Homework answers / question archive / Discussion Board Replies Each reply must be at least 250 words

Discussion Board Replies Each reply must be at least 250 words

Business

Share With

Discussion Board Replies Each reply must be at least 250 words. These include one in-text citation, and one biblical integration citation. All posts must strictly comply with APA style standards. All sources used except the textbook required sources must be published within the past five years. Reply to Kentaya Data is the new form of wealth. With advances in abilities to analyze massive data to extract actionable insights that can make significant competitive gains to organizations, collecting more high-quality data had been an obvious pre-requisite. Bartlett (2013, pp. 198-199) noted that datasets of interest depend on the problem data analytics is trying to resolve. For a better solution, it is essential to have the most relevant data. Sandmann, Jit, Robotham, and Deeny (2018) researched a number of gastroenteritis patients in England and concluded that change in the data collection has a high impact on the value of the insights from the analysis of the data. Insights were different from the data taken during the regular weekdays from that of weekends; similarly, the inferences from the winter data provided new insights that were missing when analyzed the data from other seasons. The importance of using technological advances is evident in the data collection process also. Ahmed and Pathan (2019, p. 308) noted that data collection is a systematic way of collecting and evaluating data from multiple sources to resolve a specific problem through analytics. Through data collection, individuals or organizations can gather responses to a set of pre-defined questions, validate the analytics results to determine predictions and trends relevant to the problem. Truong (2018) noted that many IoT devices such as sensors, actuators are used for data collection. Some human interaction would help to improve the quality of data collection. For example, changing the type of sensors, changing the configuration of the sensors based on critical analytics results may improve the quality of the data collected. The intervention may be on-demand than pre-planned. Bartlett (2013, P.226) noted that many software tools are used for efficient collection and storage of data. Many software tools capable of tagging and indexing at the data input stage are slowly replacing simple data warehousing and searching. One critical aspect of data collection that is gaining the attention of many is information privacy issues. Exponential growth in the usage of smartphones and applications automatically opened channels for stealing the personal data of the user. Lack of awareness of consumers in regulations and usage of applications are other reasons. Petrescu (2018) argued that data privacy is a genuine concern of data collection from a consumer viewpoint. Big data can pose many privacy issues, including the data of people who are not directly involved in the active data collection process. Increasing use of IoT (Internet of Things) that enables synchronizing of home appliances with smartphones, the risk of private information collected and shared is also increased. The problems that are promised to be solved using the information from the IoT devices are at the cost of privacy violations. Government regulations for protecting consumer privacy need to be in line with technological advances. Gospel truths are the data for a Christian leader to use for following God's word to glorify this world. Merida (2015, pp. 197-198) stated that we all need God's word to check our sinfulness through gospel truths. We need to give up the acts that are evil from God's perspective and become righteous by Him. We can only receive righteousness through Christ. God's word provides us with personal guidance, in addition, to help us pointing out gospel truths. We should seek God's word for all aspects of life. God's word is the essential resource for our life that needs to be cross-checked for every act in our life. The psalmist reminder in this regard is, "Your decrees are my delight and my counselors" (Ps. 119:24). References Ahmed, M., & Pathan, A. K. (2019). Data analytics. Concepts, Techniques, and Applications. CRC Press, Taylor & Francis Group. ISBN-13: 978-0367570989. Bartlett, R. (2013). A Practitioners Guide to Business Analytics: Using data analytics tools to improve your organization’s decision-making and strategy (1st ed). McGraw-Hill. ISBN: 9780071807593. Merida, T. (2015). Christ-centered exposition: Exalting Jesus in 1 & 2 Kings. Retrieved from https://ebookcentral-proquest-com.ezproxy.liberty.edu Petrescu, M. K., A.S. (2018). Analyzing the analytics: data privacy concerns. Journal of Marketing Analytics 6, 41–43. Sandmann, F.G., Jit, M., Robotham, J.V. & Deeny, S.R. (2018). Revisiting the winter burden of acute gastroenteritis on hospital beds in England: change in data collection supports an analytical method for previously missing values. Journal of Hospital Infection,100 (1), 115-117 Truong, H. L. (2018,). Integrated analytics for IIoT predictive maintenance using IoT big data cloud systems. In the 2018 IEEE. IEEE International Conference on Industrial Internet (ICII) (pp. 109-118). Reply to geoffrey Data Collection and Software As corporations ponder innovative ways and techniques for data collection, attention to must be prioritized throughout the procedures. The collection of data allows corporations to store and analyze critical information (Hoffman et al., 2020). Before corporations invest monetary resources into purchasing, sound judgments and screening for potential hazards should be enacted. Statistical diagnostics are substantially crucial to the corporation regardless of the industries because they impact numerous facets. The accurate and correct identification of deficiencies is critical for corporations’ progression. According to Bartlett (2013), statistical diagnostics provide enormous benefits that assist corporations with screening by providing five crucial benefits: *Detecting mistakes or weaknesses *Measuring the accuracy of an analysis ((Sivarajah & Weerakkody. 2016) . *Measuring the reliability of an analysis. *Providing insight into interpreting the results *Providing insights into potential solutions Throughout the context and ideology of business problems, there are situations where individual analysis can be unreliable. According to Blackett (2013), the “fog of war”, uncertainties with situations, can cause massive unexplained judgments and activities that are hidden weak areas, the underbelly of the process. The identification of errors and shortcomings is crucial for corporations’ progression. Advancement is an essential and crucial process for corporations (Irani et al., 2018, pp 263-286). Organizational and corporation growth can signify increases in performances and productivity using analytics for decision making. The art of progressing to the next phase is enormously crucial for corporations. According to Bartlett (2013), statistical reviews assist corporations in advancing to the next level as corporations’ decision-making ability increase significantly. It is enormously crucial for comprehensive screenings of the data, software, and other components of the analytical-based business decisions process. The need for a clear and thorough overview to ensure compliances should also extend to timeliness, reliability, accuracy, and cost (Bartlett, 2013, pp 177-178). Statistically review, which can be conducted by internal or external personnel, is an essential tool that provides beneficial services to corporations (Ali, 2016, p 662). These critical evaluations and assessments significantly impact corporations in various facets beyond decision-making. The examinations can uncover hidden defaults or deficiencies in other areas of corporations that were not visible to leaders. Collectively, the statistical review is a powerful tool for organizations that improve decision-making and encourages quicker executions. The end goal, aim, and final objective is critical for any process. The aim at times can quickly become lost in the narrative of select topics. The purpose of statistical reviews is to gain proper understandings of the procedures used in the decision-making procedures. After the extensive analysis, effective paradigms or concepts may not be possible for some of the deficiencies. Providing clear, precise, and valuable evaluations is substantially critical to the effectiveness of statistical reviews (Irani et al., 2016, pp 263-286). In conclusion, the collection of data is crucial to small and large corporations (Cessie et al., 2016, pp 25-27). Corporations must use statistical diagnosis to screen for critical deficiencies and errors in the analysis process. Performing internal or external statistical reviews substantially eliminates and identifies errors within the decision-making procedures. Statistical reviews are instrumental tools for corporations (Ali & Bhaskal, 2016, p 662). Inside of Merida (2015), God allows his followers to see his grace by opening their eyes. It is time to understand that prayers are protection. We must not take God’s spiritual impact for granted because Christian history is filled with prominent stories of God delivering blessing in remarkable ways. He wants us to seek him because he will always answer our prayers (King David Bible, 1769/2017, Ps 34:6). For centuries, God has been a solid and steady refugee for his people. As he stated to Prophet Elijah, do not be afraid, because he will never leave us alone. God protects his people all the time (King David Bible, 1769/2017, Ps 50:15). Sometimes God allows things to happen that we do not understand, or we question, but he still protects us without us knowing. God sees what we do not see; He is our shield. References Ali, Z. & Bhaskar, S. (2016). Basic statistics tools in research and data analysis. ResearchGate: Indian Journal of Anesthesia, 60(09),662, DOI:10.4103/0019-5049.190623 Bartlett, R. (2013. A practitioner’s guide to business analytics: Using data analysis tools to improve your organizations decision making and strategy (1st ed), McGraw-Hill Cessie, S., Huebner, M. & Vach, W. (2016). A systematic approach to initial data analysis is good research practice. Science Direct: The Journal of Thoracic and Cardiovascular Surgery. 151(01), 25-27, https://doi.org/10.1016/j.jtevs.2015.09.085 Hoffman, K.A., Lobe. B., Morgan, D. (2020). Qualitative data collection in an era of social distancing. SAGE Journals, https://doi.org/10.1177/1609406920937875 King James Bible (2017). King James Bible Online. https://www.kingjamesbibleonline.org (Original work published 1769) Merida, Tony (2015). Christ-centered exposition: Exalting Jesus in I kings & 2 kings. B & H Publishing Sivarajah, U. & Weerakkody, V. (2016). Critical analysis of big data challenges and analytical methods. Journal of Business Research, 70, 263-286, https://doi.org/10.1016/j.jbusres.2016.08.001 Copyright © 2013 by Randy Bartlett. All rights reserved. Except as permitted under the United States Copyright Act of 1976, no part of this publication may be reproduced or distributed in any form or by any means, or stored in a database or retrieval system, without the prior written permission of the publisher. ISBN MHID 978-0-07-180760-9 0-07-180760-8 The material in this eBook also appears in the print version of this title: ISBN 978-0-07-180759-3, MHID 0-07-180759-4 All trademarks are trademarks of their respective owners. Rather than put a trademark symbol after every occurrence of a trademarked name, we use names in an editorial fashion only, and to the benefit of the trademark owner, with no intention of infringement of the trademark. Where such designations appear in this book, they have been printed with initial caps. McGraw-Hill eBooks are available at special quantity discounts to use as premiums and sales promotions, or for use in corporate training programs. To contact a representative please e-mail us at bulksales@mcgraw-hill.com. TERMS OF USE This is a copyrighted work and The McGraw-Hill Companies, Inc. (“McGraw-Hill”) and its licensors reserve all rights in and to the work. Use of this work is subject to these terms. Except as permitted under the Copyright