Companies looking for innovative ideas to improve their products have what amounts to a free and easy focus group right at their fingertips with consumer comment sections on e-commerce websites like Amazon.

Weiguo Fan, professor of business analytics at the University of Iowa’s Tippie College of Business, says comment sections are loaded with information companies can use to improve their products, with customers giving ideas on how they think the product can be more useful.

“Mining product innovation ideas allows a manufacturer to proactively review customer opinion and unlock insights about new functionality and features that the market expects, in order to gain a competitive advantage,” says Fan. “This type of information is particularly important for product functionality improvement and new feature development from a manufacturer’s perspective.”

Finding those nuggets of consumer wisdom can be a challenge, though. Few companies have the resources to sift through tens or hundreds of thousands of comments across dozens of e-commerce sites, only a small number of which offer any kind of information that could be used for product innovation.

But Fan says machine learning might be able to help. An algorithm can be designed that looks for words and phrases that offer innovative ideas. The more reviews the algorithm analyzes, the more it can generate semantic and contextual representations of words, learning as it goes and becoming increasingly accurate in identifying comments that offer innovative ideas.

To test the idea, Fan and Min Zhang, a University of Iowa graduate student in informatics, developed an algorithm using machine learning methods and then used it to analyze 10,000 randomly selected comments from Amazon.com. The comments had already been manually reviewed by the researchers, who highlighted 243 sentences that suggested innovative ideas. The algorithm flagged 91 percent of those sentences.

Some of the reviews were more helpful than others. “I wish this pen had night vision and motion detection” is not likely an innovative idea that many companies are going to act on. But another comment suggesting that a child’s toy be made wider so it’s more stable for toddlers is something the company might be able to use.

Fan says the study shows that using machine learning to analyze online reviews is an effective and inexpensive tool to improve existing products and develop those that are new and innovative.

Fan’s and Zhang’s study, “Mining Product Innovation Ideas from Online Reviews,” will be published in a forthcoming issue of the Journal Information Processing and Management. It was co-authored by Brandon Fan of the University of Michigan, Ning Zhang of Qingdao University in China, and Wenjun Wang of the University of Arkansas.

Media contact: Tom Snee, tom-snee@uiowa.edu, 319-541-8434