For Matthew Newton, text mining tools have the potential to connect the dots between unstructured information to get a better understanding of customer needs.

Text mining tools dig info from unstructured corporate data

For Matthew Newton, text mining tools have the potential to connect the dots between unstructured information to get a better understanding of customer needs.

Newton is the manager of consumer advocacy and quality for Benton Harbor, Mich.-based consumer appliance maker Whirlpool Corp. The firm was looking for a way to improve its warranty claim process. With an increasing paperwork load, there was a growing disconnect between the Whirlpool call centre and its repair technicians, Newton said.

Text files account for a large percentage of Whirlpool’s data warehouse; the firm wanted to be able to quickly access and analyze this potentially useful data.

Text mining tools enable firms such as Whirlpool to extract data from large unstructured data sets. “We’ll be able to dig through the service technician text, create value from that and link it back to our call centre text so we can see what the customer said, and to connect the dots to create better analysis on the call centre end,” Newton said.

Access to this information allows call centre staff to better analyze the issue over the phone, without having to send out a technician, Newton said.

Michael Turney, program manager of industry solutions at Toronto-based SAS Canada, noted text mining tools help enterprises to get a sense of common themes and common words within unstructured data to better understand customer behaviour.

The tools are intended to be used with traditional BI data for another layer to customer analysis, Turney said. “Every time you add another means of segmenting data you’re getting a finer and more granular view of customer segments…for that greater level of insight, that greater level of predictive intelligence as to what they’re going to do next,” Turney said.

The ultimate goal of text mining at Whirlpool, according to Newton, is to “go through all our unstructured data that we have within the warranty system to find emerging issues and early signs of quality issues that we hadn’t noticed through the structured data.” For example, the codes on a warranty claim form couldn’t help staff in detecting patterns in the root causes of product failures. “By better understanding the root causes of [product] failures…we’ll be able to have a richer database,” Newton said.

The firm is operating within a SAP ERP environment and recently migrated from a Teradata warehousing system. Whirlpool uses a business intelligence (BI) warranty analysis tool and is running a pilot program with SAS Institute Inc.’s SAS Text Miner tool. The SAS tool “sifts” through unstructured data, formats it, and attempts to establish patterns and relationships between documents. “The long term goal is to extract data from the SAP systems without having to put it in data sets,” Newton said. “Integrating together the warranty analysis tool with text mining was extremely attractive.”

The software is currently being run as a pilot project before being slowly rolled out to the wider enterprise. “It’s installed on a desktop and we’re experimenting with some of the data we’ve got,” Newton said.

The use of text mining software is growing, but the tools have yet to reach the type of functionality offered by traditional data mining offerings. Similar text mining tools are available from IT vendors (including IBM Corp., Oracle Corp., SAP AG, Interwoven Inc. and niche players Inxight Software and Stratify Inc.) and are offered as stand-alone products or embedded as part of a larger software suite. According to Warren Shiau, a software analyst for Toronto-based The Strategic Counsel, text mining isn’t one of the highest-ranking priorities for IT right now but that might change, particularly as the market matures.

Organizations typically operate within a heterogeneous IT environment, Shiau said, and have the need to standardize in order to search this information in an effective way. “You can set up different data stores for text and data but it doesn’t become usable to a knowledge worker until you’re able to get at what’s in those stores,” Shiau said. “If you have to do coding every time you search for something…you are running into problems.”

QuickLink 059814

Related Download
The Case for Business Analytics in Midsize Firms Sponsor: IBM Canada Ltd
The Case for Business Analytics in Midsize Firms
Organizations that apply analytics outperform their peers – it’s a fact.
Register Now
Share on LinkedIn Share with Google+ Comment on this article