Organizations need to continuously improve their processes to stay competitive and relevant. Process mining provides data-driven insights into business processes, typically conveyed to stakeholders using BPMN models. A common use case is to compare such BPMN models to real-life execution records of the process, called event logs. Since different process scenarios appear with different likelihoods, event logs are inherently stochastic. Adding this stochastic perspective to BPMN can yield more accurate conformance checking results. Furthermore, existing stochastic conformance checking techniques consider totally ordered traces, which does not align well with BPMN’s explicit parallel gateways semantics. In this paper, we introduce the first stochastic conformance checking method for BPMN, a formalization of Stochastic BPMN (SBPMN), and a backward-compatible extension of the BPMN 2.0 specification. Our method accounts for partially ordered behavior and behavioral errors like livelocks and deadlocks in SBPMN models. We implemented our approach in ProM and evaluated it on three real-life event logs. Our experimental results demonstrate the computational feasibility of our method and highlight the importance of incorporating the stochastic perspective while preserving concurrency in conformance checking.