As contemporary asset‐pricing models have expanded to include new factors, the empirical literature has carefully studied the time‐series fit of competing model specifications. In contrast, comparison of the roles played by characteristics and factor loadings in explaining cross‐sectional return variation has received less attention. This paper re‐examines the characteristics versus covariance debate. The findings overwhelmingly support the characteristics explanation. Firm size, book‐to‐market, profitability and investment characteristics are important for understanding differences in returns, over and above how a stock loads on common risk factors. Portfolios designed to be a pure play on a given characteristic generate economically significant trading profits.