Bank filtration is a cost-effective and sustainable method of improving surface water quality for drinking water production. During aquifer transit, natural biodegradation and physiochemical filtration improve the quality of the raw water by removing sediments, pollutants, and pathogens. Strict regulations prohibit the use of substances that can be used to estimate aquifer residence times to define water protection areas for bank filtration. In this study, we present a novel measurement and modeling framework for deriving mean aquifer residence times for bank filtrate using the natural tracer radon-222. The method is intended for application in the drinking water sector, where extraction wells are screened over the entire aquifer and pumps are operated at high production rates. Mean aquifer residence times are estimated using composite residence time distributions that account for flow path mixing and non-uniform residence times with multiple components including bank filtrate, shallow groundwater, and deep groundwater. The mathematical framework is demonstrated for a drinking water production facility. Radon activities for the six monitored extraction wells ranged between 4,400 and 8,400 Bq/m³. Estimated mean aquifer residence times for the wells range from < 5 days to 110 days and strongly depend on i) the type of residence time distribution model (exponential, gamma or piston flow), ii) the mixing ratio between bank filtrate and local groundwater, and iii) the heterogeneity in the groundwater endmember. By accounting for mixing processes, we can show that radon can be used beyond the "5-fold half-life" (~20 days) commonly described in the literature as the upper limit for age dating purposes for radon. This method provides a simple and cost-efficient way to quantify residence times of bank filtrate on a regular basis without any addition of external substances to the aquifer.