R Square in SmartPLS: How to Read, Evaluate, and Report It
What R Square is used for in SmartPLS
R Square, usually written as R², shows the proportion of variance in a dependent variable explained by the independent variables in the model. When you open the SmartPLS results, you will mainly see R Square for endogenous latent variables, meaning variables with arrows pointing into them from one or more other variables.
For example, a model may include three independent variables, Service Quality, Perceived Value, and Trust, that jointly explain Purchase Intention. If the R² of Purchase Intention is 0.52, the independent variables in the model explain 52% of the variance in Purchase Intention. The remaining variance comes from factors not included in the model and measurement error.
R Square does not tell you whether a hypothesis is supported. To determine whether a path is statistically significant, you also need to read Path Coefficients, T Statistics, and P Values in the bootstrapping results. R² answers how much the model explains, while the path coefficient and p-value help you assess each specific relationship.
If you are still getting used to the software, you can also read about SmartPLS, Partial Least Squares, and SEM to understand where R² fits into the full PLS-SEM procedure. When using these links, remove the space after the opening parenthesis to create valid Markdown syntax: SmartPLS, Partial Least Squares, and SEM.
Preparing the model and data
Before running R², separate the two parts of the model. The measurement model describes the relationship between latent variables and indicators, such as PU1, PU2, and PU3 measuring the Perceived Usefulness construct. The structural model describes relationships between latent variables, such as Perceived Usefulness affecting Usage Intention.
R² appears only for endogenous variables in the structural model. An exogenous variable with no arrow pointing into it will not have an R² to interpret. Therefore, if you do not see R Square for a variable, check the arrow direction and the role of that variable in the model.
Use consistent indicator names
Indicator names should be short, use no diacritics, and contain no spaces, for example PU1, PU2, PU3, BI1, and BI2. Each column in the data must correspond to one indicator in the SmartPLS model. Do not rename a column after building the model unless you have checked the data links again.
Prepare the CSV file
SmartPLS 4 commonly imports data from a .csv file. The first row contains variable names, and each following row represents one respondent. Data cells should contain numbers, such as Likert-scale values from 1 to 5. Do not place notes, measurement units, or long item wording in the header row.
You need to check missing data, values outside the measurement range, and reverse-coded items before importing the file. If an indicator was designed in the opposite direction, recode it according to the questionnaire's coding rules. R² can be affected by incorrect input data even when the model and analysis procedure are correct.
Check the measurement model first
Assess outer loading, reliability, and convergent validity before drawing conclusions about the structural model. For measures such as CR and AVE, you can refer to the explanation of what AVE is. If the scale is not stable, R² may still appear for an endogenous variable, but the model interpretation will lack a sound basis.
Steps in SmartPLS 4
The steps below apply to SmartPLS 4. Some menu locations in SmartPLS 3 differ from version 4, so do not compare button names mechanically across the two versions.
Step 1: Create a project and import the data
Open SmartPLS 4, create a new project, and choose the option to import a data file. Select the .csv file, then check the preview of the column names and data types. If an indicator name is truncated or duplicated, or SmartPLS reads Likert data as text, correct the original file and import it again.
Step 2: Create the PLS-SEM model
Create a new model, drag the latent variables into the workspace, and assign the relevant indicators to each construct. Use arrows to connect the variables according to your research hypotheses. For example, connect Service Quality to Satisfaction if hypothesis H1 tests the effect of Service Quality on Satisfaction.
Check the arrow direction again. If you connect the path in the wrong direction, the variable you intend to explain may become exogenous and will not have the expected R². You should also check VIF before interpreting the path coefficients. See also what VIF is.
Step 3: Run the PLS-SEM Algorithm
Select the model, click Calculate, and then select PLS-SEM Algorithm. Keep the default settings if your proposal does not require another configuration, then run the analysis. The first run allows you to inspect outer loading, path coefficients, and R Square.
In the results, open the Quality Criteria group and find the R Square table. You can also view R² directly on the diagram if you turn on the appropriate results-display mode.
Step 4: Run Bootstrapping
Select Calculate and then Bootstrapping. For a thesis study, you should report the number of bootstrap subsamples clearly. The current PLS-SEM procedure commonly uses 5,000 bootstrap subsamples according to (Hair et al., 2022).
