This paper presents a new empirical approach for courts to identify anticompetitive consequences of trademarking product features, so-called “trade dress”. Granting a firm rights in its trade dress (imagine a candy company employing a heart-shaped Valentine’s Day box) can diminish competition and harm consumers (consider the high price a heart-box monopolist could charge). Therefore, U.S. courts recognize such trade dress as “aesthetically functional” and refuse to grant a trademark right, even if the trade dress identifies the source of the product to consumers. However, it is practically difficult for courts to discern what kind of trade dress, if trademarked, would diminish competition in specific cases. Despite plentiful doctrinal commentary on this issue, no empirical studies have explored the anticompetitive consequences of trademarking a disputed trade dress. This paper fills this gap by data mining and a human subjects experiment on color trademarks, a subcategory of trade dress. The paper first mines the data of different colored products on the Amazon website, revealing several colors might have larger market shares than others. It further experiments on relations between prices and colors, showing that some colors imply a higher price than other colors. Based on these studies, the paper proposes an empirical approach for litigants and judges to employ in deciding aesthetic functionality in trade dress litigations.