Move over, Big Data. Here comes Big Text.
Or, as it’s more commonly known, “text analytics.”
Or, sometimes, “text mining.”
Essentially, it’s the analysis of all the unstructured data a bank absorbs through email; social media posts; call center notes and transcripts; surveys and feedback forms—any information provided by or about customers not structured in some type of database.
More specifically, it could be a snarky comment on Facebook; the jottings of a call center representative; scribbled notes at the end of a survey; or a mobile text from a grateful customer to a helpful teller.
All those kinds of things produced by the combined population of current and potential customers provides usable insights—if they can be captured, sifted, sorted, digested, aggregated, and presented in some meaningful way.
Big Text is a growing area that’s generating quite a lot of literature by business and banking observers. Give credit, though, to Beyond the Arc with coming up with the catchier name of “Big Text.”
Things you can do with all those “little bits”
What’s causing businesses—including banks and financial services firms in particular—to start paying attention to text analytics is its potential to fine tune their understanding of customer expectations; identify at-risk customers; measure feedback and reactions; and uncover hidden business opportunities.
Big Text sounds a lot like what big data, and its application through analytics, promises. The problem is, unstructured data is many times more complicated to cram into formats that can be accurately analyzed in a timely manner.
FICO and Infosys, in a joint online guide they’ve coauthored—Analytics in Financial Services—devote an entire chapter to text analytics. The companies put it this way:
“Unlike structured transaction data—which tells what customers did—unstructured data provides insights into why they did it, what else they want to do, and what problems they may have.”
So what’s involved in text analytics—or Big Text? Beyond the Arc says it can:
• Find out what people are saying by scanning text for names, places, dates, and other important words and phrases to learn what people are saying about the organization or its competitors.
• Identify themes and trends by grouping similar information or topics for nuanced segmentation or to spotlight emerging trends.
• Uncover opportunities for improving the customer lifecycle by connecting events and entities to one another and finding where and when the customer experience needs improvement.
• Understand public sentiment by linking emotions implied in textural comments to outcomes.
Making sense of social media (maybe)
A big part of text analytics can be found in its subset, social media analytics.
The FICO/Infosys publication provides this scenario:
A bank offered several online services, but each involved multiple complex user interactions. Users reacted by deciding that the bank was not user friendly and then switched banks—with many not even bothering to fill out online survey feedback forms.
What they did do—and what the bank was able to capture—was express their negative sentiments about the bank in online forums. The bank employed some social media analytical tools and combined them with associated business intelligence tools. This enabled management to come up with a solution to adjust their services to make them more user-friendly.
This is what text analytics can do—at least in theory.
Not quite there, yet
In practice, the state of the art has much catching up to do relative, say, to where big data analytics has come so far.
Boris Evelson, an analyst with Forrester Research, says this in a blog:
“While structured data management and [business intelligence] processes like data integration, data warehousing, reporting, querying, analytics, and data visualization have matured over the past few decades, unstructured data management and analytics lag behind by as much as 30%. Our latest survey data shows that, on average, enterprises leverage about 35% of their structured data for insights and decision-making, but only 25% of their unstructured enterprise data.”
In their joint publication, FICO/Infosys say much the same thing: “Vendors in this space are still emerging and shaping their lines-of-businesses.” Still, they note that there are vendors very active in developing text analytic products and services. The main challenges they face include:
• Integrating social media analytics with text analytics.
• Making text analytics available as a component of other applications.
• Converging various types of analytics.
They conclude: “Many analytics vendors are coming up with solutions to integrate structured analytics with unstructured analytics … The convergence of unstructured analytics with structured analytics is no longer an ʻif’, but rather a ʻwhen’.”
Sources used for this article include: