Key Feature Insights

Binary Outcome (Logistic Regression)

Setup: The outcome is yes/no (e.g., "Did the user sign up?").
Coefficients: Each feature has a weight (coefficient, β). Positive values mean the feature increases the odds of the outcome; negative values mean it decreases the odds.
Feature importance: Often expressed as an odds ratio or average marginal effect, which answers:
"If this feature changes, how much more/less likely is the user to take action?"

Example: Key feature disbursement
* A coefficient of +2.306 means users with "results = Very High" are about twice as likely to convert.
Insight: Coefficients highlight which experiences drive or block conversions.

Think of the coefficients as knobs you can turn up or down on user behavior.

Continuous Outcome (Linear Regression)

Setup: The outcome is a numeric value (e.g., "Purchase price").
Coefficients: Represent the expected change in the outcome per unit change in the feature.
Feature importance: Answers: "Which features have the largest impact on the outcome`s magnitude?"

Example: Key feature disbursement
* A coefficient of +1.577 for internal users might indicate an increase in average purchases.
Insight: Identifies features that shape the scale of results, not just binary outcomes.