AI cuts down the time it takes for researchers to develop survey questions, collect responses and analyze data, meaning clients get answers sooner.Getty Images
Addy Graves had been working in market research for over a decade when ChatGPT launched in 2022. While exploring the platform’s capabilities, she wondered: How could AI improve how market research is done?
Traditionally, studies would take at least two months, says Ms. Graves, who previously held positions at Critical Mass and Ipsos Reid. That included time to plan the study, develop a survey platform and questions, gather results, analyze them and create a report.
“It really wasn’t until AI that we were able to automate those workflows,” she says.
In 2023, Ms. Graves co-founded Cashew Research as an end-to-end, AI-enabled digital market research company. AI helps Cashew’s researchers develop online survey questions, collect responses from real humans, analyze data and generate reports. “We’re cutting down the time frame by at least five to six weeks,” Ms. Graves says. “We can turn a study around in a couple of days to a week.”
Companies like Cashew, alongside major players in market research such as Leger and Nielsen, have recently embraced AI to evolve an industry that has largely remained the same for decades.
In addition to speedier results, AI also reduces the human capital required to complete studies. For example, instead of dozens of researchers working on a study, one Cashew researcher can manage 20 customers simultaneously. This allows the company to price their services significantly lower – 60 to 80 per cent less than what traditional market research studies cost.
“Something that was super sophisticated and super expensive is now opened up to mid-market brands with smaller budgets,” Ms. Graves says. “It allows these brands to grow faster.”
One of Cashew’s early customers was Joni, a plant-based period care company based in Victoria. Co-founder and CEO Linda Biggs sought out Cashew’s services as a lower-cost option and the results pushed the company to launch a new product: period underwear, which was something they hadn’t previously considered.
“It shifted our product development approach to what the consumer felt they needed,” says Ms. Biggs. She adds that she felt reassured that the data Cashew provided was reliable since it was based on responses from actual humans.
Ms. Graves says that is always the case for the data they provide to clients, at least for now.
But other major players in the industry, like Montreal-based Leger Marketing, have embraced what’s called ‘synthetic data.’ Researchers ask survey questions of AI-enabled ‘personas’ that can mimic the responses of real humans with surprising accuracy.
“We evaluated that the quality of the answers is between 85 and 90 per cent, which is very high and much more than enough to make a marketing or business decision,” says Sarah Mottet, vice-president of transformation and AI at Leger.
The accessibility of synthetic data makes it easy for companies to gain insights quickly and at a lower cost. But Leger is transparent about its limitations.
“We would not recommend a client use this approach for a big, high-stakes decision, like a brand repositioning,” Ms. Mottet says. Clients have, however, successfully used synthetic data for smaller decisions like testing a Facebook post or a new advertising campaign. For example, a chip company could use synthetic data to whittle 30 new flavour ideas down to the best five before developing and testing them with a human panel.
Another way that Leger uses AI in its work is its ‘AIMI’ platform, which stands for AI-moderated interviews. Here, there are real respondents, but an AI interviewer asks questions (as text on a screen – they’re not using human-looking AI interviewers over video just yet) and the moderator processes responses in real time to ensure they’ve been adequately answered.
“If you didn’t answer the question, they will rephrase it and ask it back,” Ms. Mottet says. “[The AI] understands the context and they are able to ask follow-up questions.”
Guneet Nagpal, an assistant professor of marketing at Western University’s Ivey Business School, formerly worked in market research at NielsenIQ. Dr. Nagpal says she is excited about AI-enabled advancements in market research but currently feels cautious about synthetic data.
“Since it’s trained on historical data, its results might not always be up to date,” she says. “The response that we get [from AI] is based on what has been fed into it in the past. Historical data doesn’t reflect what we are thinking in the future. Humans are also evolving.”
Dr. Nagpal says that like any organization adopting AI these days, market research companies should have verification strategies in place. To prevent responses from bots, companies can use behavioural pattern analysis (checking for unusually fast completion times, repetitive answer structures or inconsistent responses), device and IP address checks, CAPTCHA systems and direct questions that filter out inattentive users. “Each organization trying to adopt AI should have this method and structure built in … to minimize errors,” she says.
At Cashew, researchers go over a checklist to ensure that the data they are getting is from human sources, says Ms. Graves. For example, they have algorithms running to ensure answers aren’t coming in too quickly so they can know a bot isn’t replying.
Looking ahead to the future, Dr. Nagpal says she could see AI market research tools like synthetic data improving in reliability and AI-generated responses becoming “closer to being human.”
Tools could also become more sophisticated, she says, like facial recognition software that analyzes a respondent’s expressions for a more detailed reaction.
“I wouldn’t be surprised if it’s happening already because the tools are available,” Dr. Nagpal says. “There will be more and more usage of AI in every industry, in every part of our lives – and in market research, as well.”