What Is SERVQUAL? A Service Quality Measurement Model
What Is the SERVQUAL Model?
SERVQUAL is a model for measuring service quality based on the gap between customers' expectations and their actual perceptions after using a service. You usually use the model when your research asks which aspects shape customers' evaluation of service quality and whether those aspects affect satisfaction or the intention to continue using the service.
SERVQUAL originated from the service quality scale developed by Parasuraman, Zeithaml, and Berry. The original scale covered 10 dimensions and was later reduced to the five dimensions commonly used in quantitative research: tangibles, reliability, responsiveness, assurance, and empathy. The foundational scale is usually cited as (Parasuraman et al., 1988).
In measurement terms, respondents usually rate multiple items on a Likert scale. You can measure expectations and perceptions separately and then calculate the gap, or use perception items about service quality directly, depending on your research design. The Likert scale originated from the attitude measurement technique introduced by (Likert, 1932).
If you need a short answer to the search query “what is SERVQUAL,” remember this: SERVQUAL is a theoretical framework and measurement scale for evaluating service quality through the attributes customers perceive during a transaction.
You can also see the Research Model topic to place SERVQUAL within the process of building a research model, hypotheses, and questionnaire.
The Dimensions of the SERVQUAL Model
The five dimensions below are the constructs you need to turn into latent variables in your research model. Each dimension is usually measured with multiple items rather than one general question.
| Dimension | What it measures | Example item areas |
|---|---|---|
| Tangibles | Facilities, equipment, employee appearance, and the visible signs of the service | Service premises, equipment, materials, uniforms |
| Reliability | The ability to keep commitments and provide the service consistently | Providing the correct service on time and processing requests accurately |
| Responsiveness | The willingness to support customers and respond to them | Fast responses, timely guidance, proactively resolving requests |
| Assurance | Knowledge, courtesy, and the ability to create a sense of confidence | Employee expertise, appropriate communication, information security |
| Empathy | Personal attention and understanding of customers' individual needs | Convenient service hours, listening, attention to each customer |
When you design the questionnaire, write the items for the specific context. For example, “employees have professional expertise” may fit a bank, hospital, university, or e-commerce platform, but the wording and the person being evaluated will differ.
A common mistake is to copy the wording of a foreign scale directly into a Vietnamese questionnaire. Adjust each item so respondents understand the service you are studying. You should then check how respondents interpret the wording and whether the questionnaire is suitable before official data collection.
SERVQUAL is often used as a set of independent variables to explain satisfaction, perceived value, or loyalty. In another model, service quality may be a mediator between operational factors and customer behavioral outcomes. The specific relationships must follow your topic and selected literature. Do not assume that all five dimensions have the same effect.
The Original Model and Common Extensions
The original model developed by Parasuraman et al. focused on customers' perceived service quality. Its five dimensions are easy to convert into a questionnaire, so the model has been widely used in banking, education, healthcare, retail, tourism, and public services.
In the original model, service quality is formed from customers' evaluations of each dimension. One implementation measures expectations and perceptions separately and then calculates the difference. Another measures perceptions after the service experience. Choose one approach and explain it clearly in the methods chapter because the two approaches produce different questionnaires and different interpretations of the data.
SERVPERF is a commonly compared approach. SERVPERF focuses on perceived service performance instead of asking customers to rate both expectations and perceptions. If your research is considering the two models, read more about the SERVPERF model to compare questionnaire structure, respondent burden, and how measurement scores are interpreted.
Another related approach is the five-gap service quality model. It helps analyse the causes of gaps between expectations, service design, the delivery process, and customers' perceptions. You can consult the five-gap service quality model when your topic aims to explain service quality from the organisation's side rather than only measuring customers' perceptions.
You can also extend SERVQUAL by adding outcome variables such as satisfaction, repurchase intention, loyalty, or word of mouth. The added variables still need a theoretical argument and suitable scales. Do not add a variable simply because the model has too few arrows.
When comparing SERVQUAL and SERVPERF, focus on four points: whether you measure expectations or perceptions only, how long the questionnaire is, whether respondents have enough experience to evaluate the service, and which model fits your research question. A shorter model may still be unsuitable if it removes an aspect your topic needs to explain.
