Objectives: It is unclear how nurses adjust the double check during injectable medication administration and guarantee patient safety. We used the Functional Resonance Analysis Method (FRAM) to determine the fit between the double check according to the protocol (work-as-imagined) and clinical practice (work-as-done). We aimed to learn about process variation in order to optimize safety during injectable medication administration. Methods: A qualitative study (February–July 2018) with semi-structured group interviews. An internal medicine and a surgery ward of two Dutch hospitals participated (four wards total). We interviewed nurses about injectable medication administration practices, based on prior observations. A work-as-done model was constructed for each hospital. The work-as-imagined model was based on the Dutch protocol for safe injectable medication administration. Results: A total of 27 nurses were interviewed. In both hospitals, nurses split the double check into a digital and physical check to improve workflow. The digital check was routinely conducted. For the physical check, nurses made their own risk-impact analysis and assessed staffing, familiarity with the medication, severity of side effects, type of medication and administration route and the patients’ medical condition. Based upon these criteria, nurses decided to conduct the physical double check or not. Conclusions: We identified a lack of fit between work-as-imagined and work-as-done. Nurses adjust the double check in practice by assessing the patients’ and wards’ situation. It is unknown whether this variability also causes patient harm. We recommend to reconsider to what extent practice variation is acceptable and safe.