What happens when you get two research experts together for an interview, one a veteran high-level client-side Customer Insights professional, and the other a seasoned expert in CX/UX research, design, training, and experience strategy who is an innovator in AI-poweredVoice of Customer (VoC) analytics for a discussion about Artificial Intelligence (AI)? Magic!
Our conversation covered many topics, focusing on two areas of great promise in marketing research based on AI Large Language Models(LLMs). The first is a sophisticated tool for open-end coding or tagging. The second is the ability to look across multiple data sources for AI analysis.
Large Language Models
Large language models (LLMs) in AI are advanced machine learning models designed to understand and generate human-like text. They are trained on vast amounts of text data, learning patterns, context, and semantics to perform various natural language processing tasks. LLMs can perform a wide range of tasks, including text generation, translation, summarization, question answering, and more. They excel at understanding and generating text that resembles human language.
“What is uniquely powerful about LLMs is their training on an extremely large and diverse collection of textual datasets. This allows them to learn subtle nuances of language more profoundly than previous training approaches, giving them greater power in classic NLP tasks like information extraction, text classification, summarization, and of course, text generation.” — Phil Goddard
Open-End Coding or Tagging and Sentiment Analysis
Open-end coding or tagging is a laborious task where researchers analyze answers to open-ended questions, categorizing them into themes. For example, consider the question, “What is your overall opinion of this logo?” An answer might be, “Concise. Catches the eyes. Easy to read and understand. I like the sun as a symbol of California.” This single response could result in multiple tags such as:
Concise
Identifying all occurrences of these mentions and calculating their frequency can be extremely time-consuming, especially with large datasets. For instance, categorizing responses about types of sandals from 10,000 respondents across ten countries resulted in 133 types of sandals, requiring an extensive effort to analyze.
Another example is sentiment analysis. Traditional methods often bucket sentiment into positive, negative, and neutral categories or generate word clouds, which may raise more questions than provide answers.For example, "California" in a word cloud tells us little about sentiment.
The experts agree that we need AI tools to free up analysts for more critical tasks. Conveniently, our AI innovator is working on tools using LLMs to address this need.
“Customer research is a great example where large amounts of open-ended comments or call logs need to be analyzed regularly. It is much more efficient to watch AI perform this task and then provide guided feedback, rather than doing it manually. My application automates this process using the power of Generative AI and NLP to extract key phrases, classify sentiment, cluster insights by category, and summarize each category. Tools like this will change the way customer research is done—replacing mundane tasks while keeping the researcher in control.” — Phil Goddard
The Ability to Query Data Across a Variety of Data Sources
In marketing research and business, searching across multiple data sources is essential for tasks such as text generation, translation, summarization, and question answering. For instance, querying data from ten waves of an international Tracker Survey or exploring qualitative and quantitative data from a longitudinal research program might be required. In a matrix organization with various divisions and multiple product lines, it's crucial to determine if one division's messaging about a product influences customer perceptions of another product.
The solution for this is Agents. Agents, powered by LLMs, perform specific tasks or interact with users in natural language. These Agents can take various forms, such as chatbots, virtual assistants, recommendation systems, and content generation tools. In the context of LLMs, agents act as intermediaries between users and the underlying AI models, enabling seamless and efficient data querying and interaction.
While Agent scan offload labor-intensive tasks and handle enormous quantities of data, businesses must consider privacy and security. These concerns are significant, especially in highly regulated industries and when dealing with data subject to domestic or international privacy laws.
Buyer Beware
The development of tools and platforms using LLMs in AI resembles a gold rush, resulting in a wide range of tool and platform maturity. If a turnkey, intuitive solution is required, look for a mature, robust tool. Alternatively, working closely with AI developers can tailor the technology to your needs, possibly at a discounted rate.
While LLMs offer remarkable capabilities, they also raise concerns regarding ethical use, bias in training data, potential misuse, and environmental impact.Researchers and developers continue to explore ways to mitigate these challenges while harnessing the potential benefits of LLMs for various applications.
Conclusion
The integration of LLMs in marketing research presents exciting opportunities for efficiency and innovation. However, it is crucial to balance these advancements with careful consideration of ethical, privacy, and security concerns to fully realize their potential.
Kirsty Nunez is the President and Chief ResearchStrategist at Q2 Insights a research and innovation consulting firm with international reach and offices in San Diego. Q2 Insights specializes in many areas of research and predictive analytics, and actively uses AI products to enhance the speed and quality of insights delivery while still leveraging human researcher expertise and experience.
Kathy Townend is the driving force behind Customer Insights at Enlyte, a leading company in the property and casualty industry. With more than 20 years of marketing expertise and a relentless focus on customer experience, Kathy deciphers customers’ needs to fuel business growth. Known for spearheading collaborative transformations, Kathy optimizes outcomes for customers, products, and the company.
Phil Goddard has a background that spans three decades in CX/UX research, design, training and experience strategy. He has enjoyed the full spectrum of roles from researcher, designer, director, managing director, VP, and Principal at leading consultancies. Over the last 10 years Phil has led innovation, design and development of products and services that transform the way Customer Experience is done. Most recently Phil has founded and is Head of Product and Services forValue Architect International, his own startup offering productized managed services for businesses utilizing the latest advances in AI and NaturalLanguage processing applied to Voice of Customer Analytics.