Act of 1976 and the right to store and retrieve one copy of the work, you may not decompile, disassemble, reverse engineer, reproduce, modify, create derivative works based upon, transmit, distribute, disseminate, sell, publish or sublicense the work or any part of it without McGraw-Hill’s prior consent. You may use the work for your own noncommercial and personal use; any other use of the work is strictly prohibited. Your right to use the work may be terminated if you fail to comply with these terms. THE WORK IS PROVIDED “AS IS.” McGRAW-HILL AND ITS LICENSORS MAKE NO GUARANTEES OR WARRANTIES AS TO THE ACCURACY, ADEQUACY OR COMPLETENESS OF OR RESULTS TO BE OBTAINED FROM USING THE WORK, INCLUDING ANY INFORMATION THAT CAN BE ACCESSED THROUGH THE WORK VIA HYPERLINK OR OTHERWISE, AND EXPRESSLY DISCLAIM ANY WARRANTY, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. McGraw-Hill and its licensors do not warrant or guarantee that the functions contained in the work will meet your requirements or that its operation will be uninterrupted or error free. Neither McGraw-Hill nor its licensors shall be liable to you or anyone else for any inaccuracy, error or omission, regardless of cause, in the work or for any damages resulting therefrom. McGraw-Hill has no responsibility for the content of any information accessed through the work. Under no circumstances shall McGraw-Hill and/or its licensors be liable for any indirect, incidental, special, punitive, consequential or similar damages that result from the use of or inability to use the work, even if any of them has been advised of the possibility of such damages. This limitation of liability shall apply to any claim or cause whatsoever whether such claim or cause arises in contract, tort or otherwise. Dedicated to Wei “Cynthia” Huang Bartlett—Wife & Patricia “Patty” Rita Stalzer Bartlett—Mother (1944–2005) Contents Preface Acknowledgments Part I Introduction and Strategic Landscape Big Data Chapter 1 The Business Analytics Revolution Information Technology and Business Analytics The Need for a Business Analytics Strategy The Complete Business Analytics Team Section 1.1 Best Statistical Practice = Meatball Surgery Bad News and Good News Section 1.2 The Shape of Things to Come—Chapter Summaries PART I The Strategic Landscape—Chapters 1 to 6 PART II Statistical QDR: Three Pillars for Best Statistical Practice— Chapters 7 to 9 PART III Data CSM: Three Building Blocks for Supporting Analytics— Chapters 10 to 12 Notes Chapter 2 Inside the Corporation Section 2.1 Analytics in the Traditional Hierarchical Management Offense Leadership and Analytics Specialization Delegating Decisions Incentives Section 2.2 Corporate Analytics Failures—Shakespearean Comedy of Statistical Errors The Financial Meltdown of 2007–2008: Failures in Analytics Fannie Mae: Next to the Bomb Blast The Great Pharmaceutical Sales-Force Arms Race by Tom “T.J.” Scott Inside the Statistical Underground—Adjustment Factors for the Pharmaceutical Arms Race by Brian Wynne Section 2.3 Triumphs of the Nerds Proving Grounds—Model Review at The Associates/Citigroup Predicting Fraud in Accounting: What Analytics-Based Accounting Has Brought to “Bare” by Hakan Gogtas, Ph.D. Notes Chapter 3 Decisions, Decisions Section 3.1 Fact-Based Decision Making Combining Industry Knowledge and Business Analytics Critical Thinking Section 3.2 Analytics-Based Decision Making: Four Acts in a Greek Tragedy Act I: Framing the Business Problem Act II: Executing the Data Analysis Act III: Interpreting the Results Act IV: Making Analytics-Based Decisions Consequences (of Tragedy) Act V: Reviewing and Preparing for Future Decisions Section 3.3 Decision Impairments: Pitfalls, Syndromes, and Plagues in Act IV Plague: Information and Disinformation Overload Pitfall: Overanalysis Pitfall: Oversimplification Syndrome: Deterministic Thinking Syndrome: Overdependence on Industry Knowledge Pitfall: Tunnel Thinking Syndrome: Overconfident Fool Syndrome Pitfall: Unpiloted Big Bang Launches Notes Chapter 4 Analytics-Driven Culture Left Brain–Right Brain Cultural Clash—Enter the Scientific Method Denying the Serendipity of Statistics Denying the Source—Plagiarism Section 4.1 The Fertile Crescent: Striking It Rich Catalysts and Change Two-Trick Pony Section 4.2 The Blend: Mixing Industry Knowledge and Advanced Analytics Cultural Imbalance The Gemini Myths Notes Chapter 5 Organization: The People Side of the Equation Section 5.1 Analytics Resources Business Quants—Denizens of the Deep Analytics Power Users Business Analysts Knowledge Workers Section 5.2 Structure of Analytics Practitioners Integration Synergies Technical Connectivity Specialization Teamwork Technical Compatibility Section 5.3 Building Advanced Analytics Leadership Leadership and Management Skills Business Savvy Communication Skills Training and Experience On-Topic Leadership by Charlotte Sibley Expert Leaders (ELs)—Corporate Trump Cards The Blood-Brain Barrier Advantages of On-Topic Business Analytics Leaders Management Types by David Young Section 5.4 Location, Location, Location of Analytics Practitioners Outsourcing Analytics Dispersed or Local Groups Central or Enterprise-Wide Groups Hybrid: Outside + Local + Enterprise-Wide Notes Chapter 6 Developing Competitive Advantage Approach for Identifying Gaps in Analytics Strategy Protecting Intellectual Property Section 6.1 Triage: Assessing Business Needs Process Mapping of Analytics Needs Innovation: Identifying New Killer Apps Scrutinizing the Inventory Assigning Rigor and Deducing Resources Section 6.2 Evaluating Analytics Prowess: The White-Glove Treatment Leading and Organizing Progress in Acculturating Analytics Evaluating Decision-Making Capabilities Evaluating Technical Coverage Executing Best Statistical Practice Constructing Effective Building Blocks Business Analytics Maturity Model Section 6.3 Innovation and Change from a Producer on the Edge Emphasis on Speed Continual Improvement Accelerating the Offense—For Those Who Are Struggling Notes Part II The Three Pillars of Best Statistical Practice Blind Man’s Russian Roulette Bluff Chapter 7 Statistical Qualifications Section 7.1 Leadership and Communications for Analytics Professionals Leadership Communication Leadership and Communication Training Section 7.2 Training for Making Analytics-Based Decisions Statistical “Mythodologies” Section 7.3 Statistical Training for Performing Advanced Analytics The Benefits of Training Academic Training Post-Academic Training—Best Statistical Practice Training Through Review Section 7.4 Certification for Analytics Professionals The PSTAT® (ASA) (Professional Statistician)— ASA’s New Accreditation by Ronald L. Wasserstein, Ph.D. Professionalism Notes Chapter 8 Statistical Diagnostics The Model Overfitting Problem Section 8.1 Overview of Diagnostic Techniques External Numbers Juxtaposing Results Data Splitting (Cross-Validation) Resampling Techniques with Replacement Standard Errors for Model-Based Group Differences: Bootstrapping to the Rescue by James W. Hardin, Ph.D. Simulation/Stress Testing Tools for Performance Measurement Tests for Statistical Assumptions Tests for Business Assumptions Intervals and Regions DoS (Design of Samples) DoE (Design of Experiments) Section 8.2 Juxtaposition by Method Paired Statistical Models Section 8.3 Data Splitting Coping with Hazards K-Fold Cross-Validation Sequential Validation (with Three or More Splits) Notes Chapter 9 Statistical Review—Act V Élan Qualifications and Roles of Reviewers Statistical Malpractice Section 9.1 Purpose and Scope of the Review Purpose Scope Context Section 9.2 Reviewing Analytics-Based Decision Making— Acts I to IV Reviewing Qualifications of Analytics Professionals—Checking the Q in QDR Restrictions Imposed on the Analysis Appropriate and Reliable Data Analytics Software Reasonableness of Data Analysis Methodology Reasonableness of Data Analysis Implementation Statistical Diagnostics—Checking the D in QDR Interpreting the Results (Transformation Back), Act III Reviewing Analytics-Based Decision Making, Act IV Closing Considerations—Documentation, Maintenance, Recommendations, and Rejoinder Notes Part III Building Blocks for Supporting Analytics Chapter 10 Data Collection Randomization Interval and Point Estimation Return on Data Investment Measuring Information Measurement Error Section 10.1 Observational and Censual Data (No Design) Section 10.2 Methodology for Anecdotal Sampling Expert Choice Quota Samples Dewey Defeats Truman Focus Groups Section 10.3 DoS (Design of Samples) Sample Design Simple Random Sampling Systematic Sampling Advanced Sample Designs The Nonresponse Problem Post-Stratifying on Nonresponse Panels, Not to Be Confused with Focus Groups Section 10.4 DoE (Design of Experiments) Experimental Design Completely Randomized Design Randomized Block Design Advanced Experimental Designs Experimental Platforms Notes Chapter 11 Data Software Section 11.1 Criteria Functional and Technical Capabilities Maintenance Governance and Misapplication Fidelity Efficiency and Flexibility Section 11.2 Automation Data Management Data Analysis Presenting Findings Monitoring Results Decision Making Notes Chapter 12 Data Management Information Strategy Data Sources Security Section 12.1 Customer-Centric Data Management Customer Needs Data Quality—That “Garbage In, Garbage Out” Thing Inspection Data Repair Section 12.2 Database Enhancements Database Encyclopedia Data Dictionaries Variable Organization Notes Concluding Remarks Appendix: Exalted Contributors: Analytics Professionals References Index About the Author Preface “… true learning must often be preceded by unlearning …” —Warren Bennis A Practitioner’s Guide to Business Analytics is a how-to book for all those involved in business analytics—analytics-based decision makers, senior leadership advocating analytics, and those leading and providing data analysis. The book is written for this broad audience of analytics professionals and includes discussions on how to plan, organize, execute, and rethink the business. This is certainly not a “stat book” and, hence, will not talk about performing statistical analysis. The book’s objective is to help others build a corporate infrastructure to better support analytics-based decisions. It is hard to judge a book by its cover. To get a feel for the book, look at Figure 6.1 on p. 117, which shows types of business analytics that can support decision making. Table 6.2 on p. 118 provides a glimpse of how to organize business analytics projects. Figure 6.4 on p. 123 depicts how to assess the relative technical difficulties of a set of business problems. Do these items complement how you think about your business? There is a tremendous opportunity to improve analytics-based decision making. This book is designed to help those who believe in business analytics to better organize and focus their efforts. We will discuss practical considerations in how to better facilitate analytics. This will include a blend of the big-picture strategy and specifics of how to better execute the tactics. Many of these topics are not discussed elsewhere. This journey will require continually updating the corporate infrastructure. At the center of these enhancements is placing the right personnel in the right roles. This book serves to enrich the conversation as the reference book you can take into planning sessions. It is usually difficult to find a reference that addresses the specifics of what to do. This is largely because one size does not fit all. The first part of the book provides insights into how we can update our infrastructure; the second part provides three pillars for measuring the quality of analytics and analytics-based decisions; and part three addresses three building blocks for supporting Business Analytics. This book has a great deal of breadth so that professionals, despite not possibly being on the same page, can at least be in the same book. The recommendations in this book are based upon the cumulative experience of analytics professionals incorporating analytics in numerous corporations—Best Statistical Practice. This book contains 12 sidebars relating experiences from the field and viewpoints on how to best apply analytics to the business. The more you get excited about new ideas, the more you are going to enjoy this insight-intensive book. Finally, I