Customer input for product marketing
Chobani, which is well-known for its yogurt products, has now expanded into other products such as oat milk and coffee creamer. As a result, it moved beyond its initial sales locations with retailers and convenience markets to a variety of other outlets. With the broadening of its markets, Chobani wanted to obtain and analyze feedback from its growing base of customers to support decisions in marketing and product innovation. It selected Canvs MRX from Canvs AI to support this initiative. Canvs MRX analyzes text responses from a variety of sources, including surveys, product reviews, and customer comments.
Chobani carried out numerous analyses to better understand input provided by its customers. The Chobani insights team partnered with a retailer to look at survey feedback that included both open-ended and closed-ended questions about how Chobani displays its product. Canvs produced five actionable “buckets” of suggestions: more variety of flavors, better signage, improved shelf organization, greater attention to expired and out-of-stock issues, and more promotions. The use of Canvs reduced the time that it would have taken to analyze the data manually by a factor of 10.
In another survey, Chobani sought insights about what cus- tomers were looking for in their “dream smoothie” products. The resulting data offered quantitative information on how often each ingredient was mentioned, listings of what fruits were disliked, and suggestions, such as requests for both dairy and nondairy versions.
The combination of the analyses of open-ended and closed-ended questions allows Chobani to discern whether the comments are being made by individuals within certain demographics, such as age, gender, and geographic location. Canvs analyzes the text and then runs the statistics across these categories. This produces information that would not otherwise be possible through text analytics alone, including unlocking emotional responses.
Canvs AI holds patents related to measurement of emotions as expressed in language. “Our breakthrough was to not just analyze positive and negative sentiments,” said Trip Kucera, EVP for marketing at Canvs AI, “but to detect a variety of core emotions.” The underlying AI and language model was trained on public social media data. “We refined the ability to interpret social media comments, even addressing misspellings and emojis,” he added.
Canvs MRX is used to analyze text from customer data, primarily from surveys. Two other products, Canvs TV and Canvs Social, analyze public social media comments. “All the product modules utilize the same core ontology, AI-powered text analytics engine, with language models tuned for various research types and industries,” Kucera explained. “And they all benefit from the years of social media conversation Canvs AI has analyzed.” Additionally, the classification model can be customized with user-defined rules.
“Our model has been trained for over a decade, and we have a trillion terms in our ontology,” observed Kucera. “Manual coding of comments is very labor-intensive and not feasible on a large scale. In addition, human coders can cherry-pick the verbatims they use in their reports, which creates a bias.” Use of text analytics improves consistency and allows all customers to weigh in so the company can get a comprehensive understanding of their customers.