Driver analysis helps you discover which qualities within your business are affecting others and quantify what has the strongest impact on your culture. Our driver analysis feature uses statistical techniques to identify the relationship between different variables.
The feature offers two types of analysis and will default to the best fit for your organization. The approach used depends on factors such as the size of your survey population and the variability of your results.
Multivariable linear regression: This model looks at a combination of independent variables such as our culture qualities (i.e. Communication), and how each quality is correlated with an outcome metric, such as eNPS. It considers the impact of the qualities as a group.
In a multivariable linear regression, certain qualities may be dropped from the model if the variability of the outcome metric can be better explained without them. Dropped qualities may be less correlated with the outcome metric, but they can still have an impact on your culture strategy.
Univariate linear regression: This is the simplest regression model. The driver analysis will default to this method only if the variability in your outcome metric is not well explained by a combination of qualities, which tends to happen in surveys with small populations. This approach analyzes each quality individually and calculates how much each is correlated with an outcome metric such as eNPS. Usually all CultureIQ Qualities are highly correlated with eNPS, but this analysis may reveal that some are more highly correlated than others. If an input variable is not included in the regression output, it means the variable is not highly correlated with the outcome metric (if you’re really curious, we exclude any variable with an r^2 < .6).