To provide a better understanding of the approach, Fig. 1 shows a simple PLS path model with four latent variables, Y1 to Y4 (represented by circles), determined as the weighted sum of their assigned indicators x (represented by the rectangles). In other words, in the measurement model (also called the outer model), a block of directly observable indicators represents each latent variable that is not directly observable. In the structural model (also called the inner model), the latent variables have pre-defined and theoretically/conceptually established relationships.

The goal of the PLS-SEM approach is to generate latent variable scores that jointly minimize the residuals of the ordinary least squares (OLS) regressions in the model (i.e., maximize the explanation). The resulting latent variable scores are unique and determine the case values of each observation (i.e., the algorithm provides determinate latent variable scores). They also make it possible to predict the indicators (x7-x12) of the endogenous or dependent latent variables in the structural model (Y3 and Y4). In short, PLS-SEM is a variance-based method that estimates composites representing latent variables in path models. Hair, Hult, Ringle, and Sarstedt (2017), for example, provide additional explications of PLS-SEM, including details on how to create and estimate PLS path models and how to evaluate the results (also see Chin, 1998, 2010; Falk & Miller, 1992; Haenlein & Kaplan, 2004; Hair, Ringle, & Sarstedt, 2011; Henseler, Hubona, & Ray, 2016; Roldan & Sanchez-Franco, 2012; Tenenhaus et al., 2005 ).