Chapter One
FROM INTUITION TO ALGORITHMS A Brief History of Using Analytics to Make Better Decisions
Late on a November night in 2006, along New York City''s Bruckner Expressway in the South Bronx, a solid azure blue, brightly lit new billboard declared, in a single line of bold white block text:
THE ALGORITHM KILLED JEEVES.
The billboard stood out among the others hawking car dealers, reality TV shows, and sex clubs. Although it wasn''t hard to get the "whodunit?" message, the billboard''s sponsor was a mystery. A quick search, though, revealed that it was Ask.com, the search engine owned by the website conglomerate IAC Search and Media, Inc. Apparently its marketers had decided that a billboard along the Bruckner-the roadway home to the suburbs for the search engine''s target user-would be a good place to announce that the dapper info-butler Jeeves had been dismissed for a better and faster model: the algorithm. The billboard was meant to draw attention to Ask.com''s new and improved website-ranking algorithm called ExpertRank, and to contrast it with archrival Google''s search algorithm, PageRank.
Geeky highway billboards, sponsored by cheeky web search engine marketers, are certainly signs of the times. Mathematical moguls are making vast fortunes by differentiating models that compute complex equations with extraordinary speed and precision. "Once upon a time, the most valuable secret formula in American business was Coca-Cola''s. Today it''s Google''s master algorithm," wrote Randall Stross, author of multiple books on internet-era moguls, in his New York Times column "Digital Domain." An algorithm is a set of mathematically derived instructions to accomplish a defined task. Algorithms running on powerful computer networks are not just a part of the digital revolution; they are spawning a revolution in how business decisions are managed and made.
Of course, the seeds for this revolution-and for the digital technology that enables companies to apply such mathematical rigor to operational decision making-were planted long ago. "Predicting short-term changes or shocks is often a fool''s errand. But forecasting long-term directional change is possible by identifying trends through an analysis of deep history rather than of the shallow past. Even the Internet took more than 30 years to become an overnight phenomenon," writes Ian Davis, chairman and CEO of McKinsey & Company.
Today''s digital data management discipline known as analytics began with the first mainframe computers in the 1950s. In this chapter, we look back over the past sixty years, not because the history of analytics and decision management is so fascinating (though much of it is, as you''ll see), but to show you how far companies have come in using computers and analytics to achieve all of the following goals:
To sort through the enormous amount of data they have about their businesses Which helps them make better decisions about serving their customers Which in turn improves the value they offer their customers as well as their overall profitability
If you share these goals, read on.
The Pioneers of Decision Management
Long before marketers were posting arcane mathematical terms on highway billboards, business pioneers were using math and computers to make better decisions. These business visionaries promoted a union of computing power, powerful equations, and brainpower to achieve business insights from deep and diverse analysis of operational data.
The First Use of Computers to Improve Decision Making
Back in the 1950s, at MIT''s Sloan School of Management, computer scientist Jay Forrester argued that a large corporation is a complex social system far too abstract for human beings to manage effectively without the aid of computers. He asserted that we literally need technology to understand the relationships and interactions among people in big organizations. In 1961, Forrester published Industrial Dynamics, his seminal book on systems dynamics-an analytical, problem-solving methodology he developed that employs computer-based simulations to help managers visualize and understand cause-and-effect relationships in decision making and business processes that would otherwise be invisible and inestimable.
Forrester also used the term mental models to describe how people tend to make decisions based on instinct and interpretation rather than on fact. Forrester believed that management decisions based only on mental models and human judgments are inferior to decisions derived from computer models that can represent complex relationships and predict outcomes that the human mind can''t. In the 1970s, Donnella Meadows, a prot?g? of Forrester''s from MIT, applied his theories of systems dynamics to produce a global model for the Club of Rome that was the basis for the controversial book Limits to Growth, which predicted all of the long-term trends in population growth, economics, and the state of the earth''s environment that have since come to pass. Another Forrester prot?g?, Peter Senge, popularized systems dynamics in a management context with his book The Fifth Discipline. Decision management arises from the same notions of systems complexity.
