Onsite Analytic Consulting

• Predictive Modeling
• Analytic Datamarts

Morgan Stanley, Wells Fargo Bank, Intuit and eToys have benefited from our profitability focused approach to customer relationship analysis.

Analytic consulting can blend into full fledged Onsite System Development projects. For instance, end-to-end email campaign tracking systems, including analysis of outcomes and revenue impact. Or datamarts focused on profitability or delivering targeted messages and other interactive marketing. In these cases, we are architect, implementor, and analyst.

Our analytic approaches are best suited to clients with many thousands of customers, many millions of dollars in revenue, and a million or more dollars in customer acquisition expenditure. Typically, our clients have a million or more customers, engage in many multi-million dollar marketing efforts, and strive for or have revenues of a billion dollars or more.

Our Predictive Modeling Process

When you invite us to create a predictive model for your company we enter into a process of consultation, concrete demonstration, and infrastructure enhancement. The initial consultation, and often this entire process, is free and without obligation. Because we are able to demonstrate the real world value of our predictive models when applied to specific business problems, without giving the model itself away, we are able to provide our prospective clients with a way to see the value of our work before making a financial commitment.

The process works as follows:

Step 1: A free Consultation

We meet with you to identify the greatest opportunities for benefit from predictive models. For example,

• parameters which lead to the fewest defects in manufacturing processes;
• identify customers most at risk of switching to competitors;
• identify prospects most likely to respond to a marketing offer, and the most profitable offer to make to them.

Step 2: Produce a Dataset of Predictive Variables

We assess your existing database systems for their ability to support predictive analysis.

If sufficient data exists to produce a predictive model, we help to define a predictive dataset for our use in creating a predictive model. Otherwise, we can help to define and build the processes and databases necessary to create an analytic datamart. Producing a comprehensive dataset can be very time consuming. Frequently, we can help you produce a simple dataset that will serve the purpose of achieving substantial improvements over your existing processes.

Step 3: A Custom Model that Addresses your Specific Needs

Given a dataset for our use, we can produce a free custom model applicable to your immediate business needs. The model results are presented in the form of specific predictions abou the outcome of individual events. We also provide measures of the overall predictive power of the model.

Step 4: Independent Evaluation of Our Model

To see that our models really work as claimed, we ask our clients to give us explanatory data, but not the outcomes, for several thousand cases. We will confidently identify the small group of cases that contain virtually all of the outcomes of interest.

For example, a business might lose 3% of their customers each period. One of our models could predict the probability that each customer would leave. The prediction would use past data given to us by the client. Typically, we could use this data alone to provide a list of 25% of the customers that we will contain 80% of the customers who end up leaving. Although we are not given knowledge of the actual outcome, our client knows which customers left and can independently confirm the power of our predictions.

A model such as the one in this example can produce substantial financial benefit when used with current data to predict the future. Assume $10 were going to be spent on 100,000 customers to help retain their business. Using our model, $20 could be spent on the 25,000 customers most at risk. This new strategy would likely be more effective and cost $500,000 less to implement. Alternatively, the original $10 campaign could be rolled out in a more focused fashion to the smaller 25,000 customer segment, resulting in a similar outcome at a quarter of the cost.

Having seen a demonstration of our ability to predict outcomes for today using only knowledge of yesterday's data, we provide independent validation and confidence in our ability to predict tomorrow's outcomes using data from today.

This ends our free, no obligation, demonstration which can transition into a rapid implementation of this and other models.

Step 5: Using our Models

If a client wishes to proceed, our next step is to make predictions about future outcomes which address urgent business needs. We can 'score' a set of existing data and provide a dataset that can be used to prioritize customers. We can also provide parameters and a program that can be included in existing datawarehouses and business intelligence systems to make realtime predictions and decisions. In either case, the scores and parameters will be useful for making decisions for the next few months, or few years, depending on the business problem and characteristics of the particular business. We will continue to work with you to implement effective and industry leading prediction-based decison making.

For more information, call call business intelligence consultant, Lawrence Sinclair, at (415) 407-1475.

Analytic Datamarts

Frequently, clients do not have analytic datmarts or other data sources that can be readily used to make predictions about their current business problems. Before we can do our predictive magic, these systems need to be in place. We frequently build, assess, and re-engineer analytic datamarts for our clients. Creating such systems can take anywhere from a few weeks to a few months. Analytic datamarts support a wide range of reporting, analysis, quality assurance, and decision making in addition to predictive modeling.

Once in place, an analytic datamart enables us to provide a free, no-obligation demonstration of our predictive models.