Feb 01, 2013

Applying Unstructured Data Analytics to Customer Feedback Data

Spreadsheet or database driven analysis has been around for a long time, and last month in DCR TrendLine we shared what we can do with textual data. Firms can collect and analyze huge amounts of data from the web for business insights using a combination of text mining for extracting unstructured data in combination with big data for analysis. This month we share another application, now from internal data sources of a company, that can trigger some thought and also spawn new applications.

“Shallow methods work extremely well… except when they don’t.  Language is replete with underlying structure.” ~Philip Resnik, University of Maryland

One key source of textual information is the customer helpdesk. Many enterprises employ tens of thousands of call center employees to service customer for technical support, sales ordering, complaints and other needs. They capture a goldmine of customer insights that remain locked in transactions but never reach the higher value functions like product designers, SCM planners and sales teams. Even if it were to be available, the problem would be akin to looking for a needle in haystack, given the quality of such data. Customer feedback could be on a continuum from ridiculous to sublime and being textual it has to be read, understood, inferred and used. Further more, it has to be aggregated in a manner to tap wisdom of crowds, and not just individual. It is here that structured data mining can efficiently read and classify data based on structured data analysis algorithms.

For example, a computer manufacturer when launching a new line of laptops, was keen to know how the product was received in market so that it could fine-tune future campaigns of promotion in other geographies. It had a launch with much fanfare, and trended on twitter too on the day of the launch. The volume of feedback exceeded the expectations due to a hyped promotion as also some unexpected problem with the microprocessor cooling in the product.

What was remarkable about the analysis was the quick identification of top problems, top positives, and demographics of target market, that were quickly derived from unstructured data. The product team could latch onto some transparent, live feedback and guide the call center to proactively reach out to the customers even as they made changes to the product usage literature and design.

“The role of NLP [Natural Language Processing] is not “understanding”. It’s helping people do useful things with language.” ~Philip Resnik, University of Maryland