Whereas Forrester advocated for more computer-guided management of business systems in the 1950s and early ''60s in Cambridge, the International Business Machine Corporation-now known simply as IBM-was making its transition from punch card processors to electronic computers. Thomas J. Watson, Jr., bet the company''s future on the thinking machines his father had dismissed as too expensive and unreliable. Taking charge in 1952, the younger Watson hired hundreds of electrical engineers to start designing the first mainframe computers. Little did he know that this decision to commit IBM''s business machine vision to computers would kick-start the information technology revolution in business and the beginnings of decision management in large corporations.
Fair Isaac''s Formative Days with Decisions Management
At about the same time, in California, two young process management scientists-William R. Fair, an engineer, and Earl J. Isaac, a mathematician-were starting their careers in the new field of operations research. Then, as now, operations research involved applying advanced mathematics and statistics using computers to analyze complex operational business processes to improve the process through better decisions. Bill and Earl met in 1953 at the Stanford Research Institute (SRI), a think tank that primarily did operations research for the military. Bill and Earl spent their days as operations research scientists at SRI, helping the U.S. Defense Department figure out how to contain the destruction of an atomic bomb. They created elaborate mathematical models to run on SRI''s behemoth computer in order to answer basic questions. Their concern was not how to build missiles and atomic weapons, but how to operate them. How do you carry them? Where do you aim? How close to the target?
Bill had studied engineering at the California Institute of Technology in the 1940s. During World War II, he had been a radar technical representative for Sperry Gyroscope and had served in the Pacific with the Marine Corps. As a civilian, he also applied his engineering skills repairing night fighter radars on aircraft carriers. After the War, Fair finished his schooling at Berkeley and Stanford. Isaac, who studied mathematics at the U.S. Naval Academy and UCLA, had been part of the team that developed the initial programming for one of the first electrical computers-the U.S. Bureau of Standards Electronic Eastern Automatic Computer, otherwise known as SEAC.
As Bill progressed in his career at SRI and his analysis of operational processes for missile systems and atomic weapons, he became convinced the research they were performing for the military could be just as valuable to businesses. Why couldn''t the operational analysis performed for the Defense Department be applied in other contexts, like corporations serving consumer product and service markets? He visualized the corporation as a sensitive machine similar to the radar systems he had repaired during the war. Like Jay Forrester, he believed that managers needed computers and mathematics to solve tough operational problems and to make consistently better managerial choices.
Bill founded SRI''s first nonmilitary operations research practice, and he asked Earl to join his group. It wasn''t long before the independent and ambitious duo decided to leave SRI to form their own consulting business for the private sector. In an interesting turn of fate, Bill had taken half his courses at the business school while working toward a master''s in engineering at Stanford. Combining Bill''s head for business with Earl''s rare computer talents and passion for mathematics, in 1956 they each chipped in $400 to start Fair, Isaac and Company, Inc. According to Fair Isaac lore, Bill and Earl decided to combine their own last names to come up with a name for the company, but they were concerned that "Isaac Fair" sounded like one person and "Fair Isaac" sounded like a used car salesman. As Fair tells the story, they "settled on the lesser of two evils." Bill Fair was among the attendees at the First International Conference on Operations Research in 1957, just after they named their new company.
Bill Fair and Earl Isaac founded Fair Isaac because they believed, as did their business-minded engineering and mathematician peers, that the operational processes of corporations conceal a treasure trove of information to help managers run better companies. For an organization to be the best, its operational management decisions must be methodical and data-driven-not just guided by gut feelings and consensus. Their vision was to create computer-based mathematical tools for use by corporations to sharpen operational decision making and make process management the foundation for achieving consistently better business results. Fair and Isaac knew they could do the math and the analysis. The only glitch was that computing technology was still too primitive, too scarce, and too expensive.
In the 1950s-when men wore hats, not headphones, and computers were the size of a tank-few companies even used computers or would have known what to do with it if they had one. Even Bill and Earl didn''t have their own a computer, so they worked out a time-share deal with the Standard Oil Company of California (today''s Chevron) to use its mainframe during nonpeak evening hours to conduct their research. The SEAC machine Earl had worked with had been a physical monster with a grand total of only 512 words of high-speed memory. Earl contributed to the development of many of the early computer languages, but his thorough grounding in machine language and even in bit programming, along with his natural talent for the subject, gave him an understanding of the nature of the computer that was equaled by few people in the world at that time.
