Beyond Sentiment: The Value and Measurement of Consumer Certainty in Language

Abstract

Sentiment analysis has fundamentally changed marketers’ ability to assess consumer opinion. Indeed, the measurement of attitudes via natural language has influenced how marketing is practiced on a day-to-day basis. Yet recent findings suggest that sentiment analysis's current emphasis on measuring valence (i.e., positivity or negativity) can produce incomplete, inaccurate, and even misleading insights. Conceptually, the current work challenges sentiment analysis to move beyond valence. The authors identify the certainty or confidence of consumers’ sentiment as a particularly potent facet to assess. Empirically, they develop a new computational measure of certainty in language—the Certainty Lexicon—and validate its use with sentiment analysis. To construct and validate this measure, the authors use text from 11.6 million people who generated billions of words, millions of online reviews, and hundreds of thousands of entries in an online prediction market. Across social media data sets, in-lab experiments, and online reviews, the authors find that the Certainty Lexicon is more comprehensive, generalizable, and accurate in its measurement compared with other tools. The authors also demonstrate the value of measuring sentiment certainty for marketers: certainty predicted the real-world success of commercials where traditional sentiment analysis did not. The Certainty Lexicon is available at www.CertaintyLexicon.com.