Attend this session to learn about a state-of-the-art modeling tool that unlocks the hidden knowledge in your data using ArcView Business Analyst. Learn what scoring science is and the value added it can provide in spatial analysis.
Scoring Science provides high performance statistical tools that can be used by the layman to understand complex business relationships, build optimal models of behavior, profitability, and employ this understanding in ArcView Business Analyst and ArcView. The seamless integration of Scoring Science's modeling tools, Business Analyst's data tools, and ArcView's GIS environment allows you, the user, to easily identify high valued customers and areas, to quickly and efficiently assess the risks of site location, and to optimally allocate limited resources among clients that need assistance. Only the statistics you need, when you need them, where you need them, and in the way you can actually use them.
Executive Summary
Scoring Science puts sophisticated analytical tools in an easy-to-use software module that enables you to create powerful behavioral models directly from your desktop. With Scoring Science, marketing personnel and other decision makers can now carry out the same type of statistical analysis previously affordable by only the largest corporations and firms.
Depending on your company's market or industry, you can use Scoring Science to answer such questions as: Where should you open your next store or clinic? And which of your customers are getting set to transfer their credit card balances to another company? From health care to banking, from sales and marketing to risk management, there is no end to the actions you can predict using Scoring Science.
On its own, Scoring Science would be an extraordinary business tool. Combined with Esri's ArcView Business Analyst, it is a marketing director's dream. Scoring Science unites some of the most effective and robust modeling techniques currently available, with ArcView Business Analyst's vast library of data and unparalleled geographic display. As a result, it is not only easy-to-use, the findings are just as easy to interpret. Scoring Science even includes features that enable you to view alternative scenarios in real time.
What is Modeling
Think about the processes you use every day to make decisions and judgments--given a wealth of past experiences and a variety of observations about the question at hand, you weigh the facts and make a determination. You are, in effect, modeling. Statistical modeling is the standardization and automation of this process, wherein statistical methods are called upon to help identify relationships and make decisions.
In particular, statistical modeling often seeks to explain the relationships between some particular target variable, such as projected response to a marketing campaign, and a group of predictor variables, such as the demographic and psychographic characteristics of the individuals to whom it will be directed. Such an approach replaces the guesswork in the decision-making process with objective analytical methods.
The Modeling Process
Usually, the choice of target variable is determined by the nature of the question one is hoping to answer. For example, consider the case of a healthcare corporation, which must decide where to allocate scarce resources among its numerous hospitals and clinics. Such a resource could be in the form of a medical specialist, an expensive piece of equipment designed to treat a particular disease, or emergency services. This problem could be phrased as the following question: Which neighborhoods served by the healthcare corporation are at the highest risk for a particular disease?
Thus, the required target variable is one that specifies high-risk and low-risk neighborhoods. The modeler's task is to generate this target variable, select relevant predictor variables, determine the training set (i.e. the data used to build the model), apply appropriate statistical methodologies to build the model, verify the efficacy of the model, and implement the model to identify potentially high risk neighborhoods.
In the absence of a modeling tool such as Scoring Science, this process requires a broad range of skills on the part of the modeler. He or she must have both breadth and depth of knowledge regarding available statistical methods, as well as the ability to build advanced techniques into a successful, efficient model. With Scoring Science, however, the layman can produce sophisticated statistical models and scored results without any specialized knowledge.
How it Works
Scoring Science enables you to predict whether an individual will respond in the affirmative or negative to a Yes or No/True or False type question such as whether or not he or she is likely to buy from you again. To do this, it begins with data with a known outcome--for example, the results from a previous sales campaign. This information is used to construct a statistical model that correlates the way in which individuals previously responded with behavioral, psychographic, or demographic characteristics selected from the data available through Esri's ArcView Business Analyst, or from a company's own proprietary database.
As mentioned above, Scoring Science employs some of the most robust and effective modeling techniques available, including multivariate nonlinear regression methods. Furthermore, because the models are based on real data, you can be confident of their accuracy and relevance. The output of this model is then applied to new data, data for which the outcome is unknown, to calculate the probability of each individual in the new data set responding in the affirmative. The results are then automatically ranked, or "scored," according to this probability. In this way companies can quickly and easily identify the top 10 or 20 percent of likely customers, or, similarly, exclude the bottom third.
Furthermore, Scoring Science is not limited to questions of customer behavior. Models can be created to predict aggregate, as well as individual, behaviors by utilizing the census-level data available through Esri's ArcView Business Analyst. In fact, Scoring Science can be used to answer almost any type of question that can be expressed in the form of a Yes or No/True or False type question.
