image of connected people and their activities online indicating customer experience


Revolutionizing Customer Insights: The Role of AI in Accelerating VOC

Revolutionizing Customer Insights: The Role of AI in Accelerating VOC

A deep, beyond-the-obvious, understanding of customer needs is an essential component of creating competitive advantage, successful new products, and an unparalleled customer experience. With the rapid growth of Generative AI, the insights community is in a transformative period. This period raises thought-provoking questions about how AI and LLMs can optimize insights gathering, analysis, and customer understanding.

What is the Relationship Between Customer Needs, AI, and VOC?

Customer needs are the benefits sought by the customer when they use your product, interact with your brand, or conduct business with your company. The goal of VOC market research is to systematically understand in-depth, beyond the obvious, important, unmet customer needs – and to let those needs drive product development and customer experience strategy. In other words, VOC allows you to focus on solving the customer problems worth solving.

Traditionally, customer needs are best collected through in-depth customer interviews. However, many companies may be sitting on a gold mine of existing, already available, customer insights: this data comes from over 3 billion active social media users, over 100,000 online forums, and over 265 billion customer service calls placed every year for questions, comments, and concerns about products.

The average adult spends more than two hours a day on social media and has 11 customer service interactions per year (that’s about one per month). Over 60 million people have posted reviews online about products and services that they use regularly.

The question becomes: how can AI help with VOC insights? Researchers from Applied Marketing Science (AMS) explored and continue to explore this question in partnership with researchers from MIT and Northwestern.

Can Machine Learning Replace Traditional VOC for Consumer Products?

Early-stage research indicates yes. In their first experiment, the team tested the hypothesis that online reviews could serve as a source for consumer VOC insight. In other words, the online product reviews could produce the same results as traditional, one-on-one interviews? In this experiment, the team mined customer verbatims from over 200,000 Amazon reviews for personal oral care products.

Here is an example of a typical Amazon review. The customer says, “I replaced an old brush with a new one, but the description doesn’t say that this model no longer has a 30-second timer. The brush shuts off after two minutes, but the 30-second timer is missing. I would not have purchased this product if I had known.” By looking at this review, VOC methods, applied by human analysts, could formulate a customer need, such as: “I know the right amount of time to spend on each step of my oral care routine.”

In this process, the machine acted as a sophisticated data reduction tool, identifying the sentences in each review that were and were not informative, e.g., those that contained customer needs. The machine synthesized a massive amount of data into relevant content to be reviewed by a team of research analysts.

This experiment found that online product reviews contained the same–and in some cases even more–information than researchers identified in traditional interviews. Moreover, this AI method identified customer needs quickly and effectively from an extremely large body of textual content.
It is important to mention that the algorithm surfaced both frequently and infrequently mentioned needs. For example, many customer needs centered on the desire for whiter teeth and fresher breath. However, the AI algorithm can identify lesser-mentioned needs such as having oral care products that match one’s bathroom decor.

Using AI to process volumes of readily available text presents unique and exciting opportunities for those collecting insights to inform innovation and customer experience.

Machine Learning’s Application in VOC for Business-to-Business Markets

Where sufficient text data is available, machine learning can fuel VOC insights in complex B2B markets. By way of example, AMS successfully utilized its machine learning algorithm to uncover systematic and comprehensive insights in the snowplow and snow spreader category. In this case, the objective was to identify customer needs to inform product development and marketing. With the help of machine learning, the research team identified 117 distinct insights through this process.

The client for this engagement was a veteran of the snowplow industry. At the onset of the engagement, he was admittedly skeptical that the methodology would yield insights that he hadn’t heard before.

Through the use of machine learning combined with human analysts, AMS was able to unearth several new and exciting insights. Notably, we pinpointed a critical pain point related to visibility –the exact moment when the snowplow operator needed to have the best visibility was actually the point where they had the worst visibility.

Specifically, the customer said: “The reason I don’t angle it [referring to the plow] is that I hate to have one side sticking out so far, and it will block the headlights at the same angle.” The customer need formulated by human analysts was: “Being able to maintain full visibility at all times, even while my plow is angled.”

Other novel insights uncovered included the need for a sidewalk plow that is easy to turn around a corner; being assured the user has de-icing material for different types of roads (gravel, paving, etc.); and sand that will not strip away paint when applied (e.g., over parking lines).

In this B2B example, the algorithm was able to identify new insights that were critically important to the industry.

The Future of AI-Powered VOC

With the recent popularity of large language learning models (LLMs) like ChatGPT, The AMS research team, in partnership with academics at Northwestern and MIT, is now investigating whether LLMs can formulate abstract, conceptual customer needs for customer verbatims. Previously, this is a task that only human analysts have been able to accomplish.

Preliminary results from this new research are promising. We see early evidence that LLMs can formulate customer needs; indistinguishable from human-formulated needs. Notably, these needs are sufficiently detailed for product development. This specificity is critical for innovation.

By harnessing the power of AI and LLMs, researchers can quickly and effectively synthesize massive amounts of qualitative data to unearth the nuanced customer needs necessary for effective innovation. While machine learning algorithms will never completely replace insights and product development professionals, they can certainly improve efficiency and effectiveness, allowing the team to focus on the most value-added portions of the insights and product development process.

AMS prides itself on being at the forefront of human and AI collaboration for VOC insights. We are excited to be so instrumental in these exciting changes that impact the way we collect, analyze, and understand the Voice of the Customer insights.

Author Profile
Kristyn Corrigan
Kristyn Corrigan

Principal 250-6326
Let's stay in touch

Subscribe to receive the latest market research insights from AMS