SmartPLS Step by Step: PLS-SEM for Your Thesis

By the Mastermind PhD team · Last updated: 3 November 2026 · Editorial policy

Your model has five constructs, each measured by four to six questionnaire items, and your supervisor has suggested PLS-SEM. Where do you start? SmartPLS is the most widely used software for partial least squares structural equation modeling in business, management and information systems theses. This SmartPLS tutorial follows the order in which the analysis is done and reported: prepare the data, draw the model, assess the measurement model, then assess the structural model.

This article is part of our guide How to Write a Master’s Thesis by the Mastermind PhD team and relates to our Statistical Analysis service. Sources last checked: November 2026.

The thresholds quoted below come from the guidelines of Hair and colleagues, which most examiners in this field use as a reference.

When is PLS-SEM the right choice for a thesis?

PLS-SEM suits a thesis when the aim is to explain or predict key target constructs, when the model is complex (many constructs or indicators), when the sample is modest, or when some constructs are formative. Covariance-based SEM is more appropriate when the aim is to confirm a well-established theory and test overall model fit (Hair et al., 2019).

If your supervisor expects fit indices such as CFI and RMSEA, discuss the choice before you start; our comparison of AMOS and SmartPLS covers this decision in more detail.

Step 1: Prepare the data before this SmartPLS tutorial begins

Clean the data first in SPSS or Excel: remove incomplete or careless responses, check that reverse-worded items have been recoded, and give each item a short name without spaces (LEAD1, LEAD2, SAT1). Code missing values with one number that cannot occur in the data, such as -99. Save the file as CSV, which SmartPLS imports reliably.

Check the minimum sample size. The ten-times rule is no longer recommended. A better approach is the inverse square root method (Kock & Hadaya, 2018): with a 5% significance level, the minimum sample is (2.486 ÷ smallest expected path coefficient)². For a smallest path of 0.20 this gives (2.486 ÷ 0.20)² ≈ 155 cases. Report the method and the result in the methodology chapter.

📱 Ask a statistician about your data — Send a short description of your instrument and sample size on WhatsApp and get a suitable analysis plan. Every analysis comes with the raw SPSS file and a results report.

Step 2: Create the project and import the data

1. Open SmartPLS 4, create a new project and give it a name.
2. Import the data file. In the data view, set the missing value marker (for example -99) so that those cells are treated as missing.
3. Check the indicators list: each item should appear with its name, and the descriptive statistics (mean, standard deviation, missing values) should match what you saw in SPSS.

Step 3: Draw the path model

At this point in the SmartPLS tutorial you turn your hypotheses into a diagram. Create a new PLS-SEM model from the data file. Place each construct as a latent variable, rename it, and drag its items onto it. Draw arrows from predictors to outcomes exactly as your hypotheses state; every arrow is a hypothesis you must report. Reflective constructs have arrows from the construct to the items, which is the default. Switch a construct to formative only when its items define it rather than reflect it, and justify that choice in the text.

Step 4: Run the PLS-SEM algorithm

Choose Calculate, then PLS-SEM algorithm, keep the default path weighting scheme, and run. The results report gives outer loadings, construct reliability and validity, discriminant validity, path coefficients and R². In this SmartPLS tutorial the next two steps read these results in a fixed order: measurement model first, structural model second.

Step 5: Assess the measurement model

This is the step of the SmartPLS tutorial that examiners read most closely. For reflective constructs, the criteria in Hair et al. (2019) are:

The heterotrait-monotrait ratio (HTMT) was proposed by Henseler, Ringle and Sarstedt (2015) because the older Fornell-Larcker criterion often fails to detect discriminant validity problems. Many examiners still ask for the Fornell-Larcker table, so report both. If you delete a weak item, say which one and why, and re-run the algorithm.

