The productivity measures used so far in the innovation literature reflect firm performance, which confounds profitability and technical efficiency. Nevertheless, the distinction between the two is crucial. We propose a decomposition model to improve our understanding about how firms profit from innovation. We apply this model using a unique dataset with product level data on prices and quantities next to information on innovation activities and traditional firm performance measures. We find that firms benefit differently from demand and cost driven innovation. Demand innovators charge high prices, have high marginal costs and low technical efficiency; the opposite is true for cost innovators. We show how these findings reconcile the heterogeneity in estimated returns from innovation in earlier empirical studies.