Platform

Decision Tree algorithm finds the variable that makes the most difference on the dependent variable and the break-point by analyzing the variables in the data set. Continuing the analysis step by step, the tree diagram with the most divergence of the dependent variable is created.

Random Forest algorithm is a wider scanning version of the decision tree algorithm. All combinations of variables and many more tree structures are tested, allowing us to calculate the effect on the dependent variables.

Both algorithms provide insights to take action from the data we have, thereby providing guiding recommendations for improving target KPIs.

Acsight obtains action-oriented insights by using Structural Equation Modeling (SEM), one of the multivariate analysis techniques, to determine the variables that affect the NPS.

SEM analysis is a technique that takes into account not only the direct relationships of the variables on the NPS but also the indirect relationships. Thus, strategies that can guide the communication strategy can be determined.

The numbers above the arrows show how many units that element will increase in the variable (the dependent variable) to which it is linked if I increase the element by 1 unit. These effects are –standardized and range from 1 to 1. Thus, each effect can be compared with each other.

It is the effect that we directly observe the effect of that variable on NPS. In this example, a 1 unit increase in Element2 gives a 0.3 unit increase on NPS.

Although the variable does not seem to affect NPS directly, it shows the observed effect on other variables. For instance, Element3 triggers the NPS via the Element2 and Element1 variables.