What will a mobile manufacturer or a car company do if it is flooded with thousands of e-mails and chats, complaining about poor quality of products or feedback that will be useful in fine-tuning the products?
Culling out information and feeling the customer pulse is not that easy, says Dr Manish Gupta, Director, IBM Research (India) and Chief Technologist of IBM (India and South Asia).
"People could be angry, happy, sarcastic, showing myriad feelings. You need to have a very big team to do that. You need to know what they (customers) feel about your company and products. If you do that fast, chances are you can do some fire-fighting and improve image. This process will help you identify some common issues that need to be attended to first.
"It is not easy to sieve the opinions from voluminous information pouring in, every minute. Using IT to read and understand feelings is a good idea to understand their views in no time," says Dr Gupta.
The quantity of data that companies and organizations receive is mind-blowing. IBM says data that comes through e-mail, chats and telephonic transcripts constitutes about 80 per cent of enterprise customer data.
"We have developed a tool called VOCA (Voice of the Customer Analytics) that can do sentiment mining. What you usually get in email is highly unstructured. We will tell our customers how to filter information automatically," he says. IBM has done a pilot for a car manufacturer, using sentiment mining solution. "We have taken this beyond pilots and are offering this commercially," Dr Gupta says.
Data cleansing tool
Another solution that IBM has developed is a Data Cleansing tool that will help in having a single view of the customers. "Companies get information from a variety of sources on their customers. They need to put together all this information and funnel it to remove redundancies in information in order to have a single view of customers," Dr Gupta says.
Unless you have single and compact information, it is difficult to communicate with customers and take necessary steps. "Data Cleansing tool involves understanding of complex set of information. For one, take the way people write addresses. We randomly use 'near', opposite, adjacent, while writing the same address, giving different coordinates," he points out.
Matching of name and addresses would help in doing away with duplication and set apart one user from another. Another problem is, some proper names are interchangeable. "We do not have a fixed format of writing addresses. For example, Sansad Marg and Parliament Road are interchangeable in address and other references. The tool will understand such complex rules, while sieving through information," he says. "We keep in mind such complex rules in several countries while we deploy it in various countries," he adds.
-- The Hindu Business Line, October 2010