Key driver analysis is vital research for most businesses and is used to identify the most important aspects driving crucial business decisions. It can answer the common questions that may arise within your company, such as “What features of this product are influencing sales?” or, “Which areas of our service affect customer satisfaction the most?”
There are numerous methods of quantifying the variables that influence a key outcome, including various multiple linear regression techniques. However, linear regression can result in biased, spurious or suppressed results, most notably when some of the independent variables are closely correlated. The problem that occurs with highly correlated independent variables is multicollinearity, which is the inability to separate the contribution of the independent variables on the target.
Shapley Value regression analysis has been derived by Lipovetsky & Conklin from the Shapley Value method developed within the context of game theory. The Shapley Value gives an indication of the prospects of winning when co-operating with another player; the higher the Shapley Value, the better the player's prospects. The Shapley value analysis provides consistent results in the presence of multicollinearity because it calculates a score for all possible combinations of the independent variables, thus determining the effect of each one separately. Lloyd Shapley with Alvin E.Roth won the 2012 Nobel Memorial Prize in Economic Sciences for "the theory of stable allocations and the practice of market design."
Aside from the benefits of being less affected by multicollinearity, Shapley Value analysis also offers easily understandable results; it shows the contribution of each independent variable as a percentage of the whole, so no negative results. Furthermore, this technique is stable when measuring effects over multiple waves.
After Shapley Value regression has been applied, significance testing can be implemented in order to see which variables have an influence on the dependent variable and to exclude irrelevant variables from the model. Different combinations of independent variables can be interrogated to see how they affect the dependent variable and each other, providing a deeper understanding of how the variables interact.
Remove the risk of erratic and illogical results by investing in Digitab’s Shapley Value Regression tool.
For more information on Shapley Value Regression or to speak to one of our consultants please call Mike Wright on tel +44 (0) 20 7031 0287 or email email@example.com.