Retailers ‘Listen’ to What Customers Are Typing
The latest generation of text understanding technologies translates consumer comments into actionable business intelligence.
Did you know that patients tend to refer to their physicians as “doctor” when they are happy with the care they have received, and as “he” or “she” when they are not? Or that a pricey bottle of wine often smells different to consumers than the same wine offered at a lower price?
These and other curious insights into consumer behavior were identified by a “text understanding” solution developed by researchers at the MIT Media Lab. This SaaS analytics technology, offered commercially by the Cambridge, Mass.-based startup Luminoso, gives computers the ability to understand human text communication the way people understand each other. It captures online consumer comments about companies and products and turns them into actionable business intelligence.
“This system can pull in chat, surveys, emails, and news articles and quickly offer insights into the opinions of those who wrote them,” says Catherine Havasi, Luminoso’s co-founder and CEO. “It can even understand allusion, metaphor, and the jargon of specific industries such as biotech or pharma.”
According to Marcus Shingles, a principal at Deloitte Consulting LLP, such solutions are the result of an ongoing effort by CPG companies and others to use technology to make sense of the avalanche of comments customers leave online. “Over the last few years we’ve seen numerous text analytics products enter the market that use word recognition and basic sentiment analysis capabilities to understand customer intent,” he says, while adding that “few have proven to be effective.”
Shingles says that the latest breed of text understanding solutions deploys artificial intelligence (AI), machine learning, and natural language processing capabilities that, when used in concert, can take text understanding to a whole new level. “Engineers and researchers are applying a more sophisticated, scientific approach to this challenge, which is what it needs,” he observes. “As a result, they are continually developing a more nuanced understanding of language and of the context in which it is used.”
Consumers Say…
According to Havasi, the Luminoso solution was born of her work with an ongoing research project at MIT Media Lab called the Open Mind Common Sense Initiative. In the mid-2000s, she and other researchers began building a database of simple English sentences, which a computer running text analytics software then analyzed to connect concepts and draw conclusions. The more data they fed into the system, the greater its understanding of vernacular, speech patterns, intonations, and other factors of common speech became. Simply put, Havasi and her colleagues taught AI how to make itself smarter.
“It can easily learn new things,” says Havasi of Luminoso’s technology. “It uses words it already knows to understand the context of the way new words and phrases are used, and then deduces their meaning in that context.”
The ability to understand customer sentiment more precisely and accurately—and to understand the myriad ways consumer demographic groups use language—can add value to product development, customer care, marketing, and strategic planning efforts. “There is a big opportunity for marketers in understanding the opinions, views, and preferences of customers,” says Shingles. “With that insight, they can develop empathy for consumers, and use that to inform marketing initiatives and product design.”
Moreover, the ability to automate certain data management tasks that have traditionally been laborious and expensive may transform the way companies approach consumer research. “A lot of people think there are large costs associated with building an ontology, and that they will have to write a lot of complicated rules,” says Havasi. “That’s not necessarily the case anymore. Text analytics has gotten a lot smarter and more agile than it was just a few years ago.”
Data Quality and a Measured Approach
Even with recent advancements in text analytics, the value of business insights is still heavily dependent on the customer data companies amass, says Havasi. “Many organizations have collected massive amounts of text data that do not reveal much about customer experiences with a product or service. This is low quality data, delivering few signals despite its volume.”
Havasi says that by focusing on certain data sources, companies may be able to unearth more valuable information. “Many companies are finding troves of valuable data on social media sites. People also share their personal experiences on custom blogs and in comments on YouTube. Twitter, with its character limitations, is not such a good source.”
One other thing, says Havasi: “Computers don’t understand sarcasm and snark. They are rarely expressed in the same way, which makes them look like noise to a computer.”
Beyond data quality, organizations deploying text understanding technologies may face internal challenges, such as how to ingest insights from this kind of analysis. “Many companies are deploying these tools, but they are delegating junior-level people in the organization to experiment with them,” observes Shingles. “Achieving the kind of business insights companies want is equal parts art and science. It requires experienced data analysts versed in the marketing science discipline to capitalize on this type of research. Moreover, companies should consider developing business use cases to guide their efforts. Text understanding initiatives might be more useful for some brands in a company’s portfolio than others.”
Havasi predicts that, in the near future, text understanding technologies will become more commonplace. “Expect advances in artificial intelligence that will support human interaction with cars, television sets, and computers,” she says. “It will continue to help companies listen to consumers, but it also will become a part of our everyday lives.”
Related Content: “How to Make Analytics Projects Pay Off”
Questions? Write to Deloitte CIO Journal Editor
This publication contains general information only and Deloitte LLP and its subsidiaries ("Deloitte") are not, by means of this publication, rendering accounting, business, financial, investment, legal, tax or other professional advice or services. This publication is not a substitute for such professional advice or services, nor should it be used as a basis for any decision or action that may affect your business. Before making any decision or taking any action that may affect your business, you should consult a qualified professional advisor. Deloitte shall not be responsible for any loss sustained by any person who relies on this publication. Copyright © 2013 Deloitte Development LLC.
