3 Steps to big data success with text analytics


Text Analytics 2A huge proportion of big data is unstructured text (such as client interactions, product reviews, call center logs, emails, blogs and tweets). Organizations starting to invest in advanced analytics often overlook the value text analytics could add to the process. But when data scientists or analysts get to work exploring the available data to solve specific business problems, they often find that unstructured text contains the more comprehensive information.

In fact, the demand for text analytics has skyrocketed. Forrester finds that text analytics implementations have doubled since 2012. Every organization in every industry has unmet needs and opportunities – and therefore growing interest in tapping into unstructured text. Organizations across industries have adopted text analytics for a variety of use cases, and the ones that have been most successful followed these steps:

  1. Start with the end in mind. What are your goals, both broadly in the business domain and for the specific analytics initiative? Start with the business objective and then look for opportunities to analyze data previously ignored because it was unstructured. A well-defined objective will tie into the following questions: How do you want to deploy the analytics? Who are the consumers of the results? How will they interpret the analysis, and what decisions will they make? For example, what are the top priorities for the customer experience that you want to deliver? Are they customer service, then product quality, and then price? Some text analytics methods, including sentiment analysis, are often called into question because they’re applied to the wrong problem.
  2. Define your success criteria. Success may be expressed in terms of improving a specific business performance metric through an analytics model of specific accuracy. A clear definition of success helps the analytics team find the most direct path. For example, if success means increasing customer satisfaction by a certain percent by accurately categorizing call center notes or service records into relevant groups, the analysis can perhaps be highly automated with minimal impact to operational systems and processes. However, if success means injecting analytics into call center operations to make real-time recommendations using text and predictive analytics, this will be a more resource-intensive effort involving process change and should be worth the effort.
  3. Start small and think big. Your organization may get its feet wet with modest or off-the-shelf applications of text analytics, but plan for the future; have a business vision and a capability expansion strategy. There are questions that have gone unanswered because they were buried in unstructured text, and now technology is not the limitation. Do your due diligence and research text analytics capabilities along with the pros and cons of different approaches. Anticipate how data sources and volume and movement will grow. Also anticipate how you'll scale text analytics technologically, and perhaps globally. Be realistic about filling your talent and technology gaps, and get help where needed from trusted advisors.

Get the bigger picture around these three ways to harness big data with text analytics by downloading this paper by the International Institute of Analytics, Text Analytics: Unlocking the Value of Unstructured Data. You'll learn how to get the most from your data, including what makes this technology so powerful, as well as pitfalls to avoid when implementing it.



About Author

Simran Bagga

Principal Product Manager - Text Analytics

Interviewed by KMWorld and the International Institute of Analytics, Simran is described as a trusted advisor and thought leader on advanced analytics and operational deployment. With 14 years of experience, she has helped organizations see the value of enhancing business processes and making data driven decisions. Simran’s vision and technology direction of SAS Text Analytics is driving innovative ways to incorporate NLP, machine learning and search into AI and the overall analytics process.

1 Comment

  1. Anthony J Clink on

    I have a hard time understanding exactly how to get between step 3 and 1...

    I have several large end goals and very few focused starting goals...

    What ... besides expensive trial an error could assist in the bridge?

Back to Top