In a more sophisticated, real-life example, one insurance company applied text analytics combined with traditional structured data analytics to construct a predictive model for claims costs. The company wanted to improve its estimation of claim costs at first notice of loss in order to focus on reducing high-cost claims expenses. The company worked with consulting and vendor partners to build a predictive claims-cost model. Initially, the model used only structured data, but when key words from the unstructured textual claims description were added as model features the predictive performance jumped by 14 percent. This performance gain helped the company assign its most experienced agents to its most complicated claims, which resulted in significant reductions in claims expense.
Challenges in Implementing Text Analytics
Although text analytics presents unique opportunities to capture more value from data, it also presents challenges -- some related to its novelty and others related to the difficulty of working with unstructured content. Also, some of the challenges with a typical analytics project are magnified when text comes into the picture.
The first challenge is actually working with the data. Access, storage and retrieval of textual data may require a different approach than structured data. The data is likely to be highly variable, noisy, uncontrolled and heterogeneous, and privacy issues may come into play, especially if email is involved. Finally, query languages focused on semi-structured and unstructured data require a new skill set for analysts to master.
Next, there would be a learning curve to navigate the new text analytics toolsets on the market. They come with new vocabulary, including document categorization, information extraction and sentiment analysis. Scalability is likely to become even more of an issue than with structured data due to the large volumes of text that often need to be analyzed. Finally, integrating textual data rules and taxonomies with existing enterprise applications and business rules may be needed to obtain the most effective outcomes.
To mitigate the impact of these challenges and reduce the effect of the learning curve to the organization, a practical course may be to implement a text analytics initiative with a limited pilot project. Likely pilot areas would be those that have large volumes of readily accessible textual data, such as customer analytics. Once the value of using text analytics is shown with the pilot, the initiative can be expanded to other areas, with the benefit of knowledge gained and successes won.
Wrapping it Up
The majority of a typicalbusiness’s stored information is in an unstructured, textual format. Most businesses do not leverage this information to improve their bottom line. Opportunities for capturing early signals of customer sentiment and financial outcomes can first emerge as textual data. Text analytics provides automated, repeatable solutions that identify useful information hidden in those unstructured documents.
The use of text analytics has great potential but requires moving carefully for there is not a one-size-fits-all solution to the business problems. It also requires that business needs should lead the initiative, instead of investing in the technology simply because it’s new.
Finally, the skills and technologies needed to succeed with text analytics are still evolving, so talent may have to be recruited or retrained. However, by starting with a well-defined problem and demonstrating its value and impact to the business, opportunities will emerge and can be leveraged. These new areas could include contract performance analysis, R&D support, workplace safety analytics and drug or medical care safety analysis. The potential is virtually endless.
This publication contains general information only and Deloitte is not, by means of this publication, rendering accounting, business, financial investment, legal, tax, or other professional advice or services. This publication is not a substitute for such professional advice or services, nor should it be used as a basis for any decision or action that may affect your business. Before making any decision or taking any action that may affect your business, you should consult a qualified professional advisor. Deloitte shall not be responsible for any loss sustained by any person who relies on this publication.
Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited, a U.K. private company limited by guarantee, and its network of member firms, each of which is a legally separate and independent entity. Please see www.deloitte.com/about for a detailed description of the legal structure of Deloitte Touche Tohmatsu Limited and its member firms. Please see www.deloitte.com/us/about for a detailed description of the legal structure of Deloitte LLP and its subsidiaries. Certain services may not be available to attest clients under the rules and regulations of public accounting. Copyright 2010 Deloitte Development LLC. All rights reserved. Member of Deloitte Touche Tohmatsu Limited.
Cindi Thompson, Specialist Leader, Deloitte Consulting LLP, leads the Text Analytics team within Deloitte Analytics. She has more than 13 years R&D and project management experience in industrial, consulting, and academic settings. Her areas of expertise include text and social analytics, machine learning/data mining, and adaptive recommendation systems. Cindi has a Ph.D. and master's in Computer Sciences from the University of Texas Austin, and a B.S. in Computer Science from North Carolina State University.