Abstract
PLS predictionoriented segmentation (PLSPOS; Becker et al. 2013) is a distancebased segmentation method. It follows a clustering approach with a deterministic assignment of observations to groups and uses a distance measure for the reassignment of observations; as such, it has no distributional assumptions.
Description
PLS predictionoriented segmentation (PLSPOS) (Becker et al. 2013) is a distancebased segmentation method that builds on earlier work on distance measurebased segmentation (i.e., the PLS typological path modeling, PLSTPM, approach, Squillacciotti 2005, and its enhancement, the responsebased detection of respondent segments in PLS, REBUSPLS, Esposito Vinzi et al. 2008).
PLSPOS algorithm introduces three novel features: (1) it uses an explicit PLSspecific objective criterion to form homogeneous groups, (2) it includes a new distance measure that is appropriate for PLS path models with both reflective and formative measures and is able to uncover unobserved heterogeneity in formative measures, and (3) it ensures continuous improvement of the objective criterion throughout the iterations of the algorithm (hillclimbing approach).
PLSPOS follows a clustering approach with a deterministic assignment of observations to groups and uses a distance measure for the reassignment of observations; as such, it has no distributional assumptions. The segmentation objective in a PLS path model is to form homogenous groups of observations with increased predictive power (R² of the endogenous latent variables) of the groupspecific path model estimates (compared to the overall sample model).
A repeated application of PLSPOS with different starting partitions is advisable to avoid local optima.
Becker et al. (2013) and Hair et al. (2024) describe the PLSPOS method in detail.
PLSPOS Settings in SmartPLS
Number of Segments
The number of predefined segments for which the segmentation will be performed.
Maximum Iterations
The maximum number of iterations that the segmentation algorithm will perform. Should be sufficiently high for a good segmentation solution.
Search Depth
The maximum search depth is the maximum number of observations in the sorted list of candidate observations for reassignment that will be tested if they improve the PLSPOS objective criterion.
This number may not exceed the number of observations in the overall sample. In initial explorative research stages, one may use a reduced search depth for reasons of performance. However, to determine the final segmentation result, the search depth should equal the maximum number of observations to ensure that the segmentation solution minimizes the PLSPOS objective criterion.
Initial Separation
The initial separation of data into the prespecified number of groups can either be based on a Random Assignment to groups or on a prior FIMIX segmentation solution.
If the FIMIX Segmentation is chosen, the user also has to specify the nessesary FIMIXPLS setting in a separate settings tab.
PreSegmentation
If this option is selected, the algorithm will perform a presegmentation in the first round that assigns all units to its best fitting group according to the distance measure.
It will not be checked whether this improves the objective criterion.
Optimization Criterion
The optimization criterion (also objective criterion) will be optimized when estimating the segments in the PLSPOS algorithm. There are two options:

Sum of All Construct RSquares: Uses the sum of all RSquares in the model for all segments as the PLSPOS objective criterion that will be optimized (maximized) when reassigning observations in the course of the segmentation.

Sum of Target Construct RSquare: Uses the target constructs sum of Rsquare values over all segments as the PLSPOS objective criterion that will be optimized (maximized) when reassigning observations in the course of the segmentation.

Sum of All Construct Weighted RSquares: Uses the sum of all weighted RSquares in the model for all segments as the PLSPOS objective criterion that will be optimized (maximized) when reassigning observations in the course of the segmentation The weighting of the RSquares is done by using the relative segment sizes.

Sum of Target Construct Weighted RSquare: Uses the target constructs sum of weighted Rsquare values over all segments as the PLSPOS objective criterion that will be optimized (maximized) when reassigning observations in the course of the segmentation. The weighting of the RSquares is done by using the relative segment sizes.
Target Construct
If Optimization Criterion is Sum of all Construct RSquares or "Sum of All Construct Weighted RSquares*, then this option does not have to be specified.
If Optimization Criterion is Sum of Target Construct RSquare or Sum of Target Construct Weighted RSquare, then this option defines the target construct for which the outer residuals or Rsquare value is calculated.
References

Becker, J.M., Rai, A., Ringle, C. M., and Völckner, F. (2013). Discovering Unobserved Heterogeneity in Structural Equation Models to Avert Validity Threats, MIS Quarterly, 37(3): 665694.

Hair, J. F., Sarstedt, M., Ringle, C. M., & Gudergan, S. P. (2024). Advanced Issues in Partial Least Squares Structural Equation Modeling (PLSSEM), 2nd Ed., Thousand Oaks, CA: Sage.

Esposito Vinzi, V., Trinchera, L., Squillacciotti, S., and Tenenhaus, M. (2008). REBUSPLS: A ResponseBased Procedure for Detecting Unit Segments in PLS Path Modelling, Applied Stochastic Models in Business & Industry 24(5): 439458.

Squillacciotti, S. (2005). Prediction Oriented Classification in PLS Path Modeling, PLS & Marketing: Proceedings of the 4th International Symposium on PLS and Related Methods, T. Aluja, J. Casanovas, V. Esposito Vinzi and M. Tenenhaus (eds.), Paris: DECISIA: 499506.
Please always cite the use of SmartPLS!
Ringle, Christian M., Wende, Sven, & Becker, JanMichael. (2024). SmartPLS 4. Bönningstedt: SmartPLS. Retrieved from https://www.smartpls.com