Back to Learn

The Math: Key Driver Analysis

Key Driver Analysis identifies which independent variables (attributes) have the strongest impact on a dependent variable (e.g., Overall Satisfaction).

Why not simple correlation?

Survey attributes are often highly correlated with each other (multicollinearity). If you just look at correlation, everything looks important. If you run standard OLS regression, the coefficients can become unstable or even negative.

Our Approach: Ridge Regression

Krosstabs uses Ridge Regression (L2 Regularization). This adds a penalty term to the regression equation that shrinks coefficients, reducing variance and handling multicollinearity effectively.

β = (X'X + λI)⁻¹ X'y

We automatically standardize all variables (Z-score) before running the regression, so the resulting coefficients (Beta weights) are directly comparable as "Importance Scores".

Relative Importance

We rescale the absolute values of the coefficients to sum to 100%, giving you a clear "Relative Importance" percentage for each driver.