Step 6: Bootstrap and assess the structural model

Choose Calculate, then Bootstrapping. Use a large number of subsamples (10,000 is commonly recommended for final results), percentile confidence intervals, and a two-tailed test unless your hypotheses are directional and your field accepts one-tailed tests. Then assess the structural model in this order:

• Collinearity: inner VIF values ideally below 3.
• Path coefficients: the size and sign of each path, its t value, p value and bootstrap confidence interval. A path is significant when the interval excludes zero.
• Explained variance: R² values of 0.75, 0.50 and 0.25 are often described as substantial, moderate and weak, though acceptable levels depend on the discipline.
• Effect size f²: 0.02, 0.15 and 0.35 as small, medium and large.
• Predictive power: PLSpredict, reporting Q²predict and comparing prediction errors with a linear benchmark.

For mediation, read the specific indirect effects table from the same bootstrap run. If your model has no mediators, this SmartPLS tutorial step ends with the path table.

Step 7: Report the results in the thesis

The SmartPLS tutorial ends where examiners begin: the write-up. A clear results chapter usually contains:

1. A sample and data-screening paragraph, including the sample size justification.
2. A measurement model table with loadings, alpha, composite reliability and AVE for every construct.
3. A discriminant validity table (HTMT, and Fornell-Larcker if requested).
4. A structural model figure with path coefficients and R² values.
5. A hypothesis table: path, coefficient, t, p, confidence interval, f² and the decision (supported or not supported).

Mention the software and version (SmartPLS 4), the algorithm settings and the number of bootstrap subsamples, and cite the software in the format given on the “How to cite SmartPLS” page of the official site.

Mistakes that weaken a PLS-SEM chapter

• Reporting only path significance and ignoring the measurement model.
• Deleting many items to push AVE above 0.50 without considering content validity.
• Using the ten-times rule as the only sample size justification.
• Treating SRMR as decisive model fit evidence in the way CFI is used in CB-SEM.
• Calling all relationships “effects” in a cross-sectional design without noting the causal assumption.
• Running bootstrapping with very few subsamples.

Followed step by step, this SmartPLS tutorial gives you a results chapter in the order examiners expect, and you should be able to explain every number in it at your defense.

📱 Ask on WhatsApp now — 5 years of experience across all disciplines and 14 academic services. Two-stage payment (80/20); revisions are free until supervisor approval, within 30% of the chapter.

Frequently asked questions

Is SmartPLS free for students?

SmartPLS offers a time-limited free trial and several paid license types, and the conditions change over time. Check the current pricing page on the official SmartPLS website, and ask your university whether it holds a license, before you plan your analysis.

What sample size does this SmartPLS tutorial assume?

None in particular. Use the inverse square root method or a power analysis based on the smallest path you expect. A smallest path of 0.20 at the 5% level needs about 155 cases.

Can I use Likert-scale data in SmartPLS?

Yes. Likert items are commonly treated as approximately interval indicators in PLS-SEM, and the method does not require normally distributed data. Report the scale format in the methodology.

Do I need to check common method bias?

If all constructs come from the same questionnaire, examiners often ask. A full collinearity test, with all VIF values below 3.3, is a common approach in PLS-SEM; procedural remedies in the questionnaire design should also be described.

What is the difference between rho_a, rho_c and Cronbach’s alpha?

All three measure internal consistency. Cronbach’s alpha assumes equal loadings and tends to underestimate reliability, rho_c (composite reliability) tends to overestimate it, and rho_a usually lies between the two. Reporting all three is common.

Sources

• Hair, J. F., Hult, G. T. M., Ringle, C. M., & Sarstedt, M. (2022). A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM) (3rd ed.). SAGE.
• Hair, J. F., Risher, J. J., Sarstedt, M., & Ringle, C. M. (2019). When to use and how to report the results of PLS-SEM. European Business Review, 31(1), 2–24. https://doi.org/10.1108/EBR-11-2018-0203
• 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
• Kock, N., & Hadaya, P. (2018). Minimum sample size estimation in PLS-SEM: The inverse square root and gamma-exponential methods. Information Systems Journal, 28(1), 227–261. https://doi.org/10.1111/isj.12131
• 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
• SmartPLS. SmartPLS documentation. https://www.smartpls.com/documentation/

Sources last checked: November 2026. This article is general educational guidance and does not replace your university regulations or your supervisor’s instructions. How we prepare our content.