We develop a firm-specific measure of valuation uncertainty from the distribution of valuations predicted by an empirical multiples-based valuation model. The measure is effective in summarizing the information in existing proxies and offers substantial incremental variation. Among many possible applications, we use our measure to test the hypothesis that valuation uncertainty is conducive to valuation mistakes. A value-like long-short strategy is particularly profitable among high valuation uncertainty stocks. Stocks in the short leg earn average returns indistinguishable from the risk-free rate – turning negative following periods of high investor sentiment – and their future earnings disappoint. Insiders trade against the presumed valuation mistakes.