AI and Market Research: A Match Made in Heaven?

You’re racing to gather timely consumer feedback to keep your product launch on track without costly delays, but the data is overwhelming. Perhaps you’ve got multiple product iterations, and the testing cycles are getting increasingly expensive—and less efficient. SOMETHING needs to change.

Like many product researchers, you’ve probably heard AI can help. Perhaps more than simply “help,” you’ve been told it can “save the day,” “transform your research,” and allow you to “sit back and relax while those key customer insights roll in.” Sounds tempting!

AI for market research is more than just a tool (so they say); it’s a fully-fledged work partner, something (someone?) you can rely on to get those nuanced insights that you’ll never, ever come to with your paltry human brain. Maybe it’s time to ditch traditional market research techniques and let AI take over.

Before you say “After all… Why not?”, let’s explore what using AI for market research actually means, and how you can make the most of it without losing control of the situation.

What is AI in market research?

AI in market research refers to the use of artificial intelligence to analyze and interpret market data about product potential and consumer behavior, leading to more informed decisions about features/attributes to offer, customer segments to target, and product positioning strategies to employ.

In this sense, AI is much more than just the chatbots (in technical terms, “generative AI”) that took the world by storm in the latter half of 2022. AI-powered market research tools include things like:

• Natural language processing for analyzing open-ended survey responses and customer feedback
• Predictive analytics for forecasting market trends and consumer behavior
• Sentiment analysis for measuring brand perception and customer satisfaction

Tying all of these elements together is an inherent ability to crunch extremely large amounts of data and give a result based on an analytical process that’s more or less a black box. 

Although we know how AI algorithms work in general, we don’t know how a particular model is coming to its conclusions or generating its output. This black box isn’t necessarily bad—after all, there’s no way a human could go through all this data—but it’s definitely something to account for.

Another caveat is data quality. If the data used to train an AI model is incomplete, inaccurate, biased, or irrelevant, the results will be misleading.

Why are market researchers so gung-ho about using AI?

When you’re in charge of getting a real-time pulse on your industry’s market, one unrelenting issue is that there’s so much data and so little time. From review boards to the latest social app, it seems like there are more sources of market data springing up each year that can provide incredible value—IF you can analyze everything.

AI allows us to take all this stuff into account, offering a virtually unlimited source of consumer insights and the ability to segment target customers into hyper-niche categories. For product development teams feeling like their market research is always one step behind competitors, AI can help anticipate consumer trends before they become obvious.

AI can also help sift through unstructured, qualitative data and find patterns in it, even though this data isn’t easy to analyze statistically, perform mathematical operations on, or graph. This means that you can take a lengthy interview video, submit it to an AI market research tool, and get a nuanced summary of trends that the interview may hint at. Now, that’s impressive!

How AI has gradually captivated market research

It may seem as though AI just recently sprung onto the scene, but it’s really something that’s been in development since the 1950s. Early work on neural networks then advanced into machine learning in the 1980s, deep learning in the 2010s, and generative AI using large language models (LLMs) in the 2020s.

AI is also just one more (albeit disruptive) stage in the lengthy evolution of technology in market research. Starting with web analytics in the 1990s, tracking consumer behavior online has now turned into social media monitoring and mass surveys. Before the internet came along, researchers were already using data-driven methods to segment the market and predict what customers would do in the near future.

Nowadays, AI is a powerful tool for the fully tech-enabled product research team. Taking advantage of the latest technologies for market research can support real-time monitoring, competitive intelligence, and scalability.

Practical AI market research applications

Enough of the abstract—what are some concrete ways that AI-based market research can really make a difference? Let’s explore a few examples.

• Natural language processing (NLP). You can use AI to derive consumer insights from unstructured text like social media comments, forum discussions, and customer reviews. Segmenting based on how people use language can also help you target more precise consumer groups. (Are people who use lots of emojis more likely to buy your red wine chocolate ganache bonbons or your “birthday cake” chocolate bar?)
• Sentiment analysis. Get a sense of how people feel about your brand by using a subset of NLP to analyze the emotional tone of the aforementioned unstructured text data. You can get a real-time pulse on customer satisfaction or track brand perception over time.
• GenAI for survey questions. Ask the generative AI chatbots to help you come up with spot-on market research questions (and, in some cases, generate and ask follow-up questions) when you’re surveying people about their preferences for new flavors, fashion designs, and so on.
• Predictive analytics. This subcategory of AI uses statistical algorithms and machine learning techniques to identify patterns and trends in data. It’s a great way to forecast future trends in consumer behavior.

As you can see, there’s a lot that can be done with AI and market research put together. That said, there are some very important reasons why you shouldn’t just leave it all up to AI. Your human brain is still very important!

The “black box” concern: You’ll never know exactly how AI comes to its conclusions

AI models are trained on extremely large datasets. Through the many rounds of training in which the models are guided towards a “desired” or “aligned” output, the algorithms gradually adjust the weight of billions (or even trillions) of parameters, crystallizing the patterns that they perceive in the data they’re learning from. In the case of LLMs, models are essentially a collection of mathematical relationships between words—a representation of the words’ likelihood to show up in the same context and in a certain order.

We humans create the models, but through the learning process, the models themselves arrive at their own internal representation of patterns in the training dataset. This representation—which is so complex that it’s virtually unknowable and impossible to reverse-engineer—is what the model uses to draw conclusions about any future data used as an input.

This is what makes AI a black box. We never really know, down to the individual calculation, why a model outputs what it does. We just know that the result is often (but not always) useful.

The data quality issue: Bad data brings bogus insights

To make sure an AI model is as representative of real life as possible, you need to feed it highly representative data. This is harder than you might think, and biases in the data can be really, really tough to scrub out.

Take, for example, “social listening” using NLP to analyze billions of social media posts and comments. Even if the AI were to perfectly interpret the sentiment and meaning of this text-based data, there’s still a high likelihood that the data is skewed. Not everyone posts about brands on social media, so you’re already filtering your data to hyper-focus on the type of people who do. It’s hard to know what you’re missing when you’re only hearing the loudmouths.

For this reason, and for the black box issue mentioned above, it’s good to be somewhat cautious about using AI for market research. You can use the tools, but don’t forget to seek out human input when crafting nuanced research questions, devising product testing methodologies tailored to specific audiences, interpreting complex data with context that AI may overlook, and validating and refining AI-generated insights with experiential and industry knowledge.

A better way: Balancing AI with human insights

You should choose your AI-powered DIY research tools carefully, and use them for specific needs (social listening, early-stage survey question brainstorming, etc.). While AI can point out patterns and trends you might not have seen otherwise, relying too much on it and ignoring authentic human input can put you in a bind of missing other important things, but not knowing what you’re missing.  

Highlight’s product testing platform lets you take advantage of AI-powered market research tools without ceding control to the “robot overlords.” While the tools help us understand our data better, there’s no substitute for the high-quality input of real consumers testing products firsthand in their natural environments.

Highlight’s insights are authentic, nuanced, and most importantly, they come quickly—so you can get to your next product development stage without delay.

Find the optimal blend of AI and human insights

Just because AI can give you a huge leg up, that doesn’t mean you should let it run the show. You can take advantage of its incredible power to quickly analyze near-unlimited amounts of data that are pertinent to your industry, while leaving the more make-or-break product testing questions up to committed, enthusiastic, human product testers.

Caution around using AI for market research is a good thing—it means you’ll still work with real people when it really matters.