wish to add that the way companies approach analytics is evolving. Big Data is accelerating this evolution. I fully expect disagreements and respect different opinions,1 and so should you. To optimize your reading experience, you should retain those ideas that fit into how you think about your business, and leave on the shelf, for now, those ideas that do not complement your approach. Do you want to win? Do you want your company to gain market share? Of course you do. Now is your opportunity to take your game to the next level! Notes 1. This is a contentious topic and I will not go unscathed. Acknowledgments It takes a team effort to write a book by yourself. I am indebted to Isaac “Boom Boom” Abiola, Ph.D.; Jennifer Ashkenazy; Cynthia “Wei” Huang Bartlett, M.D.; Sigvard Bore; Bertrum Carroll; H. T. David, Ph.D.; Karen Fender; Les Frailey; Hakan Gogtas, Ph.D.; James W. Hardin, Ph.D.; Anand Madhaven; Girish Malik; Gaurav Mishra; Robert A. Nisbet, Ph.D.; Sivaramakrishnan Rajagopalan; Douglas A. Samuelson; Tom “T.J.” Scott; Prateek Sharma; Charlotte Sibley; W. Robert Stephenson, Ph.D.; Jennifer Thompson; Ronald L. Wasserstein, Ph.D.; Brian Wynne; and David Young. Their specific contributions are listed in the Appendix. A reviewed book provides a better reading experience. Part I Introduction and Strategic Landscape The ambition of this book is to take up the challenging task of addressing how to adapt the corporation to compete on Business Analytics (BA). We share discoveries on how to transform the corporation to thrive in an analytics environment. We cover the breadth of the topic so that this book may serve as a practical guide for those working to better leverage analytics, to make analytics-based decisions. Big Data There has been a great deal of large talk about Big Data. One sensible definition of Big Data is that it comprises high-volume, high-velocity, and/or high-variety (including unstructured) information assets.1 The threshold beyond which data becomes Big is relative to a corporation’s capabilities. As we grow our abilities, the challenges of Big Data diminish. The application of the term, Big Data, is evolving to include Business Analytics and the term is overused at the moment, so we will write plainly. The opportunity stems from the volume, velocity, and variety of the information content. This torrent of information is collected in new ways using new technologies. It can add a different perspective and provide synergy when combined with traditional sources of information. This new information has stimulated fresh ideas and a fresh perspective on (1) how business analytics fits into our business model; and (2) how we can adapt our business model to facilitate better analytics-based decisions. The first challenge is to wrestle the data into a warehouse. This involves collecting, treating, and storing high-volume, high-velocity, and highvariety data. We address these growing needs by improving our operational efficiencies for handling the data. Although Business Analytics can help in a data-reduction and organizational capacity,2 this is largely an IT issue and not the subject of this book. IT has introduced exciting new solutions for expanding hardware and software capabilities. Brute force alone, such as continually purchasing hardware, is not a long-term plan for avoiding the Big Data abyss. The second challenge is to handle the explosion of information extracted from the data. This is largely a business analytics issue and it is addressed by this book. If the volume, velocity, and variety of the data are difficult to manage, then how well are we handling the volume, velocity, and variety of the information? Previous authors have made the case for improving Business Analytics. One implication of Big Data is that we need to accelerate our development of BA. This book’s best practices will facilitate increasing our capabilities for performing Business Analytics and integrating the information into analytics-based decisions. Part I of this book will inform our strategic thinking, enabling us to develop a more effective plan. 1 The Business Analytics Revolution “All revolutions are impossible till they happen, then they become inevitable.” —Michael Tigar3 W e are poised to enter a new Information Renaissance that involves making smarter analytics-based decisions. A grove of recent books4 and articles has made the case for competing based upon business analytics (BA). These books reveal a potpourri of success stories illustrating the value proposition. It took a generation or longer to take full advantage of some past technological revolutions, such as the automobile, electricity, and the computer. Business analytics has been introduced to corporations, yet most lack the infrastructure to fully capitalize on the abundance of high quality decision-making information. This progression requires significant changes. Foremost among these are changes in personnel, organization, and corporate culture. The right infrastructure will facilitate moving from tactical applications hither and yon, to integrating analytics into the corporation. Recent interest in business analytics has been characterized by a growing awareness of analytics applications, mature IT (Information Technology), ubiquitous electronic data collection devices, increasingly sophisticated decision makers, more data-junkie senior leadership, shorter information shelf life, and “Big Data.”5 We are experiencing such a deluge of data that, in the future, there is the potential for corporations to be buried in it. Corporate concerns arising from the inefficient use of analytics extend beyond just leaving money on the table because of missed opportunities. Ineffective corporations will not see “it” coming—their demise. They will not know why they suddenly lost their customers one night or why their product is still on the shelves. They will have the data to explain it, yet they will struggle to put the pieces together in time because they will not be prepared. In addition to the need to face Big Data, there is a second layer to the problem. Corporations will continue to be awash in dirty data and filthy information. In a future emergency, they will race to clean the data, filter information from misinformation, and interpret the findings. In this book, we dispel stubborn myths and provide a perspective for understanding the organization, the planning, and the tools needed for business analytics superstardom. We have seen analytics in the trenches of effective and ineffective corporations. We leverage the perspectives of analytics professionals charged with making it happen—that is, those leading their corporations in how to apply analytics, those basing decisions upon analytics, and those providing data analysis. Business Intelligence = Information Technology + Business Analytics6 Information Analytics Technology and Business Information technology and business analytics both involve professionals leveraging data to provide business insights, which, in turn, facilitate better decisions. They provide complementary benefits, and we emphasize the synergy of the two. Concept Box Information technology—Gathering and managing data to build a data warehouse and providing data pulls, reports, and dashboards. (Bringing the data to the business) Business analytics—Leveraging data analysis and business savvy to make analytics-based business decisions. (Bringing the business questions to the data) IT involves data collection, security, integrity, management, and reporting. It begins with gathering data and ends with either constructing a data warehouse or with using the data warehouse for data pulls, reports, and dashboards. In reporting, IT measures a consistent set of metrics to track business performance and guide planning. IT places a great deal of emphasis on efficiency. BA is focused upon supporting and making business decisions by connecting business problems to data analysis—analytics. It tends to work from the business need to the available or potentially available data. BA involves reporting, exploratory data analysis, and complex data analysis, and in our definition, we include analytics-based decision making. We want to minimize the distance between the decision and the analytics. BA overlaps with IT with regard to reporting. While IT emphasizes efficiency and reliability in creating standardized reports that address predetermined key performance indicators, BA scrutinizes the reports based upon statistical techniques and business savvy. The BA skill set is valuable for determining and rethinking how these key performance indicators meet the business needs. Additionally, the BA skill set includes statistical tools such as quality control charts and other confidence intervals, techniques that certainly enhance reports for making better decisions. BA is concerned with scrutinizing the data. To this end, it recognizes nuances or problems with the numbers and traces them back through the data pipeline to discover what these numbers really mean. BA includes complex data collection, such as statistical sampling, designed experiments, and simulations. These endeavors need mathematical, statistical, and algorithmic tools. We can discern IT and BA by their skills sets; their software; and their respective locations in the corporation. IT has a stronger computer software theme, and BA is about data analysis and analytics-based decision making. IT usually reports to a CIO. BA often resides in or near the same division as business operations, closer to the business decisions. BA and IT provide an important synergy. It is difficult to have BA without IT. We want to redefine the BA team to make it more inclusive and close the distance between making decisions that are based upon analytics and performing data analysis to support these decisions. The Need for a Business Analytics Strategy Running a large corporation can be compared to flying a commercial jet in a storm. Industry knowledge is the equivalent of looking out the windows, while analytics and advanced analytics—tracking, monitoring, and data analysis—comprise the various gauges, monitoring equipment, and warning devices. In some corporations, tracking reports and data analysis cannot withstand the tiniest scrutiny. This means that some portion of the corporation’s information is fallacious, and, thus, so are some of the decisions based upon this misinformation. The promise of analytics is to provide better facts and to facilitate better analytics-based decision making. Our world is becoming more complex at a dramatic rate, and our brains7 ... not so much. The importance of data analysis has crept up on our corporations over the past decades. Data is now available in abundance, and our analysis needs range from being straightforward to being extremely complex. We want to better integrate business analytics into the decision making process and thus be able to better compete in the marketplace. We want to meet the quickening pace of decision making, the increased business complexity, and the deluge of Big Data. Analytics-based decision making is essential for making the big decisions and thousands of little ones. A history of business failures underscores the need to master how to compete based upon business analytics. One highly developed application of analytics is in estimating risk and revealing how to manage it. Many of those corporations that fared the best during the 2007–2008 financial meltdown made better analytics-based decisions. First, they validated, reviewed, and refined their risk models. Second, they understood their models well enough to believe them and interpret them in the face of human behavior. To return to our commercial jet example, they understood their instruments well enough to make sense out of them when looking out the window provided the wrong answer. AIG,8 Fannie Mae, Freddie Mac, Citigroup, Bear Stearns, Lehman Brothers, Merrill Lynch, WAMU, Fitch Ratings, Moody’s, and Standard & Poor’s were all competing based upon analytics in a prominent manner. At the time, they might not have realized the extent to which their fortunes and their reputations were exposed to their ability to leverage business analytics into their decision making. The Complete Business Analytics Team Facing the next phase of the Information Age will require