Submit your Friday Insights request for a chance to get a quick (and free) report! Learn More

The Future of AI Driven Market Research

Oct 13, 2021
Side profile made up of tiny bright specs representing artificial intelligence

A sit-down conversation with GroupSolver Co-Founder & CEO, Rasto Ivanic, and Senior Research Team Lead, Adrian Del Bosque

What is Artificial Intelligence?

Over the past few years, the world of market research and surveys has seen the emergence of Artificial Intelligence (AI). AI tools, which are finding their way into almost every industry, hold the potential to change the game entirely.

Artificial intelligence, broadly speaking, is the development of machines and computers that can perform tasks that typically require human intelligence. This includes the ability to understand the human language as it is spoken or written– an instrumental tool for an industry that relies largely on consumer feedback.

In the survey space, artificial intelligence and machine learning offer huge opportunities for automation, enhancing data quality, and uncovering deeper insights. Although some companies have opened their arms to these new possibilities, many are still holding on to outdated market research approaches that have been around for decades.

GroupSolver has challenged this traditional approach from the start– aiming to create a better, more respondent-centric approach to market research and surveys that leverages new technology and opportunities. But how did we get to where we are today, and where are we headed? What will the future of AI-driven market research look like?

In this interview, we hear from two members of the GroupSolver team that have been at the forefront of these industry changes. GroupSolver CEO, Rasto Ivanic, and Senior Research Team Lead, Adrian Del Bosque, share a first-hand look at how AI and other technologies have impacted the industry. They offer their perspective on how these tools can be used, their challenges, and what the future of AI-driven market research will look like– through the lens of GroupSolver and beyond.

First, let’s take a look at where it all began for GroupSolver.

Using Technology to Shorten the Path Between the Question and Answer

When Rasto Ivanic first founded GroupSolver, the question many asked was: why did you start the company? Rasto’s straight-to-the-point answer was, “Because I hate surveys… at least those long-winded boring affairs that make grown-ups keep on asking: ‘Are we there yet?’.”

As the CEO of a market research technology company that lives and breathes surveys, Rasto explains today that, “Our mission in the world is to make surveys more of a conversation… and besides making surveys more pleasant for the respondent by making the conversation quicker and more interactive, I think one of our missions is just to automate the massive number of steps that occur between the business question and the answer in market research. To shorten this path between the question and answer and back,” he adds, “there’s really no reason why market research projects should be lasting two months.”

Next, Rasto explains how AI fits into this mission.

Q: What was the inspiration for using AI to solve problems and answer big research questions at GroupSolver?

Rasto Ivanic: Well, we know that language is very difficult to be handled efficiently by humans when we are asking big open-ended research questions– the why, the how and the what. With survey free-text data, there’s typically a lot of manual reading of answers and trying to put them into context and categorizing those answers in some kind of a theme or real big important answer– which human brains are actually pretty good at doing at a small scale– but if you have thousands of respondents to a survey, the human brain gets lost and you have to look to something a little more efficient.

Rasto explained that AI was one of the tools that naturally emerged as a solution to this problem.

Way back when we started–AI wasn’t a big buzzword or anything– we just knew we had started to make some breakthroughs in those methodologies, and we intuitively knew that this was the way to go.

We started using NLP models before it was sexy to be AI– way before AI was anything people knew; it was just a big scary word. At the time there were a couple of movies with Will Smith that had AI robots going crazy and that’s what AI was. But we were simply trying to figure out the solution, and AI was one of the tools we needed to get there.

Q: How was GroupSolver able to push the boundaries by using AI tools for market research?

Rasto Ivanic: I think our journey to where we are today has been very interesting, because GroupSolver has never been purely about AI– it’s a useful tool but not the only one we use– but we look at AI in a broader sense than just big models that you train on a large set of data. For us, the more important thing than learning from big models and applying them to surveys is that we wanted to be flexible within the survey.

So, whenever we have a brand-new fresh question that nobody has ever trained data on before–because nobody has thought of asking that question–we needed to have a flexible model that is able to grab what respondents are giving us and put those answers into buckets quickly and reliably. What’s unique about us is that we are taking the great work that is done in the data science space and applying it to the survey space.

A Problem Searching for a Solution, Not the Other Way Around

Elaborating on GroupSolver’s unique approach, Rasto explained:

We came from outside the world of market research or even big data. We didn’t come in with a model asking how do we apply this to market research. We were thinking about if someone is a market researcher, and they have this massive universe of data that they’re collecting, how do they organize it?

We are not a solution in search of a problem; We had a problem we wanted to solve which is ‘How do you quickly and reliably organize unstructured language data?’ and AI, broadly speaking, was one of the tools that we gravitated towards.

Q: How has AI technology changed the market research industry since you started the company?

Rasto Ivanic: I wish I could say that AI has taken market research by storm, but I would say that our industry is very conservative in terms of adopting new methodologies that are not based on your traditional statistics because that’s what we have been doing for 50-100 years. I don’t think the industry has completely quite yet opened their arms and said ‘Let’s accept and do something completely crazy different and new’– like what we’re doing.

