Result Color Thresholds in SmartPLS Reports
SmartPLS (Ringle, Wende, & Becker, 2024) uses result colors in reports to help you quickly identify values that meet, partly meet, or do not meet common evaluation guidelines in partial least squares structural equation modeling (PLS-SEM). Green, black, and red values are visual cues only. They support the interpretation of results but do not replace a full assessment of the research model, measurement model, structural model, sample, theory, and study context.
The thresholds summarized on this page are primarily based on recommendations in the PLS-SEM literature, especially Hair, Hult, Ringle, and Sarstedt (2027) and Hair, Sarstedt, Ringle, and Gudergan (2024). Depending on the research objective, model type, discipline, and publication requirements, researchers may need to justify different cut-off values.
Threshold Overview at a Glance
The following table summarizes all result color thresholds used in SmartPLS reports. Detailed explanations and the underlying literature follow in the sections below.
| Criterion | Green | Black | Red | Key source |
|---|---|---|---|---|
| Outer loading | >= 0.70 | — | < 0.70 | Hair et al. (2027) |
| f-square (f²) | >= 0.15 | >= 0.02 and < 0.15 | < 0.02 | Cohen (1988); Hair et al. (2027) |
| Cronbach's alpha | >= 0.70 | — | < 0.70 | Hair et al. (2027) |
| Composite reliability (rho_a) | >= 0.70 | — | < 0.70 | Dijkstra & Henseler (2015); Hair et al. (2027) |
| Composite reliability (rho_c) | >= 0.70 | — | < 0.70 | Hair et al. (2027) |
| Average variance extracted (AVE) | >= 0.50 | — | < 0.50 | Fornell & Larcker (1981); Hair et al. (2027) |
| HTMT | <= 0.85 | > 0.85 and <= 0.90 | > 0.90 | Henseler, Ringle, & Sarstedt (2015) |
| VIF | <= 3.00 | > 3.00 and <= 5.00 | > 5.00 | Hair et al. (2027) |
| Bootstrapping p value | <= 0.05 | — | > 0.05 | Hair et al. (2027) |
| Q²_predict | >= 0 | — | < 0 | Shmueli et al. (2019) |
| PLS-SEM_RMSE vs. LM_RMSE | PLS-SEM < LM | — | PLS-SEM >= LM | Shmueli et al. (2019) |
| PLS-SEM_MAE vs. LM_MAE | PLS-SEM < LM | — | PLS-SEM >= LM | Shmueli et al. (2019) |
| Permutation MGA p value | <= 0.05 | — | > 0.05 | Hair et al. (2024) |
| MICOM Step 2 p value | >= 0.05 | — | < 0.05 | Henseler, Ringle, & Sarstedt (2016) |
| MICOM Step 3a and 3b p values | <= 0.05 | — | > 0.05 | Henseler, Ringle, & Sarstedt (2016) |
How to Interpret Result Colors
| Color | Meaning in SmartPLS reports | Recommended action |
|---|---|---|
| Green | The value meets the recommended threshold. | Continue with the substantive interpretation and check the remaining criteria. |
| Black | The value is acceptable or noteworthy but requires attention. | Interpret the result carefully and consider the study context. |
| Red | The value does not meet the recommended threshold. | Inspect the model, data, indicators, and theoretical reasoning before drawing conclusions. |
A green value does not automatically prove that a model is valid, and a red value does not automatically mean that a model must be rejected. In some situations, deviating cut-off values are well established in the literature. For example, outer loadings of 0.60 or higher can be acceptable in exploratory research, and an HTMT threshold of 0.90 can be justified when constructs are conceptually very similar. If you rely on such alternative thresholds, justify them explicitly in your report.
PLS-SEM Algorithm and Consistent PLS-SEM (PLSc-SEM) Results
The following thresholds apply to reports generated from the PLS-SEM algorithm and, where applicable, consistent PLS-SEM (PLSc-SEM). Note that these criteria address reflective measurement models. Formative measurement models are evaluated with different criteria, such as convergent validity (redundancy analysis), indicator collinearity (VIF), and the significance and relevance of outer weights. Reliability statistics such as Cronbach's alpha, rho_a, rho_c, and AVE are not meaningful for formative constructs.
