<p>Business process optimization plays a crucial role in enhancing the production and operation efficiency of enterprises. However, most of the existing methods focus on reducing process operation costs while ignoring the impact of external factors such as supply disruptions. This paper proposes a log analysis-driven BPO method based on graph transformation with the consideration of process flexibility. First, we introduce a simplified well-formed BPMN model to normalize the process structure. Second, graph transformation rules are defined for frequently used BPO operations. Subsequently, a rework task detection algorithm and a process flexibility quantification algorithm are designed to evaluate the well-formed process. Finally, a BPO algorithm consists of optimization strategy generation and optimization strategy execution is devised to reconstruct process via graph transformation. The practical application of the proposed method is illustrated by optimizing a mortgage application process and another process extracted from a real execution log. Compared with other BPO methods, the proposed method improves process flexibility while reducing the time and economic cost of original process, compensating for shortcomings of existing methods in dealing with disruptions.</p> Graphical Abstract <p>This figure illustrates multi-metrics business process optimization. Process log data is divided into core log data and rework log data based on the presence of rework tasks. The Alpha algorithm is adopted to model the business process for the core log data, while redexes are reconstructed via the rework tasks detection algorithm for the rework log data. We construct an objective function covering cost, flexibility and risk to quantify the process. Integer programming is applied to generate feasible process reconfiguration strategies, and genetic algorithms are used to drive graph transformation, so as to improve the performance of the process in diverse scenarios.</p>