| Measurement-theoretic foundations | Can composites be assumed to have the same significance as factors in representing conceptual variables? How does metrological uncertainty contribute to this assessment? Under which conditions should composites or factors be preferred for measuring conceptual variables? | Rhemtulla et al. (2020), Rigdon et al. (2019), Rigdon and Sarstedt (2022), Rigdon et al. (2020) |
| Statistical assumptions of the standard PLS-SEM algorithm | Assessing the impact of violating a method’s statistical assumptions (e.g., cross-loadings) on parameter bias and predictive performance | Lohmöller (1989, Chap. 2) |
| Modeling capabilities | Extending the modeling capabilities, for example by allowing for relationships of an indicator to multiple composites, setting model constraints, and implementing circular and bidirectional relationships. Further extensions include different forms of moderated mediation analyses and hierarchical component models | Lohmöller (1989, Chaps. 2 and 3), Sarstedt et al. (2019; 2020) |
| Big data analytics | How can PLS-SEM support big data and machine learning research? | Akter et al. (2017), Richter and Tudoran (2024) |
| Model specification search | Improve the model specification search based, for instance, on Cohen’s path method to explore path directionality (Callaghan et al., 2007) and the fuzzy-set qualitative comparative analysis (fsQCA) in PLS-SEM (Rasoolimanesh et al., 2021). Thereby, research can benchmark their theoretically established model against model alternatives with, for example, the best predictive capabilities | Cho et al. (2022), Marcoulides and Drezner (2001), Marcoulides and Drezner (2003), Marcoulides et al. (1998) |
| Model misspecification assessment | Extending the set of model evaluation criteria, for example to identify measurement model misspecifications | Gudergan et al. (2008) |
| Congruence assessment | Introduce congruence assessment to examine whether constructs in the nomological network have proportional correlations | Franke et al. (2021) |
| Striking a balance between explanation and prediction | How can explanatory and predictive goals be best accommodated in PLS-SEM-based modeling, especially when considering model selection? When considering out-of-sample prediction, should the focus be on predicting certain specific constructs or the overall model? | Liengaard et al. (2021), Sharma et al. (2019; 2021) |
| Robustness checks | Robustness checks of the estimated model, including common method bias, endogeneity, nonlinear relationships, impact of collinearity in formative measurement models, necessary condition analysis, and fuzzy-set qualitative comparative analysis in PLS-SEM | Chin et al. (2013), Hult et al. (2018), Rasoolimanesh et al. (2021), Richter et al. (2020) |
| Latent class analysis | Improve the validity of latent class techniques by including explanatory variables as covariates in the model estimation and by analyzing the heterogeneity of intercepts and unstandardized coefficients | Bray et al. (2015), Sarstedt et al. (2022a, 2022b) |
| Longitudinal data analysis | How can researchers compare models across time in longitudinal analysis? | Jung et al. (2012), Lohmöller (1989, Chap. 6), Roemer (2016) |
| Multilevel modeling | How can PLS-SEM be used for multilevel modelling when we are analyzing data that are drawn from a number of different levels. For instance, levels such as a country’s gross domestic income and gender may be used for PLS path models on job satisfaction (Drabe et al., 2015), sustainable consumption behavior (Saari et al., 2021), and circular innovation (Saari et al., 2024) | Hwang et al. (2007), Jung et al. (2015) |