The partitioning tracer test (PTT) is a promising new characterization method for locating and quantifying subsurface nonaqueous phase liquid (NAPL); however, relatively little work has been completed to quantitatively evaluate potential error sources for this technique. For example, the NAPL saturation calculated from a PTT is based upon the observed retardation of the partitioning tracer and the tracer’s NAPL-water partition coefficient. For most sites, the released NAPL is a waste and/or mixture of solvents or fuels; therefore, its composition may vary significantly both spatially and temporally (especially after remediation efforts). However, tracer partition coefficients are typically determined in batch tests using laboratory-grade NAPL or site NAPL collected at discrete locations and times. Consequently, the partition coefficients used may be inappropriate, yielding an inaccurate estimation of NAPL saturation. For tracers with robust partition coefficients (i.e., partition coefficients that are relatively insensitive to NAPL composition), these estimation errors are minimized. PTT modeling and error propagation are used to evaluate potential error and bias in PTT interpretation related to partition coefficient error and robustness.