rethinking decision management. The turnaround time allowed for making decisions is decreasing. The amounts of data and the amounts of misinformation are rising. We need to extend the business analytics team to include senior leaders investing in analytics, those consuming the information, those performing the data analyses, and those directing these practitioners. We must include analytics professionals, who value statistical and mathematical analysis and yet their job might not call upon them to perform data analysis. By including everyone involved, we can foster more cohesion between decision makers, corporate leaders, and those supplying the data analyses. Also, we need to extend the analytics conversation about how we can apply analytics to the business. In Table 1.1, we introduce four basic functional roles. Our experience has shown that we need sophisticated analytics-based decision makers and directors of analytics with strong quantitative training to meet our business analytics needs. Six Sigma has demonstrated that (1) we must have leadership advocating change, (2) we can change our culture to better leverage analytics in decision making, and (3) it is impracticable to train all of our employees to perform data analysis. Instead, we need to build a specialized group of business analysts and business quants to provide the data analysis. Organizing and expanding the business analytics team will lead to making the other infrastructural changes needed for BA superstardom. Section 1.1 Best Meatball Surgery Statistical Practice = “Most people use statistics the way a drunkard uses a lamp post, more for support than illumination.” —Mark Twain Best Statistical Practice (BSP) is our term for our evolving wisdom acquired from solving business analytics problems in the field. We must perform a data analysis within the context of the business need. This need includes addressing considerations of Timeliness, Client Expectation, Accuracy, Reliability, and Cost. We perform the data analysis within these constraints using statistics, mathematics, and software algorithms. These tools provide business insights that support analytics-based decision making.9 Through experimentation, and some trial and error, we find solutions that are fast, client suitable, accurate, reliable, and affordable enough to meet business needs. We call this ongoing experimentation, The Great Applied Statistics Simulation. Hence, the cumulative wisdom of Best Statistical Practice includes our understanding of how to execute techniques quickly, how to meet the client expectation, what information is needed to make the analytics-based decisions, how well techniques perform for certain applications, how to measure the accuracy and reliability of the data analysis, how we can best leverage the serendipity of data analysis, and how we can provide analyses inexpensively. Figure 1.1 Business analytics workbench Much of our learning comes from performing autopsies (Chapter 9) on failed and on successful analytics-based decisions and data analyses. We infer the best techniques, judge the right amount of rigor, develop our business savvy, and foster the synergism between our training and our experience. We measure the performance of decisions and techniques where possible and extrapolate these findings to where it is impossible to measure performance. For example, a generation of analytics professionals mastered building predictive models on high-quality banking data. Then they applied their refined techniques to other applications and to industries where the data quality was too weak to facilitate mastering the techniques. Best Statistical Practice consists of know-how built upon this continual learning, which, in turn, facilitates faster, better, and less expensive analytics- based decisions. It protects us from hazards that we can not anticipate.10 We further develop our BSP by improving our training, our tools, and our understanding of the business problem. This enables us to make great advances in expanding our capabilities. Finally, we need to keep in mind that the three most expensive data analyses continue to be the faulty ones, the absent ones, and the ones nobody uses. The most expensive decisions are those that fail to leverage the available information. We wish to emphasize that analyzing the data is a technical problem within the business analytics problem. The complete problem includes the broader business needs: Timeliness, Client Expectation, Accuracy, Reliability, and Cost. We must solve the analytics problem within these constraints and work toward an infrastructure that will ease them. Our academic training ignores these business constraints, thus making it imperative that we adapt the theory to practice. BSP, combined with good quantitatively trained leadership, facilitates speed and helps avoid both under-analysis and overanalysis. Quantitatively trained leaders can be relied upon to understand the trade-offs involved in cutting corners to perform the analysis within the broader business constraints. The last six chapters of this book provide the tools necessary to perform Best Statistical Practice. Bad News and Good News First the bad news—all the exciting breakthroughs about leveraging analytics to create space-age nanite technology and revolutionize business are full of embellishments intended to impress us and the shareholders. Corporations are not as sophisticated or as successful as we might grasp from the sound bytes appearing in conferences, books, and journals. Instead opinion-based decision making, statistical malfeasance, and counterfeit analysis are pandemic. We are swimming in make-believe analytics. One major part of the problem is that corporations have difficulty measuring the quality of their decisions and the quality of their data analyses. To measure these, we often need a second layer of data analyses. This is one of the most disquieting problems because, just like brain surgery, it takes a second brain surgeon to figure out if the first brain surgeon is working the correct lobe. Even with the best analysis, it is very difficult to measure the quality of some decisions and some data analyses. At present, there is a rather large gap between obtaining the right data analysis for a decision and actually making the decision. A great deal of good data analysis is misdirected and fails to drive the business. Some of this misdirection suits special interests that want the results to match preset conclusions.11 Meanwhile, it is difficult for others to recognize when there is a disconnect between the data analysis and the decision. Now for some good news—this is all one gigantic opportunity and we can easily make substantial progress. Business analytics can build enormous competitive advantages and promote innovation. Analytics simplifies the overwhelming complexity of information12 and decreases misinformation emissions. Finally, less is more. A tremendous amount of analytics and advanced analytics can be omitted. The trick is to discern what we need from what we want. The current generation of business analysts and business quants are up to the technical challenges, and they have made incredible breakthroughs. For example, applying predictive models to banking has built more intelligent banks, which is contrasted by the fatal opinion-based decisions and sloppy analyses involved in the financial meltdown of 2007–2008. Also, today’s statistical software has evolved in efficiency and capabilities. Finally, for most corporations, IT has matured and can inexpensively provide the data. We have the talent, we have the software, and the data is overflowing. Section 1.2 The Shape of Things to Come— Chapter Summaries The corporate pacemaker has quickened and analytics is wanted to speed up and improve decisions. The ambitions of this book are to provide insight into how analytics can be improved within the corporation, and to address the major opportunities for corporations to better leverage analytics. PART I The Strategic Landscape—Chapters 1 to 6 Part I discusses the infrastructure needed to fully leverage analytics in the corporation. We will discuss changes in corporate culture, personnel, organization, leadership, and planning. Chapter 2, “Inside the Corporation,” discusses analytics inside the corporation based upon experience from both successes and failures. Section 2.1 discusses how corporations employ a Hierarchical Management Offense (HMO), which centralizes authority and decision-making. We will discuss how the right calibration of Leadership, Specialization, Delegation, and Incentives can nurture analytics. We outline the typical leaders who support analytics. We note that advanced analytics is a specialization and discuss the implications of this in a corporate environment. We review good delegation practices, pointing out that more authority and decision making must be delegated to those close to the tacit information. Analytics is a team sport, best encouraged in a meritocracy with team incentives in place. Section 2.2 provides notorious examples of failure due to the sloppy implementation of analytics. We review failures at Fannie Mae, AIG, Moody’s, Standard & Poor’s, the pharmaceutical industry, among others. Section 2.3 provides examples of triumphs in statistics. These include a success story in reviewing predictive analytics at The Associates/Citi and predicting fraud at PricewaterhouseCoopers. Chapter 3, “Decisions, Decisions,” underscores the importance of leveraging the facts. It notes the schism between opinion-based and factbased decision making. Section 3.1 discusses how corporations make decisions and how they incorporate data analysis into their decision making —that is, analytics-based decision making. It clarifies the need for both industry knowledge and analytics expertise. Section 3.2 breaks down the process of integrating the data analysis into the analytics-based decision or action. Autopsies have revealed where the mistakes occur, and we will discuss the interplay between industry knowledge and analytics. Section 3.3 discusses a long list of decision impairments, which distract us from appropriately leveraging the facts. Chapter 4, “Analytics-Driven Culture,” discusses the contents of corporate cultures that succeed in leveraging analytics. It clarifies that analytics is transferrable across all industries.13 Section 4.1 discusses what is involved in an analytics-driven corporate culture and how such cultures arise. Section 4.2 helps us to better think about blending analytics and industry expertise. It also illustrates that corporations tend to understate analytics in that blend. Chapter 5, “Organization: The People Side of the Equation,” discusses the composition (Section 5.1), structure (Section 5.2), leadership (Section 5.3), and location (Section 5.4) of analytics teams within the corporation. We note the difference between management and leadership as illustrated by Warren Bennis in his book On Becoming a Leader. Chapter 6, “Developing Competitive Advantage,” is the lynchpin of this book. It discusses how to assess a corporation’s analytics needs (Section 6.1) and evaluate its prowess (Section 6.2). In Section 6.1, we outline how to assess the analytics needs of the corporation and translate that into a strategic analytics plan. This plan will clarify the corporation’s needs on an annual basis. Next, in Section 6.2, we lead the reader through evaluating the analytics capabilities of the corporation. The difference between the needs and capabilities is the gap to be addressed. Section 6.3 discusses aggressive measures for pursuing the wanted analytics capabilities. PART II Statistical QDR: Three Pillars for Best Statistical Practice—Chapters 7 to 9 PART II of this book introduces Statistical QDR—the three pillars for Best Statistical Practice. These pillars—Statistical Qualifications (Chapter 7), Statistical Diagnostics (Chapter 8), and Statistical Review (Chapter 9)— enable the corporation to measure the quality of the analytics-based decisions and the data analyses. This is the methodology behind Best Statistical Practice. These tools create the momentum for continually improving the analytics-based decisions and analytics, and they measure our performance in delivering the same. In short, they allow us to “fly on instruments” in poor visibility.14 