Outer Loadings
| Criterion | Green | Red |
|---|---|---|
| Outer loading | >= 0.70 | < 0.70 |
Outer loadings indicate how strongly an indicator is associated with its construct. Loadings of 0.70 or higher are generally desirable because they indicate that the construct explains more than half of the indicator's variance (0.70² is approximately 0.50). Lower loadings should be evaluated in relation to content validity, reliability, AVE, and the theoretical relevance of the indicator. Indicators with loadings between 0.40 and 0.70 should only be removed if their removal increases composite reliability or AVE above the recommended threshold.
f-Square Effect Size
| Criterion | Green | Black | Red |
|---|---|---|---|
| f-square / f² | >= 0.15 | >= 0.02 and < 0.15 | < 0.02 |
The f-square effect size evaluates how strongly an exogenous construct contributes to explaining an endogenous construct. Following Cohen (1988), f² values of 0.02, 0.15, and 0.35 represent small, medium, and large effects. SmartPLS combines medium and large effects (f² >= 0.15) in green, marks small effects (0.02 <= f² < 0.15) in black, and marks negligible effects (f² < 0.02) in red.
Construct Reliability and Validity
| Criterion | Green | Red |
|---|---|---|
| Cronbach's alpha | >= 0.70 | < 0.70 |
| Composite reliability (rho_a) | >= 0.70 | < 0.70 |
| Composite reliability (rho_c) | >= 0.70 | < 0.70 |
| Average variance extracted (AVE) | >= 0.50 | < 0.50 |
Construct reliability and validity results help evaluate whether the indicators consistently and adequately measure their construct. Cronbach's alpha, composite reliability (rho_a), and composite reliability (rho_c) assess internal consistency reliability. Values of 0.70 or higher indicate satisfactory reliability; note that very high values (above 0.95) may indicate redundant indicators. The average variance extracted (AVE) assesses convergent validity. AVE values of 0.50 or higher indicate that the construct explains at least half of the variance of its indicators.
Discriminant Validity and Collinearity
| Criterion | Green | Black | Red |
|---|---|---|---|
| Heterotrait-monotrait ratio (HTMT) | <= 0.85 | > 0.85 and <= 0.90 | > 0.90 |
| Variance inflation factor (VIF) | <= 3.00 | > 3.00 and <= 5.00 | > 5.00 |
HTMT values of 0.85 or lower are commonly used as evidence of discriminant validity, that is, that constructs are empirically distinct (Henseler, Ringle, & Sarstedt, 2015). HTMT values above 0.90 typically indicate a lack of discriminant validity. VIF values of 3.00 or lower indicate that collinearity among indicators or predictor constructs is not a critical issue; values above 5.00 indicate critical collinearity problems. Values marked in black or red should be reviewed carefully because they may indicate potential overlap between constructs or collinearity issues.
Bootstrapping Results
These thresholds apply to PLS-SEM bootstrapping, PLSc-SEM bootstrapping, bootstrap multigroup analysis (MGA), and PLSc-SEM bootstrap MGA.
| Criterion | Green | Red |
|---|---|---|
| p value | <= 0.05 | > 0.05 |
In bootstrapping reports, SmartPLS marks p values of 0.05 or lower in green. These values indicate statistical significance at the 5% level. Depending on the study, other significance levels (e.g., 1% or 10%) may be appropriate and should be justified. Always interpret statistical significance together with effect size, bootstrap confidence intervals, theory, and practical relevance.
PLSpredict Results
The following thresholds apply to the manifest variable (MV) prediction summary in PLSpredict (Shmueli et al., 2019).
| Criterion | Green | Red |
|---|---|---|
| Q²_predict | >= 0 | < 0 |
| PLS-SEM_RMSE compared with LM_RMSE | PLS-SEM_RMSE < LM_RMSE | PLS-SEM_RMSE >= LM_RMSE |
| PLS-SEM_MAE compared with LM_MAE | PLS-SEM_MAE < LM_MAE | PLS-SEM_MAE >= LM_MAE |
PLSpredict evaluates the model's out-of-sample predictive power. A Q²_predict value of zero or higher indicates that the PLS-SEM model outperforms the naive indicator-mean benchmark for the respective indicator. For the root mean squared error (RMSE) and the mean absolute error (MAE), lower prediction errors indicate better predictive performance. SmartPLS marks the PLS-SEM value in green when it is lower than the corresponding linear model (LM) benchmark. The share of indicators for which the PLS-SEM model outperforms the LM benchmark indicates whether the model has high, medium, low, or no predictive power.
Permutation MGA and MICOM Results
These thresholds apply to permutation multigroup analysis (MGA), PLSc-SEM permutation MGA, and measurement invariance of composite models (MICOM).
| Criterion | Green | Red |
|---|---|---|
| Permutation MGA p value | <= 0.05 | > 0.05 |
| MICOM Step 2: compositional invariance p value | >= 0.05 | < 0.05 |
| MICOM Step 3a and Step 3b p values | <= 0.05 | > 0.05 |
Permutation MGA helps assess whether model estimates differ significantly across groups. MICOM evaluates measurement invariance, which is an important prerequisite for meaningful group comparisons (Henseler, Ringle, & Sarstedt, 2016).