After the run, inspect Path Coefficients, T Statistics, P Values, and the confidence intervals. Bootstrapping does not change the meaning of R², but it helps you assess the stability and statistical significance of the paths that produce the R² value.
Step 5: Run Blindfolding or PLSpredict when appropriate
Q² is an additional measure for assessing predictive relevance through the blindfolding logic. In SmartPLS 4, depending on the settings and type of model, you may use the appropriate prediction function rather than looking only at R². A Q² value greater than 0 is commonly used as an indication that the model has some predictive relevance, according to (Stone, 1974).
Do not combine Q² and R² into one measure. R² focuses on explanatory power in the sample, while Q² adds a perspective on predictive relevance.
Reading the R Square results table
The table below is illustrative output, not the result of a specific study. The column names are presented in the way SmartPLS commonly displays them in the Quality Criteria group.
| Endogenous variable | R Square | R Square Adjusted | Illustrative interpretation |
|---|---|---|---|
| Satisfaction | 0.536 | 0.529 | The model explains approximately 53.6% of the variance in Satisfaction |
| Purchase Intention | 0.418 | 0.409 | The model explains approximately 41.8% of the variance in Purchase Intention |
| Usage Behavior | 0.267 | 0.258 | The model explains approximately 26.7% of the variance in Usage Behavior |
The R Square column is the unadjusted proportion explained. The R Square Adjusted column adjusts the value for the number of predictors and the sample size. When a model includes many independent variables, Adjusted R² can help you avoid overstating explanatory power simply because more variables were added to the model.
For example, an R² of 0.536 can be written as 53.6%. You need to identify which variable is being explained and which variables are predicting it. The sentence “the model has an R² of 0.536” lacks information if it does not state which endogenous variable the R² belongs to.
R² is not the same as r in correlation analysis. When students search for “r and r square”, they are often trying to distinguish the correlation coefficient r from the coefficient of determination R². In simple regression, R² may be related to the square of r, but in a multiple-variable model and PLS-SEM, you should not take one path coefficient and square it as a substitute for the R² of an endogenous variable.
R Square thresholds and related measures
The question “what is a good r square” has no answer separate from the industry context, model, and research objective. In PLS-SEM, commonly used reference levels for R² are 0.75 substantial, 0.50 moderate, and 0.25 weak (Hair et al., 2011). These are interpretation benchmarks, not automatic conditions for accepting or rejecting a model.
| Measure | Reference level | Canonical source |
|---|---|---|
| R² | 0.75 substantial, 0.50 moderate, 0.25 weak | (Hair et al., 2011) |
| Outer loading | 0.70 or above | (Chin, 1998) or (Hair et al., 2022) |
| CR | 0.70 or above | (Fornell and Larcker, 1981) |
| AVE | 0.50 or above | (Fornell and Larcker, 1981) |
| HTMT | Below 0.85, or below 0.90 for conceptually close constructs | (Henseler et al., 2015) |
| VIF | Below 5 | (Hair et al., 2019) |
| f² | 0.02 small, 0.15 medium, 0.35 large | (Cohen, 1988) |
| Q² | Greater than 0 | (Stone, 1974) or (Hair et al., 2022) |
You should read R² together with f². An independent variable may increase R² while making only a very small individual contribution. f² helps you assess the change in R² when that predictor is removed from the model. At the same time, check VIF for multicollinearity before interpreting the paths.
CR, AVE, HTMT, and outer loading belong to the measurement model assessment. R², f², Q², and the path coefficient belong to the structural model assessment. When writing the results, present these in two stages, consistent with the measurement-model-first and structural-model approach described in (Anderson and Gerbing, 1988).
SmartPLS does not use CFI, TLI, and RMSEA as the main thresholds for PLS-SEM. The reference points of CFI and TLI at 0.90 or above, and RMSEA and SRMR at no more than 0.08, are associated with CB-SEM according to (Hu and Bentler, 1999). If you are running SmartPLS, do not place the CB-SEM fit indices in the R² table to claim that the model meets the standard.
How to report this in your thesis
A clear order is to introduce the endogenous variables, present the R² and Adjusted R² table, interpret the explained proportion, and then compare it with f², Q², and the path-significance results. Do not simply paste a SmartPLS screenshot without naming the variables, reporting the values, and explaining the meaning of each column.
