Drivers

Drivers provide us with an opportunity to measure and correct our performance before an objective is to be completed.  Drivers are identified using correlation analysis which is a technique that makes use of business intelligence technology.  Up until now our methodology made use of traditional planning techniques using traditional information technology.

Correlation analysis uses a mathematical algorithm to identify the relationships between attributes found in a dataset.  We are most interested in the attribute that represents our outcome and it's relationship with all other attributes.  Some but not all relationships will be statistically significant. Those attributes that have a significant relationship with the outcome are our drivers.

The example that we use to explain the Weincor business intelligence method is as follows (1)

Students that had been offered college admission are surveyed and asked to rate the importance of various factors that may have affected their decision to accept or reject the offer, using this scale; 1 - Not important, 2 - Somewhat important, 3 - Important, 4 - Very important and 5 - Critical to the decision.  Each question that a student is asked becomes an attribute in our dataset with a value of 1 - 5. Each row of data represents a student's response.  We want to know what are the attributes that drive a students decision to either accept or reject an offer of admission.  Correlation analysis identified the attributes (row a10...an) and the strength of their correlation (column B) with a decision to accept or reject an offer as follows

The further away a value is from zero (positive or negative) the stronger the correlation.  In total there were seventy attributes that were analyzed however only eleven had statistically significant correlation and were determined to be drivers.  Each of the drivers have a prefix (short name) and a description as follows

a10 - Great parties

a13 - Attending an event (e.g. theatrical, sporting, homecoming...etc.) with a faculty member

a22 - Close to home

a24 - College offers spring break (for course credit) in cool countries

a30 - Housing quality

a38 - Friends recommendations

a41 - Large college

a54 - Meeting a faculty member in their office

a56 - Hearing from the college president by telephone

a63 - Meeting a graduate of the college while shopping, working, going to a concert

a70 - College will accept all/most of credit earned elsewhere

We need to measure the distribution of values for each of these drivers as found in our dataset.  This is referred to as our baseline distribution. Later we will adjust the distribution so that can achieve our objective.  This is our target distribution.  We will take incremental measurements during the admissions cycle to determine if our drivers are on target and will know if we will meet our objective before it is to be completed.  Because we will know the results ahead of time, we refer to drivers as leading indicators.  We can therefore correct our performance if we are not on target.

However, before we identify the target distribution for our drivers, we need to create a predictive model and simulation based on the objective that we want to meet.  Click on Models to learn how we create a model that can predict and simulate an outcome.

(1) The data that was used to perform the correlation analysis was included in the book Dr. Green's New Hyper-Linked Introduction to IBM SPSS Modeler, Dr. John B. Green Jr. PHD, Left Brain Books, LLC.  The successful use of business intelligence depends on the proper identification and collection of data.  Dr. Green provides a valuable data resource to illustrate our method.