Crowdsourced collaborations: Using AI, social media to find potential brand partners

In a recent study published in Information Systems Research, Penn State researchers have developed an AI-powered framework to help businesses identify potential brand collaborations. Traditionally, brand managers rely on time-consuming surveys and screenings to find suitable matchups, but this new method leverages social media chatter and consumer sentiment to uncover promising opportunities.

Using a combination of traditional brand characteristics and social media data, the framework provides a curated list of brand pairs tailored to specific businesses. In a survey of brand managers, the model had nearly a 90% chance of selecting favorable brand pairs, compared to eliminating those deemed unlikely.

The framework operates through a four-step system, beginning with signal fusion, which synthesizes various pieces of information into cohesive insights. The first phase combines brand data with co-mentions on social media to create a comprehensive network of brand connections. The researchers analyzed 450 top business-to-consumer brands and 477,933 Twitter posts mentioning these brands together in 2020.

They filtered out complaint-related posts and used AI-based text classifiers to determine whether consumers viewed brands as complementary or competing.

The researchers then combined these insights with brand characteristics and co-mention frequencies to develop a broader network of brand relationships. The framework also included phase two, which surveyed stakeholders about the model's proposed brand alliances. In this phase, 57 brand managers evaluated 24 brand pairs and selected up to eight promising pairs, five times more often than the eliminated pairs.

Phase three of the framework categorizes proposed collaborations based on factors such as frequency of mention in posts, sentiment, and engagement metrics. The model then plots the proposals on a matrix with four quadrants: premier opportunities, high-potential alliances, hidden niches, and less promising prospects. This stage provides a more systematic view of the opportunities, helping stakeholders prioritize collaborations.

Finally, phase four uses natural language processing and generative AI to generate practical guidance for individual brand collaborations based on insights from social media posts. The generated guidance details which specific marketing elements are dominating consumer discourse, allowing businesses to make informed decisions about potential brand alliances.