At least one analytics practitioner should be responsible for overseeing and continually improving each of these pillars. Chapter 7, “Statistical Qualifications,” discusses the qualifications necessary to be competent in making analytics-based decisions and performing advanced analytics—including those qualifications needed for reviewers of this work. Section 7.1 reinforces the idea that leadership and communication skills are an essential part of performing analytics. Section 7.2 discusses the needs and training for more sophisticated decision makers and presents the training required for digesting statistical results. Section 7.3 discusses the advantages of applied statistical training. The delay in certifying statisticians for so many decades has facilitated charlatanism and a credibility problem. Section 7.4 makes the case for certifying those who are qualified to analyze your data. Chapter 8, “Statistical Diagnostics,” discusses the Statistical Diagnostics that business analysts and business quants should apply and decision makers should recognize. Here we list the usual suspects and focus on a few effective techniques. Section 8.1 outlines the various Statistical Diagnostics needed for pursuing success. Section 8.2 discusses applying multiple solutions to solve the same business analytics problem. Section 8.3 discusses the family of Data Splitting techniques, whereby we partition the data into development datasets and validation datasets—the latter are also called control or hold-out datasets. Chapter 9, “Statistical Review—Act V,” discusses what is involved in reviewing analytics-based decisions and data analyses. Section 9.1 discusses the considerations going into the purpose and scope of the review. Section 9.2 discusses the nuances of reviewing the analytics-based decisions and the data analyses. PART III Data CSM: Three Building Blocks for Supporting Analytics—Chapters 10 to 12 The transition toward an analytics-driven culture requires a number of infrastructural changes. PART III discusses the three usual soft spots that, when poorly managed, hold corporations back. Every analytics professional will recognize the importance of these three building blocks: Data Collection (Chapter 10), Data Software (Chapter 11), and Data Management (Chapter 12)—Data CSM. However, time after time corporations fail to adequately cover these areas. At least one analytics professional should be responsible for overseeing and continually improving each of them. We will clarify what is getting overlooked and dispel the usual myths. Chapter 10, “Data Collection,” discusses “the matter with” data collection. Most corporations have weak data collection abilities. They rely upon the data to find them. We will discuss the application of Design of Samples (DoS); Design of Experiments (DoE); and simulation, and juxtapose the characteristics of these techniques with those of observational, censual, and anecdotal data. Section 10.1 discusses analysis of observational or censual data—the context for data mining, where the data tend to find us. Section 10.2 discusses anecdotal means of collecting information. Section 10.3 discusses the advantages of randomly selecting a representative subset from a population—DoS. Section 10.4 discusses the advantages of randomly assigning treatments (or factors) to a representative subset from a population—DoE. Chapter 11, “Data Software,” communicates the advantages of a complementary suite of data processing and analysis software tools. Section 11.1 discusses the criteria we consider for designing a suite of software tools for manipulating data. It clarifies the importance of software breadth and emphasizes using the right tool to solve the right problem. Section 11.2 discusses the productivity benefits of automated software. Chapter 12, “Data Management,” closes the book with a discussion about what all analytics professionals need to know about organizing and maintaining the data. Datasets are corporate assets and need to be managed to full effect. Section 12.1 discusses the usual data-consumer needs that corporations overlook. Section 12.2 presents a number of database enhancements that will make the data a more valuable asset. Although these chapters build upon each other, the interested reader might skip ahead to those chapters most relevant to their needs. Chapters 2 – 4 are burdened by providing support for the more impactful later chapters. Notes 1. “3D Data Management: Controlling Data Volume, Velocity and Variety” by Douglas, Laney. Gartner. Retrieved 6 February 2001, and “The Importance of ‘Big Data’: A Definition” by Douglas, Laney. Gartner. Retrieved 21 June 2012. 2. In some situations, the winner is the first corporation to learn just enough from the data. 3. “The Trials of Henry Kissinger” (2003). 4. To name a few: Competing on Analytics by Harris and Davenport; Super Crunchers by Ian Ayres; Data Driven by Thomas Redman, and; The Deciding Factor by Rosenberger, Nash, and Graham; and Business Analytics For Managers by Laursen & Thorlund. 5. Today’s “Big Data” was unimaginable ten years ago. We expect tomorrow’s datasets to be even more complicated. 6. There are many definitions of Business Intelligence; while less popular, this one is convenient for our purposes. 7. Oh, our Stone-Age brains. Our brains have not evolved a great deal during the last hundreds of thousands of years. 8. See “The Man Who Crashed the World,” Vanity Fair, August 2009. 9. We will use the term “statistical” slightly more often because we want to keep in mind the uncertainty and the inherent unreliability of data. 10. We do not need to always know exactly how every decision or analysis will fail. In many situations, it is sufficient to know what works and under what circumstances it works. 11. Like in a court case where each side starts with a conclusion and works backward—that being the appropriate direction. 12. When analytics is making things more complex, then we are doing it wrong. 13. In statistician-speak, statistics, mathematics, and algorithmic software are invariate to industry. 14. A side benefit is that these tools expose charlatans, or alternatively, force them to work harder to fool us. 2 Inside the Corporation “There is one rule for the industrialist and that is: Make the best quality of goods possible at the lowest cost possible, paying the highest wages possible.” —Henry Ford A corporation is an association of individuals—share holders, embodying their private financial interests, yet possessing distinct powers and liabilities independent of its members. It can be a “legal person”1 with the right to litigate, hold assets, hire agents, sign contracts, etc. Over the years, corporations have needed to adapt to changing technology. To keep up with the Information Age, their assets have shifted toward intellectual property, company know-how, and more specialized knowledge-based professionals. The promise of business analytics will require greater changes. We will never fully leverage business analytics without changing the corporate infrastructure—culture, leadership, organization, and planning!2 In this chapter, we address some characteristics of corporations that affect how well they can leverage analytics. We discuss the role of analytics inside the corporation. In the last two sections, we share a number of failures and successes in applying business analytics. Section 2.1 Analytics in the Traditional Hierarchical Management Offense “I didn’t dictate ever because I really felt that creativity doesn’t come from dictation, it comes from emancipation.” —Pen Densham3 “’Politics’ comes from the Greek root poly meaning many and ticks meaning blood sucking parasites.” —The Smothers Brothers The Hierarchical Management Offense (HMO) centralizes power and decision making. It is characterized by a vertical reporting structure serving as “ductwork,” dispensing directives downward and vacuuming information upward. The speed and accuracy of communications moving up and down depends on the length and quality of the vertical chains of relationships. More hierarchy means that politics can have a greater impact on analytics ... and everything else. Leadership, Specialization, Delegation, and Incentives are pivot points for calibrating the emphasis placed upon analytics. Leadership that embraces analytics-based decision making produces better decisions. Specialization facilitates more efficient and effective analytics. Delegating decisions moves the decision closer to the tacit information and expertise. Aligned Incentive structures encourage the most productive behavior. These pivot points facilitate some immediate adjustments to the corporate culture (see Chapter 4), which can increase the productivity of knowledge-based professionals. During the progression of the Information Age, we have seen dramatic growth in IT to keep pace. Most corporations have built large, efficient data warehouses. One expectation is that the next phase will focus on better leveraging this information—this investment. This will involve a new Information Renaissance, using business analytics to make smarter analytics-based decisions. The role of analytics inside the corporation will need to be redefined and expanded. It would be easier if corporations could enhance their business analytics capabilities while changing nothing about their current business model. They would prefer to alter analytics so that it will fit their approach. They want analytics to sell in a sales culture, to manufacture in a manufacturing culture, and to build things in an engineering culture. This is reasonable up to a point. However, facilitating analytics requires change; if only because it is intertwined with the decision-making process. Complete rigidity against adapting the corporate structure will dilute the value of analytics. “General, where is your division?” —General Nathan Shanks Evans “Dead on the field.” —General John Bell Hood Leadership and Analytics To succeed in applying analytics, leadership must correctly judge the merits of analytics and how to best integrate this information into corporate decision making. There are a number of leadership roles that enhance or retard a corporation’s analytical capabilities. We will describe five general leadership roles: Enterprise-Wide Advocates, Mid-Level Advocates, Ordinary Managers of Analytics, Expert Leaders, and On-Topic Business Analytics Leaders. The first two roles are advocates of analytics; they are investors in the technology. The remaining three roles direct those performing the data analysis. We find that leaders vary dramatically in the degree to which they encourage analytics. Those most enthusiastic are likely to have a history of successfully leveraging analytics—data junkies. Some lead with their own analytics-based decision making. Such a background makes it more likely that they will push the company to the next plateau in applying analytics. Enterprise-Wide Advocates put forth the corporate vision and find the resources to make it happen. The formal name of the Enterprise-Wide Advocates is up for grabs. The ubiquitous CIOs are in the running. The less common Chief Economists would be appropriate leaders. Also, there are burgeoning new roles, such as Chief Analytics Officer or Chief Statistical Officer. In Section 5.3, we will discuss the leadership of an enterprise-wide analytics group. Enterprise-Wide Advocates are in a position to: 1. Promote examples of applying analytics-based decision-making (Chapter 3)—thus, building an analytics-based or data-driven culture (Chapter 4). 2. Take an interest in the analytics team’s organization (Chapter 5). 3. Embrace a corporate business analytics plan and make certain that corporate capabilities are evaluated (Chapter 6). 4. Insist that important analyses be performed by professionals with Statistical Qualifications, using Statistical Diagnostics, and with Statistical Review (Chapters 7 to 9). 5. Build and maintain the Data Collection, Data Software, and Data Management infrastructure (Chapters 10 to 12). 6. Remove conflicts of interest and encourage objective analysis, which might or might not fit preconceived conclusions. 7. Select like-minded mid-level managers—shrewdly. 8. “Manage a meritocracy,” as mentioned in Competing on Analytics.4 9. Spread breakthroughs in statistical practice across the entire corporation. 