We have some great customers that understand and love what we do, but the rate of progress and adoption of AI is limited not by technology itself– it is limited by the user. That user is very careful in making sure they can trust what we deliver, and they don’t take it as a matter of course that any survey will have some kind of AI capability. I think AI is still the exception rather than the rule when the industry is doing research– but it’s changing. I think we are maybe one or two big breakthrough features or use cases away from making this into the mainstream.

Q: What are the most important considerations when building AI-powered technology for market research?

Rasto Ivanic: There are so many cool things you can think of putting into a platform when you’re building the technology, but you always have to step back and look at the outcome from a user’s perspective, because you have to produce things that a user can embrace and intuitively understand.

Rasto elaborates that when dealing with complex models and N-dimensional vectors (that might mean nothing to the average user), it can sometimes be challenging to communicate how you got from point A to point B. The answer ultimately lies in the model and statistics, but the product must be communicated in other ways for users who are less familiar with these areas. Rasto notes, “That’s the bridge that we have to cross every time.”

Adrian Del Bosque, Research Team Lead: I think another consideration is to always keep in mind where we are going. If we understand where we’re going then, say there are 100 possible solutions out there, it’s just a matter of us navigating what is going to get us there faster. But there are also a lot of things you have to consider, because at the end of the day if the consumer or if our user doesn’t believe in what we are doing then we have to reconsider why we are even doing it.

So, our decisions are driven by what we think the market needs and what could be disruptive but also goes hand in hand with what we are actually able to do with our capabilities and resources. In the end, the focus has to be the user and the user experience and what they perceive from working with us.

Q: What challenges do companies using AI in market research face?

Rasto Ivanic: A lot of language models are trained on literature, on really well-written text, but we deal with surveys, and if you have seen the results of surveys, people do not write like that. They write all kinds of colloquialisms, abbreviations, emojis, sometimes they just write profanity, and sometimes they write nothing. So, one of our challenges is adopting models so that they can actually work on the dirty data.

Image of code on a computer machine learning

Q: Does AI evolve as language evolves? What impact does that have on the industry?

Adrian Del Bosque: It is a fact that the space we’re in– the world of surveys– will evolve over time. 5 years ago, it was different and 5 years from now it’s going to be different, so naturally, the machine learning that we develop has to also evolve with however the market and the data evolve.

Referring to natural language in particular– we’ve seen this in the past year and a half– our models didn’t have the word Covid-19 in any observation that was ever used to train them, so at the beginning when we ran a study about the pandemic and people’s concerns, the model understood that there was something new but it didn’t know what it was. Now, a year and a half later, these models understand that “Covid-19” is a disease and are able to associate a bunch of other concepts and ideas that are related to that.

So, the simple answer is yes it will definitely have to evolve. Now, how it does? That is a completely different ball game– that’s the meat of the problem.

Q: How has the approach to using machine learning for survey data changed?

Adrian Del Bosque: The traditional training approach was ‘I’m going to tell the model everything that I know. I’m going to teach it and hopefully expect it to learn everything I tell it.’ Now, some new research is trying a slightly different approach where they’re saying: ‘I’m only going to teach it 5% of the knowledge that I know, and I want the model or the technology to learn the other 95% by itself and then, we’ll see where that gets us.’

In simple terms, I think the approach is a lot less hands-on in terms of actually teaching the model how to behave. It’s much more efficient in terms of less data, it’s less costly in terms of computing power, and the timeline becomes shorter. This should also allow the models to evolve more naturally– so I think that’s where we’re going from a general standpoint with machine learning.

The New Age of Online Surveying is Upon Us

Hands holding an iPad with a city lights and technology in the background

Q: What will the future of market research and online surveys look like?

Rasto Ivanic: I think where we are going– and AI is a great tool along with just the basic statistics and computational power which enables all of that– is that we are moving more into listening to the customers rather than speaking at the customer or the respondent.

I think the future looks like shorter surveys, much more open-ended, much more conversational– and conversational being a tool to enable us to probe seamlessly into the questions in real-time. So, where AI comes in is that it can power some of those chat features, it allows us to direct answers and questions in such a way that we’ll seamlessly get as much out of that respondents’ willingness to give you data as we can without being intrusive. So that also changes the ratio of how much I’m speaking at you in a survey and you as a respondent back to me.

I think that’s the future, and I think without these more sophisticated models and smart designs, we would continue asking essentially paper surveys in an online format like what we have seen in the industry so far.

~

GroupSolver CEO and Research Team Lead Biographies


You might enjoy these too

GroupSolver blog about Black Lives Matter
Thought Leadership
Black Lives Matter.
Read story
GroupSolver vaccine interview part 2
Newsroom
A vaccination conversation Part II: discussing health literacy with Dr. Eileen Fry Bowers
Read story
GroupSolver Rasto Ivanic thought leadership blog about startups
Thought Leadership
Lessons from the frontier
Read story

A better way to smarter insights starts now

Request a free demo. Someone from our dedicated team will get right back to you.

Request Demo