Fair Isaac Takes Off with the U.S. Credit Card Industry
It took the fledgling company almost three years, but in 1958, Fair and Isaac and Earl Follett-a fellow mathematician, alumni of SRI, and Fair Isaac''s first employee-identified consumer credit as a process in which they could put their ideas to work. By the 1960s, as more business operations started to be computerized, and credit cards became an accepted alternative to cash, suddenly it was possible for companies to capture data on customers'' behavior. When people pay cash for goods and services, it is an "anonymous" transaction. The only record of the transaction is the receipt. For the first time, companies could capture transaction-level data on masses of people.
Credit Scoring Drives Better Decisions and Growth in Consumer Lending. Credit card issuers were interested in seeing trends (that is, what people were buying or not buying). They were even more interested in knowing more about how to manage the risks of mass market lending. Fair Isaac invented the credit score to help lenders analyze each applicant''s credit risk while handling many more applications than they ever had before. The credit score was the first big application of analytics for Fair Isaac, and the beginning of what the company today calls decision management.
A few companies had dabbled in business applications of scoring as early as World War II, but none thought of applying it to consumer lending. As operations research experts, Bill and Earl were familiar with statistical analytic techniques such as multivariate analysis and logistic regression. Earl Follet knew how to apply these concepts to managing credit risk. When Fair Isaac''s first credit scoring model was introduced in 1958, it was the first to use the historical data being captured by finance companies to predict a person''s creditworthiness based on their past behavior. The model produced a score, based on analysis of specific sets of numbers related to variables such as a person''s bank balance and payment records. The credit score was a far better predictor of a customer''s ability to pay back a loan than any decision a banker could make on his own, even if he knew the applicant personally.
Using Predictive Analytics to Make Better Decisions About Customers'' Behavior. Predictive analytics is a way to make connections between the past and the future, using historical data to predict future events. Simply put, it''s the study of how what you know at the time you make a decision relates to what you don''t know: what will happen in the future. The credit-scoring model was built based on variables such as these:
Income Bank account balances Outstanding credit Payment history Time with present employer
These variables were vetted as highly predictive of a consumer''s creditworthiness.
Credit scoring models, and the type of predictive analytics Fair Isaac is known for generally, quantify the patterns and relationships among dozens of variables. Every credit application had all the data needed to build the model. A single score could convey the risk associated with a person''s future payment behavior and the person''s risk profile relative to the behavior of many other people. Using mathematics to predict the behavior of masses of consumers was a revolutionary concept when first proposed. Today, credit scoring is a cornerstone of lending processes, and other analytic applications using data on consumer behavior are revolutionizing mass advertising, direct marketing, and customer service-to name a few business processes that are spawning new, creative analytic applications.
Fair Isaac''s first foray into credit scoring, however, took more than a decade to take off. In fact, Fair Isaac didn''t sell a creditscoring system to a bank''s credit card division until 1970. The first general-purpose FICO(r) score was not developed until 1989. It took time for business attitudes and technology to change.
In 1958, Bill and Earl sent letters to about fifty major credit grantors-mainly consumer banks and finance companies-in the United States asking for a meeting to explain credit scoring and its value. Only one institution replied. More often than not, business clients showed little interest in operational insights. All they wanted was to install their first computer and get it running. The idea of the computer as a tool for analytic computation was way ahead of what business people were thinking.
Still, the timing for scoring was right, because it coincided with growth in nonbanking businesses that were offering credit and capturing the data. Early charge cards (which were metal, not plastic) had been around since the 1930s. By the late 1950s, consumer use of cards rather than cash was growing, and metal charge cards were being replaced by the plastic credit cards we use today. Although Fair Isaac''s first credit-scoring system sale was to American Investment-a finance company based in St. Louis, Missouri-banks were initially reluctant to adopt the new credit-scoring approach.
(Continues...)
Excerpted from The Deciding Factorby Larry E. Rosenberger John Nash Copyright © 2009 by Larry E. Rosenberger and John Nash. Excerpted by permission.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.