10. Ensure one source of the facts, different corporate units are entitled to their own opinions just not their own facts. 11. Set the tone as to the value of analytics. Mid-Level Advocates are critical for projecting analytics into the appropriate areas of the business—putting the corporate vision in motion. They can 1. Embrace and advocate analytics-based decision making as the way we do business (Chapter 3)—thus, affirming an analytics-driven culture (Chapter 4). 2. Take an interest in the analytics team’s organization (Chapter 5). 3. Embrace a corporate business analytics plan and make certain that corporate capabilities are evaluated (Chapter 6). 4. Insist that important analyses be performed by professionals with Statistical Qualifications, using Statistical Diagnostics, and with Statistical Review (Chapters 7 to 9). 5. Build and maintain the Data Collection, Data Software; and Data Management infrastructure (Chapter 10 to 12). 6. Uphold the meritocracy. 7. Increase the involvement of analytics professionals. 8. Recognize and reward training. 9. Recognize statistical analysis as intellectual property. 10. Quell resistance to analytics. Typically, when a corporation has an Enterprise-Wide Advocate, it will have or find Mid-Level Advocates. This complete structure does the most to integrate analytics into the business.5 If a corporation lacks an EnterpriseWide Advocate but possesses a Mid-Level Advocate, then there will be a pocket of analytics behind them.6 This pocket will have markedly less impact throughout the company. Directors of those performing data analysis (business analysts and business quants) fall within a spectrum of management and leadership skills combined with analytics competence (Section 5.3). We will discuss three roles in this book: Ordinary Managers of Analytics, Expert Leaders, and On-Topic Business Analytics Leaders. We define the Ordinary Managers of Analytics as those with the authority to direct analytics resources, yet who possess less training in business analytics than those who perform it. An Expert Leader is someone with the training and experience to lead analytics, yet less leadership authority. Finally, the On-Topic Business Analytics Leader has the authority, training, and experience—a triple threat. These three roles are charged with anticipating the information needs of decision makers and building an infrastructure that can meet these needs on a timely basis. Corporations have schedules and must make and remake decisions based upon whatever information is available. The Ordinary Managers of Analytics tend to be less engaged in the analytics. The concerns are that they will think about the business from a perspective that is too light on analytics and that they will miss critical opportunities. These managers must delegate shrewdly in order to be successful in analytics. Most of them will spend a great deal of time managing up7—this is probably more comfortable for them. We are concerned that they will not spend enough effort leading the analytics practitioners because they might not be as comfortable with that aspect of the role. Next, we consider an informal leadership role—the Expert Leader. We define an Expert Leader as someone regarded as knowledgeable of the business, competent in analytics, and possessing leadership skills. This makes this person “bilingual”8—quant and business. They comprehend the specialization. They can review an analysis; find mistakes or weak points; and construe its reliability. A corporation can have several Expert Leaders. They possess business analytics expertise, yet with less formal people management authority. They are sometimes informally “chosen” by the other analytical professionals to boost the leadership and to fill a void as a spokesperson or decision maker. They support the other analytical professionals, and they maintain the integrity of the science. By granting more formal leadership authority to an Expert Leader, we can derive: Business Analytics Leader9 = Expert Leader + Formal Authority This is a bilingual role with sufficient formal authority and business analytics expertise. Expert Leaders and Business Analytics Leaders are necessarily trained on the topic of analytics. They can better identify talent and judge results. They understand “best practices” and can skillfully lead a team of practitioners. It is not just about technical ability; it is the way they think. They can think more statistically about the business problem. They have greater appreciation for getting the numbers right and they create less burden on the other analytics professionals on their team. These skilled leaders are usually less politically astute—a trade-off. We will discuss these three roles further in Section 5.3. Specialization Specializations facilitate hyper-productivity in the corporation; statistics is a peculiar specialization. Ordinarily the benefits due to analytics are easy to quantify. We can measure an increase in sales, the lift due to a scoring strategy, or a decrease in risk. However, there are situations where the benefits are difficult to measure, difficult to trace, and difficult to claim. It takes analytics ability to measure and trace the benefits, and it takes political sway to claim the credit due. Statistics can produce modest returns for months and then unexpectedly revolutionize the business during a single day—the serendipity of statistics. Many analytics professionals are passionate about pushing the business forward. In addition to producing facts, statistical training facilitates a “scientific” approach to perceiving the business problem. It accelerates the search for solutions, which are yet to be revealed through the trial and error approach that produced the industry knowledge of the past. Corporations invest in any specialization relative to its perceived value. Estimating the future value of analytics requires foresight integrated with an understanding of analytics. For less analytical corporations, the potential of analytics is often undervalued because of missed opportunities, which have prevented it from providing value.10 Certification for quants is nonexistent in some countries and is just beginning in others, so corporations struggle to judge qualifications. Hence, it can be a challenge for them to discern the reliability of the results. The benefits due to analytics are a function of the value of the data, the technical capabilities, the shrewdness of the applications, and the degree to which the analytics team is resourced.11 In practice, many corporations ring-fence resources (retain resources earmarked for a particular corporate need) based upon their competitors’ resourcing and advice from consultants. There is no complicated economic calculation. “Analytically based actions usually require a close, trusting relationship between analyst and decision maker ...” —Davenport and Harris12 “One important dictum is to make decisions at the lowest level possible.” —Thomas Redman13 Delegating Decisions Delegation is an important characteristic of HMO. In general, leadership needs to delegate decision making toward those who are in the best position to make the decision. A single corporate-wide decision maker is unlikely to have the most complete knowledge. Furthermore, leadership needs to delegate the execution of analytics to analytics professionals, who have “practiced.” Effective delegation requires trusting relationships. There is a burden on the leadership to build strong relationships with their analytics practitioners. Delegating decision making moves the decision closer to the tacit information and expertise. Most corporations tend to involve too few decision makers.14 Dispersing the decision-making burden fosters a smarter and less autocratic15 corporation. Involving more qualified decision makers implies greater engagement and elicits higher quality decisions. Analytics is subject to nuances that cannot be easily explained. Decisions based upon advanced analytics require more sophisticated decision makers and that the decision makers possess greater familiarity with the facts. We need decision makers who are themselves analytics professionals and can (1) trust the analytics, (2) understand analytics, or (3) build relationships with other analytics professionals and recognize analytics qualifications. Delegating analytics moves the execution closer to the training and experience. This generates dividends by getting things done faster, cheaper, and better. Part of the speed is in avoiding unnecessary or poor analysis. For specialized problems like analytics, the most successful approach for “offtopic leadership”16 continues to be straightforward: Delegate analytics to the specialists. Leadership with on-topic training has serious advantages in delegating to those performing analytics: 1. They can better predict and motivate timeliness and accuracy. 2. They can trust techniques. 3. They are in a position to delegate what they understand to experts, who they can understand and trust. 4. They can better communicate with other analytical professionals. Analytics is a way of thinking. 5. They just do not need as much time to make competent decisions about analytics. 6. They understand that the quants engage in “meatball surgery.” The focus is on the critical aspects of the analysis. 7. They recognize that specialists are not interchangeable parts. Analytics does not all look the same to them. Hence, they can match the right task to the right expert’s competencies. This enables specialization within specialization—the leap from a vertical organization to a horizontal one (Section 5.2). They do not require that everyone be an interchangeable part in order to simplify their leadership role. 8. It is empowering for the quants to work with analytically enthusiastic leadership. There are tricks for delegating to all types of specialists. Here are some tips intended to help: 1. All leaders should review and evaluate the results of the assignment. Ordinarily the means used to accomplish the task are less relevant. However, in the case of analytics, the means are an integral part of the results. Managers are responsible for making sure that both the process and the outcome of the delegated task are consistent with the goals. 2. The idea is to retain responsibility while delegating authority and accountability. The analytics professional knows what needs to be done and how to do it, and only needs the opportunity to do it. Delegate the freedom to make decisions and the authority to implement them. Managers should communicate to all individuals affected by the project that it has been delegated and who has the authority to complete the work. 3. Managers should discuss with the analytics professionals what resources they need for a task and then empower them to secure those resources. 4. Good leaders allow employees to participate in the delegation process. 5. If we are concerned that the project will take too long, then include the deadline as part of the problem to be solved. On-topic leaders are better at communicating this point. 6. If we are concerned about over analysis, then we set minimum accuracy targets as part of the problem to be solved. This trick burdens the analytics professionals with stopping unproductive data analysis. We should keep trying to improve solutions yet let the experts discard useless misleading analysis. 