There’s a lot to learn from what you ask Alexa, Tweet to followers, post to Facebook, leave in an online review, or say over the phone to a customer service rep. Combine that with company news releases and reports, news stories in the media, government information, and anything else people are saying online, and that’s a lot of text.
New research from Maryland Smith marketing professor Wendy W. Moe and five co-authors explains how marketers can best use text analysis technology to sift through all of those words to figure out what consumers really want and need. With text analysis, marketers can decode what people actually say and how they say it – rather than relying on what the researchers call “measurement straitjackets,” like scripted surveys – to make better decisions and offer better products and services that fit what consumers say they want.
“The automated part is new,” says Moe. “Researchers have been analyzing text for decades, but with today’s technology, they can do it more systematically.”
Text analysis software looks at individual words and expressions and their relationship to other words and phrases in a document and across documents gleaned from multiple sources. Techniques range from computerized word counting and applying dictionaries to supervised or automated machine learning.
Moe and her co-authors illustrate how text data can be used to both predict and understand consumer preferences. They offer a how-to guide for using text analysis, including the main tools, pitfalls and challenges that researchers may encounter. They provide a framework for incorporating text into marketing research at the individual, firm, market and society levels. And they point to how these methods can “unite the tribes” of marketing researchers to lead to more effective marketing.
Traditionally, says Moe, there are two types of marketers: the creative, design-oriented marketers who work to understand people and psychology, and the quantitative marketers who are tuned into metrics like sales models, predictions, forecasts and advertising responses.
“Text analysis forces both sides to come together,” she says. “You have the quantitative aspect with automated text analysis, but you can’t do analysis in a vacuum. You have to do it in the context of what the text represents from the consumer psychology perspective, and that’s where the two sides come together.”
Moe is co-director of the Smith Analytics Consortium and also heads up all master’s programs, where teaching students how to use technology tools to analyze data for better decision-making is a critical focus.
“One thing that I think is really important in business education is not just knowing how data analytics and technology applied, but also understanding the underlying technology that drives the analysis. This allows you to really understand where you can take the technology and where you can’t or shouldn’t take the technology,” says Moe.
Part of that is understanding how text analysis technology uses machine learning, where computers have to learn what the text means. Moe said privacy concerns are one of the biggest issues with text analysis technology as consumers grapple with how artificial intelligence listens in on conversations in homes and humans on the other side use that information to help the machines learn.
“The machines aren’t inherently intelligent; they need a teacher or coach -- which is a human being -- for them to learn what the text means,” says Moe.
She also pointed to a potential pitfall in the technology: in only offering consumers products they have expressed an interest in, opportunities for serendipitous new discoveries could be shut out.
Moe says she thinks communication will eventually shift to more images and video content, forcing technology to evolve and marketers to keep up. But for now, this research will help them navigate the best ways to use current tools to make the best marketing decisions.
Read more: “Uniting the Tribes: Using Text for Marketing Insight” is featured in the Journal of Marketing.
Media Relations Manager