The direction of the threshold differs across MICOM steps because the underlying statistical hypotheses differ. In MICOM Step 2, the null hypothesis states that compositional invariance holds. A p value of 0.05 or higher therefore supports measurement invariance and is marked in green, because the null hypothesis of invariance is not rejected. In Steps 3a and 3b, in contrast, significant differences in the equality of composite means and variances (p <= 0.05) indicate that only partial measurement invariance is established.
Important Notes
The SmartPLS result colors are intended as guidance for common reporting situations. Researchers should always consider theory, measurement quality, model complexity, sample size, data characteristics, confidence intervals, effect sizes, and predictive relevance.
When reporting results, describe the actual values in addition to the color coding. For example, instead of writing only that construct reliability was green, report Cronbach's alpha, rho_a, rho_c, and AVE values and explain what they imply for the measurement model.
Frequently Asked Questions
Why are some SmartPLS results green, black, or red?
SmartPLS uses colors to highlight whether values meet commonly recommended PLS-SEM thresholds. Green indicates that the value meets the threshold, black indicates a value that requires attention but is not necessarily problematic, and red indicates that the value does not meet the recommended threshold.
Do the colors replace the PLS-SEM evaluation guidelines?
No. The colors are a quick visual aid. Researchers should still evaluate the complete measurement and structural model using the relevant PLS-SEM guidelines and the requirements of their research context.
Can I still use my model if a value is red?
Possibly. A red value signals that a commonly recommended threshold is not met, but it does not automatically invalidate the model. Inspect the affected indicators or constructs, consider established alternative thresholds (for example, outer loadings of 0.60 or higher in exploratory research), and justify your decision based on theory, measurement quality, and the study context.
What is an acceptable HTMT value in PLS-SEM?
HTMT values of 0.85 or lower are commonly used as evidence of discriminant validity. Values between 0.85 and 0.90 may be acceptable when constructs are conceptually similar, but should be interpreted carefully. Values above 0.90 are typically treated as problematic for discriminant validity.
Is a VIF above 3 acceptable in PLS-SEM?
VIF values of 3.00 or lower are ideal. Values between 3.00 and 5.00 may indicate collinearity issues and require careful inspection. Values above 5.00 indicate critical collinearity problems among indicators or predictor constructs.
Why is AVE green when it is 0.50 or higher?
AVE values of 0.50 or higher indicate that the construct explains at least 50% of the variance of its indicators. This is commonly used as evidence of convergent validity in reflective measurement models.
Which significance level does SmartPLS use for bootstrapping p values?
SmartPLS marks p values of 0.05 or lower in green, corresponding to the widely used 5% significance level. If your study relies on a different significance level (for example, 1% or 10%), interpret the colors accordingly and report your chosen level.
Do these thresholds apply to formative measurement models?
No. Criteria such as Cronbach's alpha, composite reliability, AVE, outer loadings, and HTMT address reflective measurement models. Formative measurement models are evaluated based on convergent validity (redundancy analysis), indicator collinearity (VIF), and the statistical significance and relevance of outer weights.
References
Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2nd ed.). Lawrence Erlbaum.
Dijkstra, T. K., & Henseler, J. (2015). Consistent partial least squares path modeling. MIS Quarterly, 39(2), 297–316. https://doi.org/10.25300/MISQ/2015/39.2.02
Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18(1), 39–50. https://doi.org/10.1177/002224378101800104
Hair, J. F., Hult, G. T. M., Ringle, C. M., & Sarstedt, M. (2027). A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM) (4th ed.). Sage.
Hair, J. F., Sarstedt, M., Ringle, C. M., & Gudergan, S. P. (2024). Advanced Issues in Partial Least Squares Structural Equation Modeling (PLS-SEM) (2nd ed.). Sage.
Henseler, J., Ringle, C. M., & Sarstedt, M. (2015). A new criterion for assessing discriminant validity in variance-based structural equation modeling. Journal of the Academy of Marketing Science, 43(1), 115–135. https://doi.org/10.1007/s11747-014-0403-8
Henseler, J., Ringle, C. M., & Sarstedt, M. (2016). Testing measurement invariance of composites using partial least squares. International Marketing Review, 33(3), 405–431. https://doi.org/10.1108/IMR-09-2014-0304
Ringle, C. M., Wende, S., & Becker, J.-M. (2024). SmartPLS 4. SmartPLS. https://www.smartpls.com/
Shmueli, G., Sarstedt, M., Hair, J. F., Cheah, J.-H., Ting, H., Vaithilingam, S., & Ringle, C. M. (2019). Predictive model assessment in PLS-SEM: Guidelines for using PLSpredict. European Journal of Marketing, 53(11), 2322–2347. https://doi.org/10.1108/EJM-02-2019-0189