7. Even on-topic leaders should avoid the tendency to intervene simply due to style differences. Corporations with the advanced “power” to delegate have the advantage.17 Incentives Corporations run on their incentives. With the proper incentives, corporations can become highly efficient. Incentives are best when they are aligned with solving the problems. In the U.S., corporate incentive structures have steepened in the past few decades. About 30 years ago, U.S. CEOs received approximately 30 times the average employee’s compensation. This multiple has since exceeded 340 times. As individualist incentives steepen, this encourages much more individualism and eventually sociopathic18 behavior. This tampers with team cohesion and creates horizontal and vertical rifts in the corporation. Analytics, just like innovation, thrives in a meritocracy with team incentives.19 Dysfunctional or misaligned incentives will lead a corporation toward destruction. They can make it hazardous to do the right thing for the company. The senior management of Bear Stearns had their bonuses aligned to high-risk behavior.20 This encouraged their fatal mistake of getting over extended and trapped in a liquidity squeeze.21 Incentives for leadership roles need to be long-term, and there should be some team incentives. The concern with excessive individualistic incentives is that they encourage everyone to place their self-interests ahead of the corporation, creating more politics as employees vie for lottery prizes. We think analytics needs more teamwork and that usually individual incentives do little to motivate struggling employees. Those employees who are here for the wrong reasons are difficult to incent. Complex undertakings, like some analytical projects involving large parts of the corporation, can be more efficient with team incentives. A corporation has an incentive problem when it is hazardous or at least not in an individual’s best interests to solve the statistics problem appropriately. Section 2.2 Corporate Analytics Failures— Shakespearean Comedy of Statistical Errors “Safety is not a lucky system. It’s a system of science, analysis, and facts.” —Mark Rosenker, Chairman, U. S. National Transportation Safety Board22 “Data analysis is an aid to thinking and not a replacement for it.” —Richard Shillington “It’s easy to lie with statistics. But it is easier to lie without them.” —Frederick Mosteller Statistical malfeasance is one of today’s corporate diseases. Several corporations have gone bankrupt, missed breakthrough opportunities, and taken big losses because of statistical mistakes. These mishaps go undiagnosed even after an “autopsy.” Sometimes it is difficult to trace business mistakes back to absent or faulty data analysis, just as it is difficult for corporations to measure the quality of the data analysis. The solution to difficult decisions is not riverboat gambling. It is to measure the quality of the information and then interpret the facts. Chapters 7 to 9, Statistical QDR (Qualifications Diagnostics Review), will cover the means by which to measure the quality of the facts, including certification for quants.23 If we cannot measure analytics quality and we are unable to reasonably confirm that there are no problems with our analytics and our analytics-based decisions, then those are our problems. For the remainder of this section, we will share accounts of corporate failures due to poor analytics practice. The Financial Meltdown of 2007–2008: Failures in Analytics Financial corporations are by necessity the most analytically savvy in the global economy. Part of the banking cycle includes a financial crisis that culls the weakest. The tinderbox that facilitated the financial meltdown of 2007–2008 comprises a web of decisions that were based upon invalid assumptions and faulty analyses. Once the housing bubble reached a certain size, there was no gentle recourse and the bubble was popped by additional mistakes in analytics made by the first victims. Banks, credit rating agencies, and investors struggled to price the risk of subprime assets, which are notorious for destroying banks.24 Those highly leveraged banks with the worst algebra tended to lose the most. The subprime melodrama went as follows. First, in 1999 the U.S. Congress repealed the to Glass-Steagall Act of 1933, which protected the economy from banks becoming too big to fail. In time, investment banks were allowed to raise their leverage—the amount they owe versus their cash on hand, to obscene ratios. This increased their ROAs (returns on assets) and made them too soft to experience significant financial stress. Next, there was a glut of money looking for high-return AAA investments.25 Investment banks pooled mortgages into MBSs (mortgage-backed securities) and CDOs (credit debt obligations) for sale on this market. They took their CDOs to the rating agencies, who rated the riskiness of these investments. Then the banks sold them to these investors. This generated huge profits for the banks, who demanded more mortgages. Money was cheap. Housing prices rose. As time progressed, the supply of prime mortgages shrank and was outpaced by demand for CDOs. Exotic mortgages appeared to keep the cycle going. This facilitated a growth in subprime mortgage. People who ordinarily could not obtain credit for a single house were buying multiple houses. Real estate investors were also buying a large portion of the new purchases. Housing prices continued to rise at unsustainable rates. It was clear that this was a shift in the economy. In order to market these CDOs, the banks split them into tranches based upon riskiness. They sold the least risky tranches and tended to retain the highest-risk mortgages. Through magical accounting tricks, they put these off of their balance sheet. In order to cover the risk, most of the investment banks purchased credit default swaps (CDSs) from AIG as an insurance policy against the worst that could happen. This was their “originate to distribute” model, which was supposed to generate fees and distribute all of the risk. Fitch Ratings, Moody’s, and Standard & Poor’s They were the “arbitrators of value” for these CDOs. They received lucrative fees from the investment banks for the service of “objectively” rating these CDOs. An investment bank would solicit multiple ratings prior to selling the financial investments. The companies with the top two ratings would be chosen, and they would be paid. Any other rating agency received nothing. According to The Big Short, the rating agencies did not always have their own models or complete granular data. Instead, they relied upon models provided by the investment banks and pooled data for any supporting analysis. Through some strange alchemy, subprime mortgages were turned into AAA-rated investments. Countrywide Bank, Golden West, and Washington Mutual To maintain the pace in loan originations, shadow banks dealt in exotic mortgages for consumers lacking the usual credit worthiness. These included “no money down, interest only” loans that would balloon in payments. These loans made sense only in an economy where housing prices would continue to rise. However, the increase in housing prices was unsustainable. Housing prices rose by 124% from 1997 to 2006. The business was so lucrative that these shadow banks failed to heed the mounting warnings from their risk models. AIG Financial Products AIG was a AAA-rated corporation with a small Financial Products group that basically insured the risk for large blocks of debt. They unwittingly amassed a vast portfolio of risky CDSs, which essentially were insurance policies on mortgage-related securities. Their experience illustrates the classic risk involved in directing analytics with an ordinary off-topic manager of analytics. Years before the bubble burst, AIG changed its management of the Financial Products group. The incoming manager did not have the ontopic training in mathematics, statistics, and algorithms. This led to a cultural change from vigorous discussions about how well their models were performing toward apathy. The new head of FP managed up, and for an analytics group, this is mismanagement. During this time, AIG took on incredible risk without realizing it.26 At the onset, they were insuring tranches that contained 2% subprime mortgage, and before they realized it, they had grown this proportion to 95%.27 Fannie Mae: Next to the Bomb Blast Although Fannie Mae and Freddie Mac, the two main governmentsponsored enterprises (GSEs), continue to haunt the American financial scene, we should remember that they were not the first monoline mortgage companies to fail; they were among the last. Countrywide, Washington Mutual, and a number of others failed first. Golden West, a thrift in California, was so toxic that it killed off its buyer, Wachovia, one of the most widely respected national banks in both commercial and regulatory circles, forcing its sale to Wells Fargo. Of course, nonmortgage businesses, such as the investment banks Lehman Bros. and Bear Stearns, also failed. These banks generated enormous quantities of bad mortgage loans and failed to fully implement their “originate to distribute” model, which would have distributed 100% of their risk. In order to understand what blew up the mortgage market in 2007– 2008 (and the problems continuing to the present day), we have to go back in time. Although I’ve worked in a number of environments, my perspective is fundamentally based on the years I spent regulating multiple types of risk across a wide variety of banks at the Treasury Department. This experience gave me a strong appreciation for the logic and analytics of risk management, but without the level of intimidation often felt by non-quants around folks with Ph.D.s in things like statistics, economics (econometrics), operations research, and others. These experiences included excoriating whole classes of models that purported to be precise and statistical, but created so-called “reliability” indices with no underlying statistics on their actual reliability—to the delight of users and the chagrin of salespeople. It included catching quants at major institutions faking a model validation. My experience taught me to trust my own doubts, above all, and to always question, with the most humble of attitude, the most lettered of businesspeople. By the time I got to one of the mortgage GSEs, I had lost my youthful arrogance, but also my youthful awe. And I began asking questions— like a financial Colombo. When the dot-com bubble burst in 2001, the Fed lowered interest rates substantially. This caused a refinancing boom, which, as had happened many times before, drove mortgage banks to greatly expand their staffing to handle the temporary rate-driven increase in volume. As rates hit bottom and stayed there, the “refi boom” consumed most of the traditional borrowers (credit-worthy non-investors) and began to tail off. It was at this point that “exotic” mortgages began to be engineered as a tactic to keep the level of mortgage originations at an artificially elevated level. Many of these exotic mortgages were neg-am (negative amortization), characterized by low down payments and low monthly payments—for a limited time. A large portion were “zero down, interest only” loans. At the same time, corruption blossomed in the real estate industry, and a significant portion of housing purchases were driven by investors who, for “zero money down,” could buy a call option on the housing market. For the most part, these neg-am mortgage products were not actually new; they were simply rediscovered by an industry that had little data covering mortgage performance prior to 1995. Of course, anyone who had been paying attention in the 1990s should have remembered28 that the neg-am feature, which played a critical role in the new designs, had also contributed to the hole in Citibank’s balance sheet in the early 1990s. There were two differences between the 1990s and the 2000s. First, in the 2000s the commercial banks played a trailing, rather than a leading, role. They caught up to the “leaders,” namely the mortgage companies and their investment banking partners, only after the latter had seemingly proven the concept, with several years of low losses and high profits. Second, the recession of 1989–1992 shut down the first neg-am experiment before it could grow too large. Whereas in the 2000s, the neg-am experiment was protracted by Wall Street redirecting investment capital from the stock market. This capital came from investors seeking moderate-yield investment-grade bonds. Instead, they would receive new types of sub rosa junk bonds. Throughout the 1990s and into the early 2000s, there were two major groups in (first) mortgages: (1) conventional and government (i.e., FHA or VA-insured), and (2) conforming and jumbo (determined by size relative to GSE loan limits). Before the boom in exotics, the GSEs and FHA dominated the smaller loan product market, with the banks and investment banks dominating the rest, albeit at somewhat higher rates and stricter terms. When exotic mortgages were first introduced in the 2000s, commercial banks did not enthusiastically receive them. They tended to focus on building portfolios of loans and servicing assets, and