We propose a method for measuring the conjectural errors that inexpert consumers make relative to experts in using observable product characteristics as surrogate indicators of a valued unobservable characteristic. Observations on the unobservable characteristic, available to the researcher but not consumers, are used to divide the data into high- and low-quality subsamples. Separate hedonic estimation on the subsamples enables measurement of the relative valuations and conjectures of experts and non-experts with respect to indicators under the assumption that consumers sort across quality grades based on their appraisal expertise. The method is demonstrated using a small sample of SKU-level data on the dietary supplement black cohosh. Our exploratory findings on this sample suggest that, relative to experts, inexpert consumers underestimate the value of most observable characteristics as indicators of black cohosh chemical authenticity; however they overweight therapeutic claims on the product label as a negative indicator of authenticity.