By using a proprietary real–time daily online sales data collected in China from 10–billion consumer accounts, this paper ffnds that the ffrm–level daily online sales growth (DOSG) can positively predict future one–day to more than three–month cumulative stock returns in the cross section, implying a growth premium in contrast to Lakonishok, Shleifer, and Vishny (1994). A spread portfolio that is long on stocks with high DOSG and short on stocks with low DOSG delivers an abnormal return of around 30 basis points per week. DOSG derives its short–run (e.g., weekly) predictability from investor sentiments, tilting to a behavioral explanation. However, it derives its medium to long–run (e.g., three–month) predictability from fundamentals, voting for a rational explanation. Our further evidence indicates that stocks with high DOSG experience more intensive information acquisition from retail investors and less severe crash risk, implying online sales as a channel for retail investors to get access to daily real–time ffrm fundamentals.