had relatively strong risk-based federal regulation. It was the mortgage specialists, some non-bank mortgage originators and thrifts, who partnered with investment banks to devise these exotic negatively amortizing loans to save the mostly non-bank mortgage companies from the typical cyclical downturn and mass layoffs following a refi wave. Of course, the investment banks were also eager to bite off as much of the GSE/Govi market as they could—a market that for years they had been complaining was dominated by institutions with an unfair advantage. A confluence of several factors lead to their enthusiasm for this exotic new product, including the following: 1. Lack of publicly available data that would have underscored the poor performance of previous experiments in neg-am mortgages 2. Faith that the “originate-to-distribute” model would really allow the institutions to remove all credit risk from their books 3. A fundamental belief that, because they are backed by real property, mortgages present minimal to no risk 4. A belief that the historical lack of a national fall in home prices since the Depression meant that home prices would never fall nationwide 5. Greed and envy directed toward the sheer scale and profitability of the GSEs and a desire to take a piece of it 6. Demand for new investment classes as the dot-com crash discredited equities generally 7. An overly optimistic faith in the ability of financial engineering to add significant value So, what likely happened at investment banks is what I witnessed at Fannie Mae, when it belatedly began approving exotic mortgage products for purchase through standard channels in 2005 (it was already buying securitizations). When asked if he had built a model to estimate the credit guarantee-fee for one of these negatively amortizing products, a credit modeler, whom we’ll call Jim since he still works there, said to the SVP in charge: Yes, I’ve built a model to calculate the credit guarantee-fee, given the information we have. However, we have no historical data on this product, so we’ve had to make a lot of very questionable assumptions for the model.29 Of course, by the time this discussion occurred, these new mortgages had already reduced the GSE’s market share by 50%, and management was concerned with keeping the agencies “relevant” in this “new world.” Jim’s final comment was, “Rather than rely on this very approximate pricing, what you should ask yourself is, would you want your son to buy a house with this mortgage.” This idea was ignored by our senior management, and in their quest to be relevant, they began originating exotics. Although I haven’t had the opportunity to analyze the GSEs’ actual losses, I believe that they were pushed into failure not by their origination of these questionable loans in competition with the investment banks, or even the purchase of pools of asset-backed securities based on these exotic mortgages. Additionally, it is a well- repudiated myth that the requirement that the GSEs invest a minimum amount in loans for low- and moderate-income borrowers pushed them over the edge. While the GSEs suffered significant losses on their exotics, losses on their required low-mod product suffered minimal losses. Despite what their critics have implied, the GSEs never dominated these high-risk markets but always played catch up to the “more nimble” fully private-sector players. A careful dissection of GSE losses will likely show that the bulk of them were due to the simple fact that home prices declined at double-digit percentage rates nationwide. This decline was due to a bubble that was fueled almost entirely by an out-of-control private sector. What killed the GSEs was nothing that they themselves did, although they were guilty of some lapses in judgment. What killed them was the definition of who they were and the waters in which they sailed. As the mortgage banking/Wall Street axis pushed down the firstyear’s monthly payment on mortgages they generated so much additional purchase business that they drove home prices higher—to the point where 10% appreciation per year began to be treated as a new norm. This rate is clearly unsustainable in a world of 3% inflation and 2 to 5% growth, as housing costs would eventually crowd out all other consumption and production. These price increases could only be sustained through a Ponzi “bubble” of neg-am mortgages and house flipping financed by overrated investment vehicles. Underwriters were allowed to calculate qualifying income-to-payment ratios based on the temporarily low payments allowed under these rates—meaning that the ultimate insurance for lenders on the negatively amortizing loan was infinitely rising prices, which could only occur if new, and ever more irrational, buyers were found to pay for increasingly inflated property. Of course, this greater fool was the very same person that the buyers— an enormous number of them motivated by pure speculation—were relying on to protect their minimal investment. It is the collapse of that mechanism that ultimately brought down home prices and with it the GSEs, and severely threatened the financial health of the U.S. and many foreign economies. At this point I have to refer again to Jim. Over the 2007–2008 period, and perhaps before that, Jim would periodically distribute a very startling graph. It was simply the average nationwide price for homes from 1985 to the present, using Fannie Mae’s internal repeat-sales index, in constant dollars (I think it was indexed to 1995). It looked like an artist’s rendering of a tsunami—a little wavy line blending into a massive tsunami towering, preparing to crash. I have reproduced it as well as I can by using publicly available information in Figure 2.1 (this chart uses the national FHFA index and the monthly national CPI to deflate it to January 1991 dollars). Needless to say, the marketing department didn’t much care for Jim. In 2007 he attended an enormous marketing department meeting. At one point he asked a senior manager, in this open forum, what we were going to do when the market crashed. The manager retorted, “Jim, you’ve been saying the market was going to crash for two years. Tell me, when is it going to happen? When?” In less than six months, the overconfident manager’s query was answered and a few months after that, the manager retired. We, in the Economic Capital group, were so taken with Jim’s simple analysis that in the fall of 2007, we developed an alternative credit risk stress tool. The tool generated possible future home price stress paths for use in calculating the potential economic damage of a stress to Fannie Mae. We set just a few simple parameters, such as the depth to which prices might fall measured as a percentage below the previous real low, how long it would take to reach the real price nadir, and how long a recovery would take, as well as the inflation rate that would allow us to translate stress/mean reversion from the real price space back to nominal. This model did not claim to predict the future. What it did was to create apparently plausible stress scenarios, using simple assumptions and obvious logic, that were far worse than those generated by the Fannie Mae’s statistical models at claimed probability levels below 0.1%. Figure 2.1 Housing Price Index (HPI) A talented quant on my team built this simple tool, and we wrote up some documentation. We were listened to with politeness in the spring of 2008, but could not get such a stress tool implemented into the corporate credit modeling infrastructure. Of course, since we were taken over that summer, the model would clearly have done us no good. My point is rather the following: The failure at the GSEs was due to two factors. First, they are mortgage monolines with government charters. They couldn’t diversify their risks, and their executives would only maintain their status—and pay—if they continued to be a major force in the market. Second, and equally important, was the fact that these firms were run primarily by people whose job it was to be optimistic and whose imagination only entertained dreams of greater success—never nightmares of doom, no matter how obvious the analytical evidence. Like many successful organizations, they were dominated too much by marketing, in this case with a heavy dose of government relations. Analytics’ job was to make fine distinctions in value and risk—not to influence strategy. The GSEs missed the opportunity to integrate analytics into their strategy. The only criticism I ever heard of Jim from other, senior members of the company’s analytics community was that he was a “chicken little” who was making no friends. No one ever questioned the relevance or legitimacy of his straightforward analyses. There are only two ways the GSEs could have possibly saved themselves. The first was to simply stop lending in the mid–2000s. This could have been done by recognizing the bubble and pricing themselves completely out of the market. This would have been incredibly brazen, but because of the GSEs’ role in the market, it might well have moved the entire market back to rationality in a way that the federal regulatory authorities did not have the courage to begin until 2007, and might not have been able to accomplish earlier. The second survival path would have been to hedge our credit risk, taking short positions in the relatively illiquid Case–Schiller index, and/or buying credit default swaps on mortgage-backed ABS)—or shorting the ABX index of assetbacked securities. But the faithful don’t hedge. They believe in their business, so it was much easier for the pilots to keep running the ship downwind, following the market, then to take a stand and tack into the headwind. I believe that only a senior management with the ability to imagine tragedy as well as triumph, and an appreciation of risk analytics and the courage to follow it wherever it may lead, can be relied upon to safely steer today’s leveraged risk-taking institutions. The Great Pharmaceutical Sales-Force Arms Race by Tom “T. J.” Scott Many senior leaders are not analytically sophisticated. Some lack even a basic understanding of statistics or scientific methods. As a result, these leaders often rely on gut instinct when making decisions. When leaders use gut instinct, they are relying on long-held beliefs and personal experience. There are two obvious problems with this. First, markets are changing rapidly, making many long-held beliefs untrue. Second, relying on one’s personal experience is like doing important research with a sample size of one. In this case, ideas outside the experience of our personal sample are considered less reliable. In big U.S. pharmaceuticals during the last 15 years sales forces grew to an immense and inefficient size, maybe two to six times larger than necessary. This happened for a host of reasons including: 1. Leaders could not let go of their long-held beliefs. 2. Many of the analysts and consultants that supported them were under-qualified and all too happy to provide senior management with results that matched preconceived thinking. This provided unreliable “analysis.” Meanwhile, there were some of us working in a statistical underground with complex contradictory analysis. Unfortunately, we were unsuccessful in convincing our leadership to act on what we had found. Our leaders could not discern that our analysis was more reliable. It was difficult to convince them, because pharmaceutical leader’s long-held beliefs were formed more than 15 years ago, when sales interactions between doctor and sales representatives were critical. At that time, physicians relied on sales representatives to provide efficacy and safety information. Physicians were trained to diagnose; they had very little pharmacology training. Pharmaceutical sales representatives made up for this lack of pharmacology knowledge, and helped physicians stay current with interactions between them and physicians that were valuable and meaningful. At that time a physician might spend 20 minutes discussing clinical trials and mechanisms of action with a sales rep. Today, however, the typical sales interaction is a 45-second chat about the local sports team while the representative gets a signature for samples. Physicians rely on other sources of information—often managed care providers—to evaluate which drug...