An enhanced-oil-recovery (EOR) pilot test has multiple goals, among them to be profitable (if possible), demonstrate oil recovery, verify the properties of the EOR agent in situ, and provide the information needed for scaleup to an economical process. Given the complexity of EOR processes and the inherent uncertainty in the reservoir description, it is a challenge to discern the properties of the EOR agent in situ in the midst of geological uncertainty. We propose a numerical case study to illustrate this challenge: a polymer EOR process designed for a 3D fluvial-deposit water/oil reservoir. The polymer is designed to have a viscosity of 20 cp in situ. We start with 100 realizations of the 3D reservoir to reflect the range of possible geological structures honoring the statistics of the initial geological uncertainties. For a population of reservoirs representing reduced geological uncertainty after 5 years of waterflooding, we select three groups of 10 realizations out of the initial 100, with similar water-breakthrough dates at the four production wells. We then simulate 5 years of polymer injection. We allow that the polymer process might fail in situ and viscosity could be 30% of that intended. We test whether the signals of this difference at injection and production wells would be statistically significant in the midst of geological uncertainty. Specifically, we compare the deviation caused by loss of polymer viscosity with the scatter caused by the geological uncertainty using a 95% confidence interval. Among the signals considered, polymer-breakthrough time, minimum oil cut, and rate of rise in injection pressure with polymer injection provide the most-reliable indications of whether a polymer viscosity was maintained in situ.