In the market research industry, there are countless vendors touting “AI-powered” solutions promising to revolutionize the way we do our work. In reality, a lot of what’s on the market today is more flash than substance, offering little in the way of time savings let alone insightful research findings.
But it’s not all flash. AI is genuinely revolutionizing the way we do our work—but maybe not in the way you think. For us, AI isn’t a magic button that replaces human expertise. But it is a powerful tool that’s helping us in a number of ways, and freeing up our time to focus on the strategic work only a human can do.
So, let’s pull back the curtain and show you how we are—and are not—using AI in market research.
AI as a Pressure Tester
At Campos, we believe designing a research instrument is a team sport. Whether we’re developing a quantitative survey, a focus group discussion guide, or a mixed-method activity, we follow the same process:
- Someone on our team develops a first draft (which is informed by client input)
- They walk several other team members (usually a group of researchers and strategists) through the draft in a meeting
- Those team members each do a deep review of the research instrument asynchronously and provide feedback in a shared document
- The team members regroup to hash out lingering items, like which questions to add or cut, and how to rework questions to better meet our objectives
The human element is essential to designing a good research instrument, and we believe the collaborative element with other humans is essential too. That’s because the process of thinking, drafting, re-writing, discussing, debating, re-writing, and polishing forces us all to think deeply about the research, what we’re trying to learn and why, and that iterative, collaborative process inarguably produces a better result.
“Many users mistakenly expect AI to lead the creative process,” Geoff Woods writes in his recent bestseller, The AI-Driven Leader. “This is not how you will get value from AI. It will only lead to mediocre work.”
So how does AI generate value in this process? It calls to mind a recent article in The New Yorker about how AI is being used by college students. “Professors have renewed their emphasis on getting students to see the value of process,” Hua Hsu writes. Dan Melzer of University of California, Davis, for example, “has designed workshops that treat writing as a deliberative, iterative process involving drafting, feedback (from peers and also from ChatGPT), and revision.”
And feedback is where we have found AI can be useful when it comes to research instrument development. Once we’ve drafted a survey or discussion guide, AI can be used as a pressure tester of sorts. Before we go to a client, we can ask AI to identify any weaknesses in the instrument, such as topics we haven’t addressed but perhaps should, or language that may need to be revised to ensure the survey meets a widely accessible reading level. AI tools can also provide helpful feedback when it comes to survey flow.
Bringing (More) Qualitative Nuance to Quantitative Research
We almost always incorporate some open-ended questions into surveys to bring some qualitative nuance to our research findings. Often, those responses are critical to building a story with the data and delivering truly meaningful insights to clients.
One way in which we’re using AI now is to make those open-ended questions more engaging, through the use of AI-enabled follow-up questions. For these questions, we can feed our AI tool information about what we want to learn from the question, and what additional follow-up questions might be needed to arrive at the level of response we’re looking for. The tool then uses this information to ask conversational follow-up questions of respondents as they take the survey.
Just as you have to be selective about how many open-ends you ask in a survey (there’s nothing worse than a too-long survey!), you have to be very selective about when you use this tool to avoid survey fatigue.
Catching Bots and Fraudsters with AI
There’s probably nothing more important to a primary market research firm than data integrity. If the consumers we’re engaging with and reporting on aren’t real, well, nothing else matters.
We devote a lot of time and resources to combating fraud. We have very extensive measures in place to flag and remove bots, fraudsters, and people who are just not paying attention while taking our surveys. Our sophistication and attention to detail in this area is something we take great pride in.
We can and do use technology, including AI, as part of this process, but the level of data cleaning we do requires human involvement. Staying on top of all the new ways in which market research fraud is evolving – and evolving our quality measures to keep up with them – requires humans as well.
One interesting, perhaps unexpected use of AI in this arena is using AI to help create new methods for catching bots and fraudsters. With the right prompts, generative AI can help us draft code to help identify behaviors consistent with fraud that merit further human review.
AI, the Analysis Co-Pilot
One of the most exciting ways in which we’re using AI currently is to help us navigate large amounts of unstructured data.
Historically, scaling qualitative research has been challenging and expensive, because there was no way to make the analysis of that data more efficient. Researchers had to read through hours of interview transcripts, hundreds of open-ended responses, and sift through endless photos and videos to tease out themes and, eventually, arrive at key takeaways.
At Campos, we’ve built AI analysis tools that allow us to feed all kinds of unstructured data—text, audio, video, photo—into one place. First, we compile a closed, clearly defined source list, and ensure we have a detailed understanding of every input. Then, we query the tool to make our analysis much more efficient. “What factors are driving satisfaction?” “Which respondents had a bad experience?” “What do customers want to see more of in the future?”
Using AI like this, like a “co-pilot,” doesn’t mean we’re outsourcing analysis to AI. Rather, we’re using it to make it easier to navigate large amounts of unstructured data. We still review the data AI points us toward in response to our queries—as noted, the process of reading/watching something is critical to learning and developing a point of view. But now, we can sift through and organize that data in a more structured way.
All of this frees us up to spend more time thinking through what the most important findings are, what it all means given our objectives, and what our client should do with this research. And we don’t use AI to generate the story we want to tell with the data, to write our reports, or to generate our strategic recommendations. That still requires experience, deep thinking, collaboration, and imagination.
But once we’ve built out a story based on the data and formulated strategic recommendations, we then turn to AI again as a pressure tester. Have I missed anything important? Are there any weaknesses in the reporting that need to be addressed, or opportunities that need to be explored further before we go to the client?
The Future for AI & Marketing Research
We’re incredibly excited about how we can use AI in our work. The use cases above capture where we are in this particular moment in time, and we talk about new use cases almost every day.
While AI will continue to change many things about the way we do our work, it won’t eliminate the need for humans to drive truly strategic market research. (Bad market research–think superficial surveys and data dump reports–is another story.) When you work with Campos, you work with a team that truly gets to know your organization, builds on that knowledge over time, and leverages that knowledge in our work. You work with people who have a broad range of experiences they bring to bear in every project. You work with people who collaborate and push one another to design better methodologies, ask better questions, and produce better deliverables. In short, you work with us, and we believe there’s no replacement for true strategic thinking.
Want to learn more about how we use (and don’t use) AI? Drop us a line here.