How to write this in your thesis: “The PLS-SEM results show that the R² value of [endogenous variable name] is [R² value], meaning that the variables [list predictor variables] explain approximately [percentage value]% of the variance in [variable name]. Based on the reference levels of Hair et al. (2011), this R² level is considered [weak/moderate/substantial].”
If Adjusted R² is available, you can add: “The R Square Adjusted value is [value], which is [difference] lower than R Square, indicating that the explained variance after adjustment for the number of predictors is [percentage value]%.” Include this statement only when your output actually contains the two corresponding values.
The Chapter 4 table should include at least the endogenous variable name, R², and Adjusted R². If your thesis requires an assessment of predictive relevance, add Q² in a separate table or column, but state clearly that it is a different measure from R².
Common mistakes
Calling R² an effect percentage
R² is the proportion of variance explained by the model, not the percentage effect of each independent variable. To say which variable has the strongest effect, you need to examine the path coefficient, statistical significance, and f². Do not divide R² by the number of variables.
Using the R² of an exogenous variable
An exogenous variable with no predictor pointing into it will not have an R². If you try to assign an R² to that variable, check the diagram and the research hypotheses again.
Looking only at R² and declaring the model good
A high R² alone does not prove that the scale is adequate, that the relationships are significant, or that the model has good predictive relevance. Read it together with outer loading, CR, AVE, HTMT, VIF, the path coefficient, p-value, and Q², with each measure interpreted according to its role.
Confusing R Square with Adjusted R Square
The two columns may be close, but they are not identical. When extracting the results, use the exact SmartPLS column names and do not replace Adjusted R² with R² simply because one value is easier to interpret.
Bringing CB-SEM thresholds into a SmartPLS paper
CFI, TLI, and RMSEA commonly appear in CB-SEM literature. Including them in a PLS-SEM assessment without explaining the method can lead the committee to ask which framework you are using.
Removing variables only to increase R²
Removing an indicator or predictor must be based on theoretical reasoning, measurement quality, and diagnostic results. Do not run many model versions and keep only the one with the highest R². Save each model version and record the reason for every adjustment.
Frequently asked questions
What is R Square in SmartPLS?
R Square is the proportion of variance in an endogenous variable explained by the predictors in the model. In SmartPLS, you find this measure in the Quality Criteria group, in the R Square table.
R² does not replace the p-value and should not be used alone to decide whether a hypothesis is supported. It describes the explanatory power of the full group of predictors for each endogenous variable.
What is a good R Square?
Using commonly cited reference levels in PLS-SEM, an R² of 0.75 is considered substantial, 0.50 moderate, and 0.25 weak (Hair et al., 2011). You need to place the number in the context of the industry, research objective, and model complexity.
A low R² does not automatically make a thesis wrong, just as a high R² does not automatically prove that every hypothesis is supported. Read f², Q², and the bootstrapping results as well.
What is the difference between R Square and Adjusted R Square?
R Square is the unadjusted explained value. Adjusted R Square adjusts this value for the number of predictors and the sample size, so it is usually lower than or close to R Square.
When a model has many independent variables, Adjusted R² allows you to report explanatory power more cautiously. In the results table, retain both columns if SmartPLS provides them.
Is R Square the square of r?
In some simple regression situations, R² is related to the square of the correlation coefficient r. However, in a multiple-variable PLS-SEM model, the R² of an endogenous variable is calculated from the full model predicting that variable, not from one individual path coefficient.
Therefore, do not take r or one path and square it as a substitute for R² in the SmartPLS results table.
What should I do if R Square is low?
First, check the data, arrow directions, measurement-model quality, VIF, and predictors against the theoretical basis of the model. Then inspect f² and Q² to see which variables contribute and whether the model shows predictive relevance.
If R² remains low but the model is theoretically appropriate and the related tests are reported fully, you can present the limited explanatory power honestly. Do not add or remove variables simply to reach a desired value.
Open your SmartPLS 4 project, identify each endogenous variable, export the R Square and R Square Adjusted tables, and compare them with f², Q², and the bootstrapping results before writing the interpretation. If you need to run the full analysis on a .sav or .csv file, use M4 data analysis.