New INFORMS Marketing Science Study Key Takeaways:
- Engaging Ads Don't Guarantee Sales: Many influencer ads fail to convert because they focus more on the influencer's brand than the product itself.
- AI-Powered Metric Predicts Sales More Accurately: The "product engagement score," outperforms traditional metrics such as likes, influencer popularity or ad spend in forecasting ad effectiveness.
- Pre-Launch Testing Saves Time and Money: This algorithm enables brands to assess ad performance before launching, eliminating costly post-production testing.
BALTIMORE, MD, March 6, 2025 – Even though the future of TikTok in the U.S. hangs in the balance, a groundbreaking study in the INFORMS journal Marketing Science introduces a game-changing algorithm that predicts which TikTok ads will drive sales before they even go live.
The study, "Engagement that Sells: Influencer Video Advertising on TikTok," reveals how artificial intelligence can identify which influencer ads will convert viewers into buyers, giving brands a powerful tool to maximize their marketing investments.
"We found that many popular TikTok ads fail to boost sales because influencers often prioritize their personal brand over the products they promote. To solve this, we developed the 'product engagement score,' a computer vision-powered metric far more accurate than traditional methods like video likes or influencer popularity," says Juanjuan Zhang, one of the study's authors and a professor at the Massachusetts Institute of Technology. "This tool answers the critical question for brands: which ads will drive sales?"
Already productized through MIT's Technology Licensing Office, this breakthrough is poised to reshape influencer marketing by enabling brands to optimize ad spending, improve influencer collaborations and reduce costly trial-and-error approaches.
Powered by a cutting-edge computer vision algorithm, the product engagement score predicts sales impact more effectively than video popularity, influencer status or ad budget. By leveraging this tool, brands can determine which ads will resonate with consumers before they are launched, leading to better resource allocation and stronger return on investment.
"Content production is becoming easier and cheaper thanks to AI, but that means evaluation is more important than ever," says Jeremy Yang, one of the authors and an assistant professor at Harvard University. "Brands can now ask themselves, 'I can make a hundred video ads, but which ones will drive sales?' Our method provides that answer."
This research has already begun transforming how companies approach digital advertising. Brands can use the algorithm to predict ad performance, refine influencer partnerships and foster greater transparency between advertisers, content creators and platforms.
"This innovation not only reshapes how brands approach video advertising, but also empowers influencers to design more effective content. As the creator economy continues to grow, tools like the product engagement score could redefine the rules of social media marketing," concludes Yang.
About INFORMS and Marketing Science
INFORMS is the world's largest association for professionals and students in operations research, AI, analytics, data science and related disciplines, serving as a global authority in advancing cutting-edge practices and fostering an interdisciplinary community of innovation. Marketing Science, a peer-reviewed journal published by INFORMS, focuses on quantitative research at the intersection of consumers and firms. INFORMS fosters smarter decision-making and innovation through its journals, conferences and professional resources. Learn more at www.informs.org or @informs.
