Key driver analysis

What is actually associated with the outcome you care about?

Pick an outcome, review candidate drivers, and run a target-appropriate model on mixed survey data: numbers, opinion scales, single categories, and multi-select columns together. AddMaple ranks drivers automatically without collapsing the result into a single black-box score.

Rank the strongest relationships.

Random Forest handles non-linear mixed data. Elastic Net supports numeric outcomes, logistic Elastic Net binary outcomes, and ordinal and multinomial Elastic Net ordered and multiclass outcomes; automatic routing is available when the engine should choose. Elastic Net models return interpretable coefficients with direction and magnitude.

Check when a ranking hides an important difference.

Signed, class-specific associations provide coefficient and sign information when one global ranking would mislead on Likert or multiclass outcomes. AddMaple also surfaces statistically related columns first in the driver picker, so the candidate list starts with plausible drivers rather than the full codebook.

Use a driver result as evidence, not a verdict.

Inspect ranked importance, model fit (out-of-bag accuracy or R²), and AI-generated interpretation, then save the result to Insights or dashboards. Key-driver analysis answers which factors matter most; use regression, PCA & multivariate modelling when you need explicit coefficients or latent dimensions.

Make the model useful to the person who has to explain it.

The output is not just a ranked list. Analysts can see which variables were considered, how the target shaped the method, whether the model fit is useful, and where direction is available. That creates a better bridge from statistical output to the stakeholder question: which parts of this experience should we investigate or act on next?