25. Can synthetic data replace real market research?
AI driven market research is creating a big buzz but like every technology, has its pros and cons.
I read many articles about behavioral and data science, and this week’s newsletter is based on the most interesting thing I learned last week.
Startups often lack the time and resources to conduct sufficient market research to fully understand their customers' needs and pain points. Traditional market research can require a large budget and time. Marketers need research that is faster, cheaper and produces credible insights.
So, I was intrigued to read about a marketing startup called Evidanza that enables you to “survey AI copies of your customers to build finance-friendly sales and marketing plans in minutes, not months.”
Not your actual customers. Their AI copies. How does it sound?
Welcome to the world of synthetic market research, where “even the hardest to reach customers are now available on demand”
As a marketer trained in traditional research, my first reaction was to think this was ridiculous, but then I saw Mark Ritson on their panel of advisors, among other big names, so that counts for something.
Their founders, Jon Lombardo and Peter Weinberg (who previously founded LinkedIn B2B Institute), made this astonishing claim on the UncensoredCMO podcast:
“All market research will be synthetic in 5-10 years’ time!”
How does it work?
Evidenza’s AI platform creates digital personas of the target market you are interested in who respond to your research questions. They claim to have generated over $1 million in revenue from several clients. There seems to be a stronger use case for B2B companies, where market research is even harder and costlier than B2C.
But there are many unanswered questions:
Can synthetic data truly replace the insights we gain from real people?
Are the results accurate?
Is there any competitive advantage?
How will the AI reflect changing preferences of the real human being it’s mimicking?
If they claim to make research easy for markets where it’s hard to get data, then where do these synthetic respondents get their ‘data’ from?
Some early investigation shows that such research will have its “biases and a lack of variation in qualitative and quantitative analysis, often veering toward stereotypical answers. The effectiveness of synthetic sampling is highly dependent on the training data used to develop the AI model. When dealing with entirely new or niche topics, the responses may not be reliable.”
I found Peep Laja (of Wynter and CXL) comment on a LinkedIn post on this topic interesting:
At this stage, we need to see how the capabilities of this new AI application develop and can address the concerns. One always feels skeptical of new technology, but eventually, one learns how to get the most out of it. I hope the same for synthetic market research. After all, more customer research can only be a force for good.
If you want to learn more about this topic, this LinkedIn post has a great collection of resources.