Using regression to find factors that correlate provides a greater understanding of the associations among factors and the ability to forecast the factors that drive demand, revenues, profit, etc.
What it is: Regression analysis involves identifying the relationship between a dependent variable and one or more independent variables. Regression is used to look for associations within data sets, identifying what factors are related. Knowing the relationship between factors a manager can make business decisions about which factors they could modify to impact revenues and profitability, or they could predict the impact of changes created by the movement of independent factors that they do not control. Correlation is the measure of how closely the independent and the dependent variables move together. Perfect correlation is measured as 1 and perfectly inverse relationships are measured as -1.
What does it do:Armed with the relationship correlations among factors within a market, industry etc., a manager can predict what will happen as individual factors are impacted or modified. As an example, if you can correlate the relationship between price and demand, you can predict the increased volume demand if the price of a product was reduced by 10%.
Uses:
How is it used:Correlations are used to understand how different factors interact with a business system and how they impact each other. The correlation can be causal (A causes B to happen) or correlated (When A moves than B moves the same way or the inverse way) or have no relationship whatever (when A moves there is no indication what B will do). Knowing which levers you can pull to impact change can allow a business to make proactive moves. Knowing how the movement of certain trends or factors in the industry/economy will affect your business will allow you to make reactive, proactive, or pre-emptive actions in the marketplace.
Where:Regressions are run to find correlations in many applications in business today. Some examples include:
Find the correlation of price vs. demand to be able to predict the impact of price changes on volume demand. This same type of correlation can be used to determine the impact of price on market share between competitors.
Forecasting product demand growthbased oneconomic variables (e.g. what is the impact of changes in disposable income on automobile sales)
Forecast product demandbased on correlated factors (e.g. what is the impact of housing starts on dishwasher sales since there is an average of .95 dishwashers per housing start)
Correlations can be completed on manufacturing issues to forecast manufacturing efficiencies and costs
Relationships can be found between certain employee behaviors and success in the company (this could be used in screening potential employees)
Correlations can be used to determine inventory order quantities, order points, inventory quantities in storage, safety stocks, length of storage for perishable products, etc.
Correlations can determine the factors impactinginventory shrinkage and can indicate actions that could be taken to reduce shrinkage
Identifying factors for prospective customers (lead scoring analysis)
Churn Analysis: the analysis will predict which customers will leave.
Identify the factors driving safety failures and predicting the impact of new safety procedures in avoiding future accidents
Finding thefinancial correlationsamong price, volume, fixed costs, and variable costs in predicting the impact on the profitability of changes to these factors.
Why:Without regression and correlation insights managers would be making many decision blinds without knowing how their decisions will impact their business
Limitations:
Where it shouldn't be used:There are no situations where the use of regression analysis to find correlations should not be used.
Any restrictions:Do not use regression and correlation when the data quality is poor.
Warnings: Regression analysis to find correlations is a science, but knowing which factors to regress and interpreting the results in an art. Therefore the results must be interpreted with care and common sense tests need to be applied to the results and interpretations. It is also best to test the results to see if the predicted outcomes turn out as expected. Regressions are more effective when there is a lot of variation in the variables. There is also the risk of an omitted variable that could have a major impacted but was not included in the regression and correlation analysis.
Gathering data:The data for the analysis is gathered
Independent variables that will be analyzed
The dependent variables that will be regressed against the independent variables are gathered
Analysis of data: Run the regression analysis and look for correlated factors
Interpretation of results: Determine if any of the dependent variables have a direct or inverse relationship to the independent variable. Try to determine if there is causality or merely correlated responses. Is there a third variable that both the dependent and the independent variables are correlated with.
Presentation of results: Regression analysis is shown as a combination of the regression charts that show the scatter plot with the regression line included. Then the variability and confidence levels are defined. Then the predictive elements that drive the recommendations are demonstrated.
Regression analysis is shown as a combination of the regression charts that show the scatter plot with the regression line included. Then the variability and confidence levels are defined. Then the predictive elements that drive the recommendations are demonstrated.
