The Real Benefits from Text Mining Olivier Jouve Vice President SPSS Rebecca Wettemann Vice President Nucleus Research
Agenda SPSS and Text Mining Our analysis of text mining Identifying the biggest benefits from text mining Best practices Fine-tuning tips Examples Looking forward
Customer Adoption of SPSS Text Mining Technologies Acquisition of LexiQuest in 2002 More than 2400 unique organizations Many of the top 500 Fortune companies: Telco/ISP Pharma/Life Sciences Media Finance, Bank, Insurance Public Sector Retail Manufacturing
What are our customers insight and retention at doing Cablecom, with Text Mining? Understand customer preferences in detail by analyzing notes fields in call center applications Improve their ability to predict which customers are likely to defect or churn, and take appropriate action to prevent it Predict the offers customers are most likely to accept, increasing up-selling and cross-selling results whether in person, in the call center, or online Identify customer issues and measure the preferences that are expressed in open-ended survey responses or Blogs Federico Cesconi, head of customer said, We ve uncovered concepts and relationships in text that would be too costly or even impossible to detect by any other methods. We can now combine multiple data sources to evaluate customer expectations and improve customer satisfaction by employing more one-to-one customer contact and preemptively resolving customer complaints to keep our retention rates high. Jen Brown, equipment monitoring manager at Komatsu America, said, SPSS Text Mining software provides Komatsu the ability to combine insights from unstructured sources and business data. We are able to better understand why machines fail and why they may go down in the future, leading to improved service capabilities and more satisfied customers. Sikorsky selected SPSS Predictive Analytics text mining and data mining software and its Predictive Analytics enterprisewide enabling platform to analyze helicopter and pilot data collected by the aircraft health and usage monitoring system and flight maintenance log records. Sikorsky will be able to determine the relationships between how the aircraft is being operated and maintained and the consumption of parts. The deeper understanding of these Target influencers and/or detractors in the Web 2.0 sphere Predict when product components may fail or production equipment need maintenance, and better control both product quality and operating costs Jochen van correlations der Wal, technical will allow engineer Sikorsky at to the Dutch take National proactive Police; action said, to reduce "After implementing direct SPSS maintenance Text Mining costs. software and deploying it to a crime case, We found an essential connection within just five minutes which we couldn't have found in the past three months of investigations Predict what types of fraud, waste, and abuse are likely to occur, and Niels Schillewaert, managing where, partner at by Insites analyzing Consulting, textual a information Market Research such Company, as notes fields said, We deploy SPSS Text and e-mails Mining software to mine blogs and communities, and analyze Protect open ended public questions. safety SPSS and security more allows effectively us to track online by using buzz predictive and enrich the analysis of open text analysis to improve models of ended questions, which provides potential us with threats useful customer by individuals and groups information.
Customers on the Predictive Enterprise Journey Leading US Telco Company: Behavioral data => demographics => text => web behavior => Attitudinal => Text =>process automation => high-performance scoring =>multi-channel deployment => multiple applications Cablecom: Demographic data => behavioral => attitudinal data => Text => satisfaction modelling => multichannel data collection => multiple applications Driving strategic focus on: People data Analytical environment Deployment capabilities
Technology Slide No Technology slide!!! Text Mining/Analytics mainstream? Efficient solutions use Natural Language Processing and a few years of tuning To extract accurately sentiments, facts, events from any type of content To be used by the masses By companies with international operations BUT ROI figures
The Benefits from Text Mining May, 2008 Rebecca Wettemann Vice President rwettemann@nucleusresearch.com Nucleus Research www.nucleusresearch.com
About Nucleus Research > A technology advisory firm delivering investigative analysis and advice. > 1000 published ROI case studies > 4.7M ROI tools distributed > Research centers in Boston, Paris, and London > The only firm registered with the National Association of State Boards of Accountancy Registration #108024
Nucleus analysis of SPSS text mining > Nucleus conducted in-depth interviews with SPSS text mining users in the US and Europe. > The sample included organizations in the financial services, telecommunications, high technology, market research, public/nonprofit, and automotive sectors. > Key benefits achieved included: > Reduced churn > Improved management of promotions > Increased visibility > Increased analyst productivity > Improved product development and refinement
What s interesting here? > Users can analyze, categorize, and draw conclusions from unstructured data such as text. > Common document types as well as databases, RSS feeds, and blogs can be analyzed. > The opinion dictionary can identify and categorize sentiments such as likes and dislies. > It natively supports multiple languages and supports many others through translation. > The Clementine workbench can integrate text mining into predictive modeling techniques and strategies.
Where are the biggest benefits?
The 5 Key factors of ROI > Breadth How many people will the application affect? > Repeatability > Cost How many times a day will they use it? ------------------------------------ Is this a costly task? > Collaboration Will employees need to collaborate? > Knowledge Can I reuse the information I create?
Factor 1 - Breadth The greater the breadth of the application, the higher the potential return.
Factor 2 - Repeatability Will the application be used frequently or infrequently? The greater the repeatability of the application, the higher the potential return.
Factor 3 - Cost The greater the cost of the task, or the greater the benefit, the higher the potential return.
Factor 4 - Collaboration Does this task involve collaboration among groups? The greater the collaboration component of the task, the higher the potential return.
Factor 5 - Knowledge The greater the use of knowledge management the higher the potential return.
Best practices
Best practices > Move beyond data > Pilot with your own content > Take advantage of training > Set realistic expectations > Use Clementine for data mining first
Fine-tuning tips
Fine-tuning tips > Incorporate more sources > Evaluate new features > Introduce emotive techniques > Learn from your peers > Continue to evaluate and evolve
And
Missteps to avoid > Don t just focus on churn > Don t ignore the complexity of the dictionary > Take time to prepare your content > Don t expect text mining to do it all
Examples
Example: increased productivity Some analysts were able to increase productivity by up to 50 percent. Employees * time savings * fully loaded cost * correction factor or Number of new hires avoided * fully loaded cost
Example: increased visibility Our internal clients like managers and directors can get answers within minutes when it used to take weeks. Benefit 1: Time savings for employees to deliver information (most direct) Time savings * productivity correction factor * fully loaded cost Benefit 2: Better decision making (more indirect) - Share of increased profits attributed to decision making? - Key cost reductions enabled through better visibility into operations?
Example: reduced churn Text mining can cut customer churn in half. Benefit 1: Average annual revenue per customer * number of customers * percentage change in churn * profit margin Benefit 2: Reduced / eliminated cost of alternative retention efforts
Example: increased satisfaction It really put a lift into the morale of our analysts. Great. Can t count it?
Looking ahead
Future influences >Google >NABD >iphone
Summary > Text mining can help companies leverage all the unstructured information they have about products, services, competitors, and customers to increase customer satisfaction and loyalty. > The ability to rapidly analyze unstructured information is a growing competitive differentiator. > The most successful projects are phased rather than large and disruptive. > As the technology evolves the opportunity to provide the power to more users will deliver greater benefits. > Deploy for a rapid payback so you can continue to evaluate and evolve.
Resources Nucleus Research Web site: www.nucleusresearch.com Nucleus Research knowledge center > Tutorial > B20 ROI Quick Reference Guide > A11 Managing Payback and Risk > A10 Maximizing ROI > A21 The Strengths and Weaknesses of TCO > A4 Human Factors Impact Application Value