This paper presents a new method for analyzing product differentiation and industry competition using text-based analysis of product descriptions from 50,673 firm 10-K statements filed with the Securities and Exchange Commission. The authors develop two new industry classification systems: "Fixed Industry Classifications" (FIC) and "Text-Based Network Industry Classifications" (TNIC). FIC is analogous to SIC and NAICS classifications and requires transitivity in industry membership. TNIC, on the other hand, allows for more flexible industry classifications that do not require transitivity and can change over time. The authors find that TNIC classifications provide better explanations of managerial discussion of high competition, the specific firms mentioned by managers as competitors, and how advertising and R&D create future product differentiation. They also find that firm R&D and advertising are associated with subsequent differentiation from competitors and increased profitability, consistent with theories of endogenous product differentiation. The authors also find that their new industry classifications offer econometric gains in explaining the cross section of firm characteristics. They conclude that TNIC industries offer substantial improvements over existing methods used in the literature, and that their focus is mainly on horizontal product scope rather than vertical relationships. The authors also find that their results are robust to the treatment of firms that report producing in more than one industry (conglomerate firms). They conclude that these new text-based classifications offer economically large improvements in their ability to explain managerial discussion of high competition, the specific firms mentioned by managers as competitors, and how advertising and R&D create future product differentiation.This paper presents a new method for analyzing product differentiation and industry competition using text-based analysis of product descriptions from 50,673 firm 10-K statements filed with the Securities and Exchange Commission. The authors develop two new industry classification systems: "Fixed Industry Classifications" (FIC) and "Text-Based Network Industry Classifications" (TNIC). FIC is analogous to SIC and NAICS classifications and requires transitivity in industry membership. TNIC, on the other hand, allows for more flexible industry classifications that do not require transitivity and can change over time. The authors find that TNIC classifications provide better explanations of managerial discussion of high competition, the specific firms mentioned by managers as competitors, and how advertising and R&D create future product differentiation. They also find that firm R&D and advertising are associated with subsequent differentiation from competitors and increased profitability, consistent with theories of endogenous product differentiation. The authors also find that their new industry classifications offer econometric gains in explaining the cross section of firm characteristics. They conclude that TNIC industries offer substantial improvements over existing methods used in the literature, and that their focus is mainly on horizontal product scope rather than vertical relationships. The authors also find that their results are robust to the treatment of firms that report producing in more than one industry (conglomerate firms). They conclude that these new text-based classifications offer economically large improvements in their ability to explain managerial discussion of high competition, the specific firms mentioned by managers as competitors, and how advertising and R&D create future product differentiation.