Regardless of the industry that you're in, it is harder than ever to find truly new customer insights. Research budgets are smaller, the low-hanging fruit has already been picked so you need to dig deeper to find new insights, and traditional research can be expensive and time-consuming.
But artificial intelligence, or machine learning, is changing the game, according to John Mitchell, president and managing principal at Applied Marketing Science, a Waltham, MA-based research and marketing firm that helps its clients better understand and incorporate the voice of the customer into product development.
Between social media, online customer reviews, and customer service calls, companies already have billions of user-generated content (UGC).
"Consumers are freely volunteering insights about products and services at the moment of truth," Mitchell told BIOMEDevice Boston attendees on Tuesday.
The problem is that sifting through all of that to find valuable product development insights is simply too much for one human reader to process on their own, Mitchell said.
Citing a recently published paper written by Artem Timoshenko and John Hauser, Mitchell shared a process Applied Marketing Science uses to show how machine learning can, in fact, help companies across all industries find new customer insights.
"Artificial intelligence methods are really changing the game. Machine learning is helping us find the insights that matter, quickly," Mitchell said.
He also shared a case study demonstrating how this process would work in the glucose monitoring market.
"I've got to tell you, lots and lots of people are talking about diabetes and are talking about all of the different equipment and devices that are used to treat diabetes. People are talking about this online all the time," Mitchell said. "We collected over 1,000 posts from forums and blogs about diabetes and about living with diabetes. We pulled over 1,000 reviews from e-commerce sites that sell supplies related to diabetes. And we pulled over 500 reviews from social media websites."
Then Mitchell's team went through the five-step process that Timoshenko and Hauser describe in their paper.
From that data, Mitchell's team parsed out about 8,500 sentences, used a subsample of those UGC sentences as an input to train the algorithm, ran the machine, the machine gave the team an output of customer ideas organized into different thematic groupings (such as accuracy, connectivity, and interface and features) to help identify user needs from a product development standpoint.
Then, to figure out how important these insights are, the team conducted a traditional survey of 297 adults with diabetes who have owned or used a glucose monitor for at least three months (and use the device at least once a week). The respondents were asked to rank a list of 22 secondary needs on both importance and performance in reference to their own glucose monitors. Brands asked about included devices made by Abbott, Bayer, Dexcom, Roche, and the OneTouch brand formerly sold by Johnson & Johnson's LifeScan unit, which J&J divested last year.
After analyzing the survey results, the team created a market opportunity map to prioritize each of those 22 needs both in terms of how important they are to the user and how satisfied the user is with the market's current ability to meet those needs.
"So a need that consumers tell us is very important but they're also really satisfied with what the market's providing, we might consider that a minimum requirement and not necessarily something we need to invest a lot of time and money into making better because customers are already pretty happy," Mitchell said. "However, a need that is important but customers tell us they're not satisfied with [from the current market] is something we might consider a focus area, something that we really should try to solve in the next generation product."
For example, one important need that came out of the research is "a simple, easy-to-read interface' but patients also said that was something they are pretty satisfied with from currently available products. On the other hand, the research also identified an unmet need for external conditions to not affect the accuracy of the glucose reading.
"That's something that patients are saying is really important, but they're not really satisfied with what the market is currently giving them," Mitchell said.
Similarly, a need that customers say is of low importance but is also low in satisfaction might be considered a hidden opportunity, he said, and a way to potentially disrupt the market because there's a chance that customers ranked it that way because they just assume nobody can do it any better.
For example, the need for glucose monitoring information to seamlessly synch to the user's cellphone may be a hidden opportunity for innovation.
"That's something that patients may be telling us right now 'it doesn't really matter all that much' but if they don't have any experience with how well that could work, they may be just giving us a pass on that."