SmartPLS is the most widely used software for Partial Least Squares Structural Equation Modelling (PLS-SEM) — a technique that lets you test a whole network of relationships between latent constructs (things you measure with multiple questionnaire items) in one model. If your thesis has a conceptual framework with arrows between constructs like trust, satisfaction and loyalty, PLS-SEM is very likely the technique you'll be asked about.
PLS-SEM vs CB-SEM in one paragraph
Covariance-based SEM (AMOS, LISREL) asks: does my whole theorised model fit the data? — best for confirming established theory with larger, roughly normal samples. PLS-SEM asks: how well does my model explain and predict the outcomes? — best for prediction, theory building, complex models, smaller or non-normal samples, and formative constructs. The full comparison is in SPSS vs AMOS vs SmartPLS.
The PLS-SEM workflow
1. Assess the measurement model
- Indicator reliability — outer loadings ≥ 0.708 (items below 0.40 are removed; 0.40–0.708 removed only if it improves reliability/validity).
- Internal consistency — Cronbach's alpha and composite reliability (rho_c) ≥ 0.70.
- Convergent validity — Average Variance Extracted (AVE) ≥ 0.50.
- Discriminant validity — HTMT below 0.85 (0.90 for conceptually similar constructs).
2. Assess the structural model
- Collinearity — inner VIF values below 3–5.
- Path coefficients — size, sign and significance via bootstrapping (5,000+ resamples).
- Explanatory power — R² of endogenous constructs; effect sizes f².
- Predictive relevance — Q² (blindfolding) or PLSpredict.
3. Extensions when your model needs them
SmartPLS handles mediation and moderation analysis, multi-group analysis (does the model differ for men vs women?), higher-order constructs, and importance–performance maps — which is why it dominates management, marketing and IS research.
Sample size — the honest version
PLS-SEM tolerates smaller samples than CB-SEM, but the old '10 times rule' is a floor, not a defence. Reviewers now expect a power-based justification — start with our sample size calculator and the guidance in how to determine your sample size.
Journals expect the full chain: measurement model results (loadings, alpha, CR, AVE, HTMT), then structural results (VIF, paths with bootstrapped confidence intervals, R², f², Q²). Missing links in this chain are the most common reason PLS-SEM papers bounce. The exact cut-offs — and what to do when a construct misses one — are in convergent and discriminant validity thresholds.
If you're deciding between tools or need your model reviewed before you collect data, our SmartPLS mentoring coaches you to run and defend the analysis yourself — or book a free consult.
Related statistics guides: convergent and discriminant validity thresholds, average variance extracted (AVE), composite reliability, the HTMT ratio, SPSS vs SmartPLS, model fit indices in AMOS, interpreting SmartPLS bootstrapping and jamovi, the free SPSS alternative.
Frequently asked
Is SmartPLS free?+
SmartPLS 4 offers a free tier with limited features alongside paid licences; universities often hold licences. The free student options are typically sufficient to learn the workflow before your main analysis.
What sample size does PLS-SEM need?+
Enough for the statistical power your model requires — commonly 100–300+ in practice. The '10 times the maximum arrows' rule is only a minimum bound; use power analysis for a defensible number.
Can I use SmartPLS for exploratory research?+
Yes — that's one of its core strengths. PLS-SEM suits theory building and prediction-oriented studies, whereas CB-SEM (AMOS) suits strict confirmation of established theory.
phdguide's mentors are senior academics, former supervisors, statisticians and publication specialists with 25+ years' average experience guiding MBA, MPhil and PhD scholars from topic to viva.
Ethical, compliant guidance: We provide academic support, mentoring, analysis, editing and structuring — not authorship. Your work stays compliant with university policies.