Text analytics broadens its reach

Enterprises are using only about 25 percent of their unstructured data for insights and decision-making, while they are using 35 percent of their structured data for those purposes, according to a survey conducted by Forrester. One reason for the lag is that unstructured data is not as straightforward to analyze and interpret.

“Many steps are involved in text analytics,” says Boris Evelson, VP and principal analyst at Forrester, “from accessing the data sources to extracting and processing the data, performing advanced statistical analyses, and interpreting and enriching it with domain knowledge.” Yet the value of the analyses is so significant that more and more companies are motivated to take on the challenge.

Lenovo is a large manufacturer of PCs, with sales of 60 million machines in 2014. Its products range from desktops to laptops, workstations and servers. The company had total revenue of over $40 billion last year. To drive engineering innovation, Lenovo launched a program about two years ago that uses text analytics to uncover consumer input contained in unstructured data. The data sources include survey responses, blogs, forums, product reviews, conversations documented in notes from its customer relationship management (CRM) systems and social media.

Already a user of business intelligence (BI) products from SAS, Lenovo began using SAS Text Miner, SAS Sentiment Analysis Studio and SAS Visual Analytics to analyze information from about 50 different channels. “Wherever we can find feedback from customers, we are going to mine it,” says Mohammed Chaara, director of customer insight center of excellence at Lenovo. “We wanted to be quicker to detect product quality issues, because in the short lifecycle of electronic products, detecting problems quickly is a game changer for us.” (Download chart)

Let the customers redesign

Most of the initial text analytics initiatives centered on usability of products. For example, Lenovo redesigned the keyboard on its laptop, but found that customers were not reacting positively. Through text analytics, the company was able to pinpoint the frustration point precisely. The information was provided to the engineering team, which released a new generation of keyboards quickly

“Product development involves many design details,” says Chaara, “such as the number of USB ports, the screen resolution and so forth. One product developer said he wanted to help make our customers the ones who redesign the products. Text analytics helps us do that.” Laptops and notebooks tend to have the shortest lifecycle, so much of the focus has been on those devices.

The ecosystem of SAS tools includes a wide range of elements, from the backend systems that store the data to frontend user interfaces. “To develop the analytics, we need data scientists who understand the software and the analytical models,” Chaara says. “We have a community of stakeholders, about 400 employees, who log onto dashboards to gain insight from the results.”

Lenovo’s ultimate goal is to use text analytics to cover the customer experience end to end, from marketing to customer support. “We want to get feedback at all stages,” he says, “but we started with quality control.” That approach is somewhat unusual; text analytics is more often the purview of the marketing department, which uses the information to generate campaigns.

“We did encounter some resistance,” acknowledges Chaara, “because we were trying to do something that had not been done before—using text analytics for early detection of problems. But if we had started in a more traditional way, with marketing, it might have limited our creativity, because we would have used the traditional social listening methodologies that have been used in marketing.” Now the group has a green light to explore the use of text analytics to improve performance across the board. “There is always room to go wider or deeper, more things to do to increase our understanding,” Chaara says.

The breadth of it

The range of uses that Lenovo is planning for its text analytics is a good indicator of the breadth of the technology. “Text analytics is becoming ubiquitous,” says Fiona McNeill, global product marketing manager at SAS. “It is seen as just another form of data. Bringing text analytics into predictive quantitative models can improve their accuracy by 10 to 30 percent.” In some cases, text analytics may provide such a large part of the insights that the structured variables drop out, according to McNeill. In other cases, the structured model is so finely tuned that text analytics does not contribute significantly. “Each situation is different and needs to be evaluated on its merits,” she says.

Another use case that is becoming popular is the development of a knowledgebase from text such as the service notes that technicians write. “Companies are concerned about the graying of their populations and expertise walking out the door,” McNeill says. “By analyzing large repositories of text and creating metadata from that process, companies can create a searchable knowledgebase from internal documents.”

Text analytics can also be used to improve processes. “This example is not just a case of finding relevant content, but of locating chokepoints in a process based on customer comments,” McNeill explains. “Although business automation process software provides good monitoring of when a slowdown occurs, it does not necessarily provide insights into why.” In one office that was dealing with errors in taxpayer filings, the notes about the errors were categorized automatically using linguistic rules, which allowed quicker resolution.

Cybersecurity support

The issue of cybersecurity now touches nearly every department in an organization, and has become pervasive at all levels from entry employees to the C-suite. Because of the continually changing techniques used by hackers, preventing data breaches is considered nearly impossible. In a study of healthcare organizations, the Ponemon Institute found that 90 percent had experienced a breach, and 38 percent had experienced five or more breaches.

Because of the highly sophisticated and continually evolving techniques used by hackers, preventing data breaches is considered nearly impossible. Although organizations will continue to use and improve their protective measures, an increasing focus is on detection and remediation.