Common Scales for Each Construct
The foundational SERVQUAL scale was published by Parasuraman et al. in research on service quality (Parasuraman et al., 1988). In your thesis, state the source, the context in which the scale was adapted, the number of items, and the coding used for each item.
| Construct | Suggested scale source | Use in the questionnaire |
|---|---|---|
| Tangibles | (Parasuraman et al., 1988) | Measure facilities, equipment, and the visible features of the service |
| Reliability | (Parasuraman et al., 1988) | Measure the ability to keep commitments and provide the service accurately |
| Responsiveness | (Parasuraman et al., 1988) | Measure response speed and willingness to provide support |
| Assurance | (Parasuraman et al., 1988) | Measure expertise, attitude, and the sense of confidence created by employees |
| Empathy | (Parasuraman et al., 1988) | Measure personal attention and understanding of customers |
| Satisfaction | Select a scale that fits the context and sources for your topic | Measure the overall evaluation after using the service |
| Loyalty or intention to continue | Select a scale that fits the context and sources for your topic | Measure the intention to return, recommend, or continue using the service |
The table above is a source-selection map, not a confirmed translation for every industry. When adapting a scale, preserve the construct's meaning but change context-specific terms. “Employees provide the service as promised” needs more specific wording if respondents are evaluating an app, clinic, or digital bank.
In your data file, use consistent codes such as TAN1, TAN2 for Tangibles, REL1, REL2 for Reliability, RES1, RES2 for Responsiveness, ASS1, ASS2 for Assurance, and EMP1, EMP2 for Empathy. Short variable names make it easier to read the Reliability Statistics, Rotated Component Matrix, and model results.
After collecting data, the usual process includes checking missing data, recoding reverse-coded items if necessary, running Cronbach's Alpha, running EFA, or assessing the measurement model in SmartPLS. Cronbach's Alpha measures the internal consistency of a scale and is based on (Cronbach, 1951). Read the Corrected Item-Total Correlation together with the item's content. Do not delete an item simply to make one number look better.
Applying SERVQUAL to Your Topic
Suppose your topic is: “Factors of service quality affecting customer satisfaction at Bank X.” The five SERVQUAL dimensions can then be independent variables, while satisfaction is the dependent variable.
The conceptual model could be written as follows:
- Tangibles affect satisfaction.
- Reliability affects satisfaction.
- Responsiveness affects satisfaction.
- Assurance affects satisfaction.
- Empathy affects satisfaction.
Write the hypotheses as complete sentences in the theoretical foundation chapter:
H1: Tangibles have a positive effect on customer satisfaction at Bank X.
H2: Reliability has a positive effect on customer satisfaction at Bank X.
H3: Responsiveness has a positive effect on customer satisfaction at Bank X.
H4: Assurance has a positive effect on customer satisfaction at Bank X.
H5: Empathy has a positive effect on customer satisfaction at Bank X.
This is an illustrative model, not the result of an actual study. For your own topic, replace Bank X with the real context, identify the correct target respondents, and check whether the literature supports each relationship.
If your topic only asks how customers evaluate service quality, you can stop at a measurement model with the five dimensions. If it asks which factors affect satisfaction, you need to add a dependent variable and test the structural model using regression or PLS-SEM.
You can adapt the following paragraph for your thesis: “Based on the SERVQUAL model of (Parasuraman et al., 1988), this study proposes five dimensions of service quality, namely [list the dimensions], which affect [dependent variable] in the context of [target respondents and research setting]. The hypotheses are developed to test [the direction of the effect or relationship] between the constructs.”
You also need to decide whether to measure expectations. If you measure both expectations and perceptions, the questionnaire is longer and respondents must rate the same attribute twice. If you measure perceptions, the questionnaire is shorter, but you must explain why you selected this approach and how the variable score is calculated.
How to Test the Model with SPSS or SmartPLS
SPSS is suitable when your model uses composite variables and you need Cronbach's Alpha, EFA, correlations, linear regression, or mediation testing with PROCESS. Start with Analyze > Scale > Reliability Analysis, move the items for each dimension into Items, then select Statistics and check Scale if item deleted.
Next, run EFA through Analyze > Dimension Reduction > Factor. Check KMO and Bartlett's Test, Total Variance Explained, and Rotated Component Matrix. In EFA, check whether the items load on the expected factor, whether there are high cross-loadings, and whether the rotated structure still fits SERVQUAL.
For regression, go to Analyze > Regression > Linear. Put satisfaction in Dependent and the five SERVQUAL dimensions in Independent(s). The Coefficients table reports the coefficient, t, and Sig., while Model Summary reports R Square. Also check VIF, residuals, and diagnostic plots before drawing conclusions.
SmartPLS is suitable when your model has latent variables, multiple items, mediation relationships, or requires simultaneous assessment of the measurement model and structural model. Import the .csv file, draw the five SERVQUAL constructs and the dependent variable, connect the hypothesised paths, and then run PLS-SEM Algorithm and Bootstrapping in SmartPLS 4.
In SmartPLS, read Outer Loadings, Construct Reliability and Validity, Discriminant Validity, Path Coefficients, R Square, f Square, Collinearity Statistics, and the bootstrapping results. Separate the procedure into two parts: assess the measurement model first, then interpret the structural model. This two-step approach is presented in (Anderson and Gerbing, 1988), and the current PLS-SEM procedure is presented in (Hair et al., 2022).
