The Challenge
As markets evolve at a fast pace and information flows freely, companies face increasing pressure to identify and address emerging customer needs as quickly as possible in order to maintain a competitive edge. Traditional Voice of the Customer (VOC) research, while valuable, typically relies on in-depth customer interviews conducted by professional market research, and therefore can often be constrained by time and resources.
Applied Marketing Science (AMS) and BEHR Paint Company, a subsidiary of Masco Corporation, partnered to pilot a groundbreaking VOC study leveraging AMS’s cutting-edge AI tool that overcomes these limitations: a supervised, finetuned large language model (LLM). The goal of this study was to illustrate how AI can deliver a comprehensive list of customer needs, sifting through thousands of existing customer comments and conversations, without the constraints of traditional VOC, to help BEHR refine its marketing, product development, and customer engagement strategies.
What We Did
The LLM, developed in partnership with leading researchers from Northwestern University and MIT Sloan School of Management, was trained using customer feedback from a comprehensive panel of real-world, professional VOC studies. It underwent rigorous refinement through supervised prompt engineering and iterative finetuning via expert feedback, mirroring the learning process of a professional human analyst.
Following successful training and validation, the LLM was applied to the BEHR study focused on uncovering trends and opportunities surrounding color expertise, customer experience, and brand leadership. The study had a particular focus on BEHR’s wood stain product category and served as a benchmark to evaluate the model’s ability to extract explicit and latent needs from large quantities of data.
The model synthesized over 4,000 customer verbatims across diverse publicly available sources such as social media, reviews, and community forums. From there, the LLM extracted detailed customer needs, supplemented by some input from professional analysts for a complete analysis, where they reviewed and organized the list of needs, grouping similar insights together to create a clear, complete list ranked by importance
To ensure an objective assessment of the methodology’s validity, the research team also conducted a rigorous blind test, comparing the output of professional analysts, traditional LLMs (such as ChatGPT), and AMS’s supervised, finetuned LLM. The results showed that AMS’s supervised, finetuned LLM consistently performed on par with professional analysts, accurately identifying the core customer need based on the verbatim, whereas traditional LLMs only focused on the suggested solution and did not accurately identify the core customer need.
The Outcome
This AI-powered approach delivered exceptional results: the model found 189 unique and specific customer needs from thousands of customer comments, exceeding the typical 80 to 120 insights typically gleaned through traditional VOC studies. Notably, the researchers found the model uncovered needs that other methods missed: in one instance, it was able to identify 60 less obvious, more detailed needs that the original study overlooked.
The insights gleaned by the LLM spanned areas such as user experience, competitive positioning, and product performance, revealing both mainstream and niche customer insights. Researchers were also able to explore themes of customer expectations and purchase journeys across BEHR’s diverse market segments. These findings offer an authentic, moment-of-truth look into customer pain points and priorities surrounding the research process, customer service, purchase experience, as well as contributors to brand leadership.
For BEHR, the result was a low-lift, but high-impact, engagement. The organization received comprehensive, ready-to-use research findings to inform marketing and strategic decision-making. The insights helped this trusted name in paint and home improvement stay ahead of evolving customer needs and market trends, strengthening the company’s competitive position. The research didn’t just answer today’s questions; it equipped BEHR with a framework for ongoing decision-making as consumer priorities evolve.
Download the PDF