the correct matter such that companies can act upon it, could provide insight on production performance to unfold improvement strategies on the longer run. Accuracy of these planning parameters, such as production yields and required resources, is therefore essential in the era of continuously innovating industries with higher societal and governmental expectations and regulations regarding sustainability. This thesis research presents a framework that could be implemented to gain insight on actual production parameters, actual production performance and to present steps required to improve production efficiency and sustainability. In order to measure the performance of production processes over consecutive periods, a multi-period data envelopment analysis (MDEA) method is adapted and extended by benchmarking each inefficiently produced product in a step-wise fashion. _ese benchmark steps are then combined into a classification tree resulting in a production efficiency improvement strategy for a tactical planning level. _e framework is tested and validated for a manufacturing firm in the fast-moving consumer goods (FMCG) industry, by two distinct data sets, and shows in one overview what measures lead to most production performance improvement. We concluded that the parameters used for planning differed substantially from the parameters resulting from production, resulting in significantly different efficiency scores. We also identified significantly different efficiency scores among the different product groups and production lines. Furthermore, evaluating the production processes on a yearly basis results in nervous behaviour of the efficiency scores. Next, we found that an increase in unit selling price and a decrease in packaging costs are the two main drivers leading to efficiency improvement. The feature importances of the classification trees did not depend on the benchmark levels nor the length of evaluation subperiods. However, the performance of the constructed decision trees did depend on the benchmark levels and length of evaluation subperiods. In general, we concluded that the monthly and quarterly evaluations lead to robust strategies, but this strongly depends on the distribution and fluctuations of the production factors. As the proposed framework is completely data-driven, it must be tested with multiple datasets from, preferably, multiple industries. Finally, the proposed framework is a novelty in the sense that it combines MDEA with benchmarking and machine learning. In this study, we aim to show the potential of employing this framework in practice. To this extent, the scope in this study is limited to the framework performance with regard to dynamic behaviour and robustness but leaves room for many other research applications and topics.