Common thresholds in PLS-SEM include outer loading from 0.7, CR from 0.7, AVE from 0.5, and HTMT below 0.85 or 0.90 depending on how closely related the constructs are. You can consult (Chin, 1998), (Fornell and Larcker, 1981), and (Henseler et al., 2015) for these thresholds. VIF below 5 is stated in (Hair et al., 2019). When running bootstrapping, the use of 5,000 bootstrap subsamples is recommended in (Hair et al., 2022).
If you use CB-SEM with AMOS, CFI, TLI, RMSEA, and SRMR are model fit indices discussed in (Hu and Bentler, 1999). Do not place these indices in the SmartPLS model assessment section because SmartPLS and CB-SEM use different estimation logic.
Common Errors When Using This Model
Treating the five dimensions as mandatory. SERVQUAL is a foundation for building a model, but your Vietnamese context and service industry may make some items unsuitable. If you remove an item, record the reason based on its content and the test results.
Mixing expectations and perceptions. Some questionnaires use expectation items, while others measure perceptions only. If you mix the two approaches in one model without explaining the decision, your supervisor will ask how the score is calculated and what the variable means.
Translating the scale without adapting the context. Hospital users, bank customers, and e-commerce buyers will not interpret every item in the same way. Check the service name, the people providing the service, the transaction channel, and the evaluation time.
Deleting items only to achieve a higher Alpha. Cronbach's Alpha from 0.7 up is commonly considered acceptable according to (Nunnally, 1978), while an exploratory scale may consider a level from 0.6 according to (Hair et al., 2010). Even so, item deletion must be supported by a theoretical reason, corrected item-total correlation, and factor structure.
Drawing conclusions from correlation coefficients alone. Correlation shows the degree of association between variables, but it is not enough to test hypotheses in a model with multiple independent variables. Read the correct regression table or Path Coefficients, check the p-value or bootstrap confidence interval, and then draw your conclusion.
Reporting illustrative results as real results. Every number in a teaching example only shows how to read the output. When writing Chapter 4, replace it with the data from your own .sav or .csv file and save the cleaned-data version.
Frequently asked questions
What is SERVQUAL and what is it used for?
SERVQUAL is a model and scale for evaluating service quality through five dimensions: tangibles, reliability, responsiveness, assurance, and empathy. It is commonly used to explain satisfaction, loyalty, or the intention to continue using a service.
Adjust the items to fit the industry and target respondents. The foundational source for the five dimensions is (Parasuraman et al., 1988).
How many dimensions does the SERVQUAL model have?
SERVQUAL is usually presented with the following five dimensions after the model was reduced: Tangibles, Reliability, Responsiveness, Assurance, and Empathy. Each dimension should be measured with multiple items so you can assess the scale's reliability and validity.
In some contexts, the EFA results may not preserve the theoretical structure. Report the actual results and explain the adjustment instead of forcing the data into the original five groups.
How should you compare SERVQUAL and SERVPERF?
SERVQUAL measures expectations and perceptions, while SERVPERF focuses on perceived service performance. SERVQUAL can provide a view of the gap, but its questionnaire is usually longer. SERVPERF is more concise when your objective is to measure evaluations after the service experience.
Choose based on your research question, respondent characteristics, survey time, and how previous studies in the same context implemented the models. You can compare the SERVQUAL model and the SERVPERF model before finalising the questionnaire.
Is SPSS or SmartPLS better for SERVQUAL data?
SPSS is suitable for Cronbach's Alpha, EFA, regression, and tests based on composite variables. SmartPLS is suitable when you want to retain the latent-variable structure, assess the outer model and structural model, or test a model with several mediation relationships.
The software does not decide whether your model is correct. Choose the method based on the research design, data type, sample size, and your supervisor's requirements.
What should you do if the SERVQUAL scale has a low Alpha?
First, check the coding, reverse-coded items, missing data, and whether an item from another construct was placed in the same group. Then review Item-Total Statistics, the content of the items, and the EFA results.
You can consider removing an item if you have a clear reason and record each processing round. Do not delete several items in succession only to reach a threshold, because the final scale may lose important content from the construct.
Can SERVQUAL be used to measure satisfaction?
SERVQUAL usually measures service quality, while satisfaction is an outcome construct measured with its own set of items. You can build a model in which the five SERVQUAL dimensions affect satisfaction if you have a theoretical basis and a suitable scale.
You then need to assess the service quality scale and satisfaction scale separately before testing the relationships between them. Do not combine all items into one general scale simply because they appear in the same questionnaire.
Open your data file and list the constructs, variable codes, and scale sources before running Cronbach's Alpha or building the SmartPLS model. If you need to run the analysis on your own .sav or .csv file, you can use M4 analysis from DoThesis.