The partitioning inter-well tracer test (PITT) is a method to determine average oil saturation between an injector-producer pair. Tracer tests can be used to quantify incremental oil recovery in enhanced oil recovery (EOR) pilots and for reservoir surveillance purposes. This study provides a sensitivity analysis of the inter-well tracer test, with specific focus on reservoir geometry and heterogeneity. The uncertainty of oil saturation determination from tracer production responses is quantified and visualized. Various interpretation methods can be applied: peak arrival time comparison, Residence Time Distribution Analysis (RTDA), extrapolation methods and projection methods. Various sensitivities influence the outcome, accuracy and consistency of these methods. First and foremost, reservoir geometry and heterogeneity have significant impact on the shape of the tracer response curve, and on the accuracy of the subsequent oil saturation estimation. The presence of multiple flow paths can be clearly identified from tracer responses and oil saturation of each flow path can be determined individually by use of extrapolation and projection methods. Thus, potential permeability baffles or barriers can be identified and static reservoir models can be improved by evaluating tracer response data. Further key sensitivities are sampling duration, sampling frequency and measurement errors. An incomplete tracer response can lead to significant loss of accuracy of oil saturation determination by RTDA. A low sampling frequency has severe impact on the accuracy of oil saturation estimation, especially if large measurement errors are present. For timely execution of an EOR project, an early estimation of oil saturation is desirable. In this study, a new and robust analytical projection method is proposed that enables early time estimation of oil saturation based on limited data. The projection method is based on a translation of the non-partitioning tracer response curve to the partitioning tracer curve using a time and amplitude scalar. Robustness of this method is achieved by performing a least squares optimization that takes into account all available data in order to find optimal fitting time and amplitude scalars for tracer data translation. This projection method provides accurate early time oil saturation estimations based on limited partitioning tracer data. Especially if responses are incomplete, contain multiple peaks caused by reservoir heterogeneities, have a low sampling frequency and contain large measurement errors, the least squares projection method provides a most accurate oil saturation estimate, compared to other methods.