News Stay informed about the latest enterprise technology news and product updates.

Avoid public relations issues by mining unstructured data from within

Text Analytics Summit presenter Daniel Ziv explains how mining unstructured data from within can help prevent product complaints from going viral.

It’s a classic example of customer service gone amiss, one that’s been referenced at conferences and training sessions for years. “United Breaks Guitars,” a song recorded by Canadian musician Dave Carroll in 2009 documenting his difficult customer service experience with United Airlines, became a YouTube sensation within days after it was uploaded, pulling in more than 2 million hits in two weeks.

According to Daniel Ziv, vice president of customer interaction analytics at Verint and a presenter at the seventh annual Text Analytics Summit, in Boston last week, emerging trends or complaints that go viral, such as “United Breaks Guitars,” can be a challenge to organizations when it comes to social media. The other challenge: teasing out the useful information from the blather.

“Not only is it unstructured data, but it’s not even clean,” Ziv said.

And not only is information from social media sites unstructured and “dirty,” but by the time a complaint surfaces on Facebook, YouTube or Twitter, the impact -- positive or negative -- is pretty much set in stone, he said.

The ‘double deviation’
Information from social media sites shouldn’t be ignored, but Ziv suggests a strategy focused on integrating internal and external customer data to establish a big-picture perspective on the voice of the customer and to get in front of complaints that have the potential to go viral.

In his presentation, Ziv referenced recent research, which appeared in the MIT Sloan Management Review earlier this year, stating that many complaints gone social come from loyal customers. In many cases, the customers experienced a “double deviation,” or tried to take care of the issue on two separate occasions with the business directly to no avail.

“Those customers will give you a chance to fix the mistake,” Ziv said, “and if you fail twice, they feel justified in airing the complaint.”

In other words, the seeds for customer complaints that could pop up on social media sites may already have been planted internally. And so he recommended broadening the idea of social data to include calls, emails, chat sessions, surveys, focus groups -- any interaction an organization has with its customers.

“Social media [sites are] public,” he said. “If you look internally, within the enterprise, that data is yours. No one has access to it.”

Beyond that, Ziv said collecting customer data internally is cleaner and provides more context than what’s seen on social media sites. For example, a five-minute phone call, still the preferred method of lodging a complaint, is roughly 1,000 words, a volume of information rarely seen on Twitter or Facebook.

The problem is that organizations are better at mining structured as opposed to unstructured data, and less than 1% of unstructured data is being used effectively, he said.

The importance of unstructured data
The idea that mining unstructured data is important was less than surprising to those attending the conference. Just days before the summit, Seth Grimes, the chairman and founder of the event, released survey results revealing that the market for text analytics technology is approaching the $1 billion mark. The growing interest in text analytics was evidenced by the number of attendees at the summit. This year’s show saw a 30% increase in attendance over last year. And many of those who attended said they were not only interested in getting an overview on what’s happening inside the industry, but were planning to make a software investment as well.

Ranjani Venkataraman, a market research specialist with PepsiCo. Inc, was just one example. She said her company is still evaluating software options, but it has become interested in leveraging data from social sources -- specifically from focus groups.

“It’s the most unbiased information,” she said, “because we’re not forcing people to talk about the product.”

Venkataraman said she came with a few questions she wanted answered about challenges other businesses ran into with unstructured data and how they overcame those challenges. One of those challenges, Venkataraman said, was how to link together data that may flow into an organization through different channels, such as email, text or a website.

“We believe social media is important, but we’re also interested in leveraging other data from sources as well,” she said. “We don’t want to use social media sources in isolation.”

This was something Ziv touched on during his presentation.

“A lot of organizations talk about social media as silos,” he said, “These departments -- the social, the call center, the website -- they need to talk to each other.”

The advice should feel familiar. Many business intelligence projects are initiated because siloed information can prevent an organization from making decisions quickly. Just having the customer data -- or the analytics -- isn’t enough. The data must also be used to drive an actionable response. And, Ziv said, data coming from all of those potentially disparate channels within an organization aren’t seen as being segregated by the customer.

“All channels are part of the same customer experience,” he said.

Ziv recommended collecting customer data on everything from phone numbers, home and email addresses, Twitter feeds, Facebook pages and linking those together, paying attention to early warning signs and creating a platform to ensure that everyone who needs the data coming in from these sources has the data.

“We have to start thinking like that,” he said. “We have to start connecting the dots.”

Dig Deeper on Text analytics and text mining

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.