Could text analytics be the unsung hero of big data?
Text analytics mines reams of often ephemeral, unstructured data — from a voice recording of a customer call to emails or a Tweet — for meaningful insights that can inform business decisions, be it a branding strategy or a product launch.
And while retailers have hailed big data as the key to everything from delivering shoppers personalized merchandise offers to real-time metrics on product performance, the industry is mostly scratching its head on how to monetize all the data that’s being generated in the digital era.
One point of departure: Over 80% of all information comes in text format, Tom H.C. Anderson, CEO of OdinText, which markets its text analytics software to clients such as
So if retailers, for one, “aren’t using text analytics in their customer listening, whether they know it or not, they’re not doing too much listening at all,” he said.
Anderson and chief technology officer Chris Lehew shared how OdinText's clients are leveraging what the startup dubs its “next generation” text analytics software, while also putting a few big-data myths to rest.
Myth: Big Data Survey Scores Reign Supreme
“A lot of people think that common structured data, especially survey data, is the gold standard, especially if it comes with a large sample size,” Anderson said. “Many retailers and other businesses glommed onto something called the ‘Net Promoter Score’ about a decade ago when the term was first coined by
Client Problem: “Shell Oil’s Jiffy Lube International, with over 2,000 stores in North America, began text mining their NPS data and discovered that neither NPS nor any other structured survey metric was correlated to revenue,” Anderson said.
Text mining first helped explain why this was the case. As a general proposition, customer satisfaction survey responses shed little light on why consumers do the things they do, he said. “In other words, if they have a good experience and say they are likely to recommend your business, they also seem to think they didn’t wait in line long, that your employees are smart, friendly and well trained, etc. Conversely, those few who have a bad experience also tend to ding you on every single question, giving you very little reliable data to improve on. They either claim to love everything, or hate everything equally.”
Solution: “Using OdinText's text analytics, Shell was able to analyze the text comments left by customers," Anderson said. "These top-of-mind, unaided answers explained what things were most important to the customers, and how Jiffy Lube could affect positive change and have a real measurable impact on not just satisfaction, but also on actual revenue,” he said. “Jiffy Lube was further able to predict exactly what would happen when they addressed specific issues mentioned by their customers, from ‘coupons’ to ‘easiness of visit.’”
Myth: Bigger Social Media Data Analysis Is Better
“Many believe that simply because of the sheer volume of tweets, blog posts and public Facebook posts, social media data must be valuable, and that bigger is always better,” Lehew said. But in reality, “listening to larger amounts of noise is just more noise.”
Client Problem: “Coca-Cola’s social media hub had been analyzing social chatter for quite some time, but was uneasy about the kinds of insights they were getting,” Anderson said. The company wondered, “‘who were these comments really from?’ ‘What could be learned from the average 140-character comment?’”
Solution: Coca-Cola moved from “super noisy simplistic word cloud type findings around 100 million brand conversations, and from insights that weren’t at all useful in planning a brand strategy, to much smarter text analytics,” Anderson said. “By narrowing down their sample from 100 million unknown social media users to specific demographics of interest, such as moms, they could better understand their conversation by channel type and topic.”
The shift enabled “Coca-Cola to feel secure about the significance of their findings, and ultimately better communicate with important customer groups,” he said.
Myth: New Data Sources Are The Most Valuable
Client Problem: “Most companies have a customer-service department whose sole task it is to handle customer inquiries and complaints. Many of these departments are good at what they do, and often generate large databases of call notes from thousands of calls. Unfortunately, while they are good at answering the calls and emails, which usually contain valuable praise, complaints and suggestions, little-to-no analysis is conducted on the data,” he said.
Solution: “While other companies were still just trying to listen to social-media comments from who knows whom, Campbell’s Soup Company realized that the quarter of a million customers who reach out to them each year is even more important,” he said.
In turn, the consumer products giant used text analytics to understand the trends reflected in comments on various product lines and items down to the SKU number.
“The voice of the customer comments collected when listened to in aggregate can help companies see that a specific product line or product within a product line is doing better or worse than others, and also explain why,” Anderson said. “Often problems discovered are solved as easily as better communicating how certain packaging or products should be used, while other times these insights can open up ideas for entirely new and lucrative products.”
Myth: Keep Your Eye On The Ball, And Mind Your Own Business
“Wherever brands are gaining customer insights via large-scale customer satisfaction tracking, social media monitoring etc., many managers assume that they just need to listen to what their own customers do or say,” Anderson said.
Client Problem: “One of the problems here is that without the proper context, that data means relatively little. Take pricing, for instance. Customers will almost always tell you that the price is too high. But is it really?” he said.
Solution: “Thanks to social media, gathering data on what your competitors’ customers are saying is often just as important, and just as easy as gathering the same data on your own customers,” he said.
Starwood Hotels opted to do just that, using the company's text-analytics tools to compare discussion board comments from its guests with comments from guests at Hilton, Marriott, Intercontinental and Hyatt hotels.
“This ‘comparative’ type of analysis puts your issues into perspective and helps you understand exactly where your strengths and weaknesses are," Anderson said.
"Whether your product is sold in brick-and-mortar stores or on sites such as Amazon or both, there are various discussion boards and rating systems now associated with almost every product of consequence on the market,” he said. "Sites like Baazarvoice.com even offer ratings and customer comments on various laundry detergents. This data is usually easy to screen scrape or purchase for analysis.”