Data Preparation and Pre-Processing
One reason many analytical modules fail to return satisfactory results is their inability to deal with imperfect data--data that includes missing values, categorical variables, predictor values with values out of range, or all of the above. Scoring Science automatically prepares your data for modeling, carefully accounting for missing values and categorical variables, and applying an appropriate scaling function so that all numeric predictor variables are in the same range and can be compared on equal terms.
Definitive Results
Employing the tremendous number-crunching capabilities of today's computers, Scoring Science not only generates precise models - it does so extremely quickly. Unlike other scoring systems, Scoring Science incorporates statistical features designed specifically to generate meaningful starting points for its calculations. It also applies a complex iterative approach to find the best solution for the model in the shortest possible time. Indeed, where others systems can take an entire day or more to return a model, Scoring Science generally returns a finished model in less than a single hour, even for extremely large datasets. Scoring Science even includes a rating function that automatically evaluates the accuracy of the models it creates.
Value Added
Underlying most businesses' core strategy is a desire to identify the most probable cases of some event. That event might be an initial or return purchase by a customer, the repayment of a loan without default, or some other such action. Regardless of the specific case, the ability to assign a probability of success to each potential customer or store location is extremely beneficial. This probability is commonly referred to as a "score" (often computed between 0 and 1), with a higher score implying a higher probability of success. Typically, generating such scores both accurately and efficiently is a highly involved task. It requires the gathering and processing of data, the use of expensive and difficult-to-use software, not to mention the time of a highly trained (and highly paid) analyst.
Scoring Science, an Esri Business Analyst extension, enables business people of all kinds to build and implement sophisticated scoring models without the need of an analyst or the need to understand the underlying statistical methodologies.
Return on Investment
Two avenues are open to the company in need of such models. The first is to contract with an analytical consulting firm to develop and implement a model, the typical cost of which is in the area of $100,000. The second is to purchase the necessary statistical software (roughly $10,000) and demographic/psychographic data (roughly $80,000), and have an in-house analyst build and implement a model. With Scoring Science, the typical ArcView Business Analyst user now has the power to build and implement a successful scoring model at a fraction of the cost and time.
Application: The Allocation of Scarce Resources
Once again, consider the question: Which neighborhoods served by a certain healthcare corporation are at the highest risk for a particular disease? Our imaginary healthcare corporation operates numerous hospitals and clinics throughout the northeastern United States. It also has incidence rates for the disease in question for a number of metropolitan areas throughout the nation. This data is loaded into Business Analyst, geocoded, and used to generate an appropriate target variable. The user would then import the demographic data available with Esri's ArcView Business Analyst, and link this data to the target variable. Next, the user would launch the Scoring Science wizard to build the model. Once the model had been tested and validated, it would then be deployed using only the data from those neighborhoods that it actually served (neighborhoods for which incidence rates for the disease are unknown). Scoring Science will assign scores to the different neighborhoods, ranking the risk in each neighborhood for the given disease. In the past, a model such as this would have cost the healthcare corporation tens of thousands of dollars. Now it can be built in less than a day using Esri's ArcView Business Analyst and Scoring Science.
No More Guesswork
Whatever the field--insurance, finance, e-commerce--Scoring Science enables you to identify the critical behavioral patterns within your customer and client databases, so that instead of subjectivity and conjecture, you can now base your decisions on solid, statistical analysis. Simply put, Scoring Science helps businesses identify their most profitable current and potential customers, reducing costs and maximizing return on investment.
Stone Analytics' Expertise
Stone Analytics develops, markets, and supports productized analytical solutions that add value to any data-rich application environment. Stone Analytics' statistical modeling products are easy for software developers to integrate, and easy for customers to use. These tools bring point-and-click simplicity to powerful statistical models for rank ordering, forecasting, grouping, and similar everyday business decision-making tasks. Stone Analytics employs the latest statistical and information management techniques to provide customers with state-of-the-art solutions that add statistical modeling functionality to their applications.
Stone Analytics supplements its products with a wide range of service offerings customized according to customer needs. These services include integration assistance, custom model building, and consulting in the areas of data management and analytics.
Christy Joiner-Congleton
President
Stone Analytics, Inc.
3665 Ruffin Road, Suite 300
San Diego, CA 92123
Phone: (858) 503-7540
Fax: (858) 503-7541
cjc@stics.com