Abstract: Like brands, political parties must position themselves against their competitors to capture market share, that is, votes. Likewise, these positions are not static but rather frequently change in response to external events or shifts in public sentiment. Understanding when/how such changes in party positioning occur and how they relate to voter support can provide campaign managers with valuable insights for their positioning or branding strategies. Despite this relevance, marketing research has mainly developed approaches for studying brand positioning and given little attention to political parties. To fill this gap, we introduce an approach for estimating high-frequency time series of party positions from textual data. Our approach leverages recent innovations from natural language processing and market structure analysis (transformer models and dynamic mapping). We apply our approach to thousands of parliamentary speeches in Germany over 24 years and use these estimates to quantify party positioning along various dimensions, such as their temporal consistency or degree of differentiation. Our text-based estimates uncover multiple significant shifts in Germany’s political landscape, such as the con-/divergence of competing parties. We show that these shifts in party positioning can substantially help to predict shifts in voter support, measured via weekly election polls, above and beyond established indicators.

Abstract: Consumers routinely use search engines like Google to satisfy their information needs. These searches provide brands with ample opportunity to connect with consumers. Yet, brands first need to understand their target consumers’ information needs to create relevant content or target them via ads. Yet, identifying consumers’ information needs is challenging, as consumers’ online searches are typically not observable to brands. This article provides a new approach to identifying consumers’ information needs. Our approach tracks consumers’ search behavior via a custom-developed search simulator and uses textual data analysis to infer consumers’ information needs from the recorded search queries. We apply our approach to identify consumers’ information needs when searching for a new credit card. Our analysis reveals eleven broad information needs, including information on benefits, costs and fees, interest rates, specific brands, or traveling. These insights helped our practice partner target consumers who generated 10,000s of additional monthly clicks at below-average ad prices.