Service Quality Scale: Sources, Indicators, and Validation

What is a service quality scale, and how is it measured?

You may have a broad concept such as service quality, but SPSS cannot analyze an abstract concept directly. You need to turn that concept into a group of indicators, such as statements that respondents rate on a 5-point or 7-point Likert scale. This group of indicators is called a service quality scale.

In quantitative research, service quality is usually measured through customers' perceptions after they use a service. Indicators may cover reliability, responsiveness, assurance, empathy, and tangibles. These are the five dimensions commonly associated with the SERVQUAL model developed by Parasuraman et al. (1988). You can review the background model in the SERVQUAL model before deciding whether to retain or adapt this structure.

A good scale must represent the concept accurately, use wording that respondents understand, and have the potential to demonstrate reliability, convergent validity, and discriminant validity in the actual data. A scale that has already been published cannot simply be copied into your questionnaire. The industry context, customer type, location, and service delivery method can all change how respondents interpret each indicator.

You can process a service quality scale in SPSS if you use Cronbach's Alpha and EFA, or in SmartPLS if your model includes latent variables and structural relationships between constructs. When you read more about this topic, you can start with Research model topics to check how the scale connects to your model and hypotheses.

Sources of the original scale

The most frequently cited source for service quality is SERVQUAL. Parasuraman et al. (1988) developed the scale to measure customers' perceptions of service quality through several groups of indicators. When you apply it, state clearly whether you use the full structure, part of the structure, or only the measurement approach.

The table above only helps you locate the sources. Read the original study or the materials required by your lecturer to determine the exact number of indicators, wording, and factor structure. Do not state that the scale has a fixed number of indicators until you have checked the original version or a version used in the same context.

If your topic concerns banking, hospitals, schools, e-commerce, or public services, explain why you selected SERVQUAL. For example, some tangibles indicators may fit a hotel but need to be rewritten for a delivery application. This explanation shows the committee that you adapted the method instead of copying a questionnaire.

You should also distinguish a service quality scale from models used to analyze the business environment. SWOT analysis and Porter's Five Forces model can support the context analysis, but they do not replace a quantitative scale that measures customers' perceptions.

Reference set of indicators

The following is an illustrative set of indicators to show how you might code variables in an .sav or .csv file. It is a starting point for adaptation, not a validated translation or a scale validated for your topic. Change the service name, customer group, and specific context before placing these items in your questionnaire.

You can use a 5-point Likert scale, from 1 for strongly disagree to 5 for strongly agree. The Likert scale originates from the attitude measurement technique presented by Likert (1932). In the questionnaire, the instructions must state clearly which experience respondents should evaluate, when that experience occurred, and which service provider they should consider.

Variable codes should be short, consistent, and free of Vietnamese diacritics, such as TC1, TC2, and DU1. In SPSS, Variable View should contain the variable name, variable label, coded values, and appropriate measurement level. In SmartPLS, the column names in the .csv file must match the indicators that you drag into the model.

Do not combine two ideas in one indicator. A statement such as “Employees are friendly and process requests quickly” measures at least two things. If a respondent agrees that employees are friendly but disagrees about processing speed, you will not know which part they evaluated.

Adapting the scale to the Vietnamese context

The first step is to identify the service and the respondents. “Banking service quality” needs different wording from “food delivery service quality.” Respondents must also have used the service recently enough to have a basis for their evaluation.

Next, read each indicator and replace overly general terms with a specific subject. “Service provider” may become the brand name or the type of service, depending on how you preserve neutrality. Avoid technical terms that ordinary customers do not commonly hear.

You need to decide whether to use reverse-coded items. Reverse-coded items can help detect mechanical answering, but they also increase the risk that respondents will misunderstand the statement. If you use a negative statement, reverse-code it before running Cronbach's Alpha. In SPSS, the new value for a scale from 1 to 5 can be created using the rule 6 minus the old value. Record this operation in your data-processing log.

After drafting the questionnaire, check its logic with a small group that fits the research population. The aim is to identify unclear wording, display errors, missing response options, and unusually long completion times. This checks the questionnaire, but it is not evidence that the scale has achieved reliability.

When you collect the official data, check responses with extensive missing data, rows containing the same answer for nearly every variable, and completion times that are too short. Set your data-exclusion criteria before analysis instead of deleting rows that make the results look less attractive. If you use a separate collection tool, fillform.info is DoThesis's own form product for Vietnamese questionnaires. Google Forms is a common alternative.

Testing the reliability and validity of the scale

With an SPSS workflow, the usual order is to check the data, run Cronbach's Alpha for each dimension, run EFA if the research design requires structure exploration, and only then enter suitable factors into regression or a later test. Cronbach's Alpha reflects the internal consistency of a group of indicators, but it does not by itself prove that the scale has the correct theoretical structure.

In SPSS, go to Analyze > Scale > Reliability Analysis. Move the indicators for one dimension, such as TC1, TC2, and TC3, into the Items box. Select Statistics, then tick Scale if item deleted to see how Alpha changes when each indicator is removed. Read both Cronbach's Alpha in the Reliability Statistics table and Corrected Item-Total Correlation in the Item-Total Statistics table.

Cronbach's Alpha from 0.7 upward is commonly considered acceptable according to Nunnally (1978). For a new scale or exploratory research, a level from 0.6 may be considered under Hair et al. (2010). Corrected Item-Total Correlation from 0.3 upward is a commonly cited threshold under Nunnally and Bernstein (1994). Do not delete an indicator only because Alpha if Item Deleted increases slightly. Check whether that indicator is important to the content of the construct.

If you run EFA, go to Analyze > Dimension Reduction > Factor. Open Descriptives and select KMO and Bartlett's test. In Extraction, specify the extraction method and the eigenvalue rule according to your research design. In Rotation, select a rotation method that fits your assumption about whether the factors are independent or correlated. The tables to read are KMO and Bartlett's Test, Total Variance Explained, and Rotated Component Matrix.

A KMO from 0.5 upward and a statistically significant Bartlett's test at a level below 0.05 are criteria commonly cited under Kaiser (1974). The rule of an eigenvalue greater than 1 is associated with Kaiser (1960). Total variance explained from 50% upward and factor loading from 0.5 upward can be referenced to Hair et al. (2010). Place these thresholds alongside theoretical meaning, because EFA output is rarely tidy the first time you run it.

If you use SmartPLS, assess the measurement model before the structural model. The usual outputs include Outer Loadings, Cronbach's Alpha, rho_A, Composite Reliability, AVE, and Discriminant Validity. Outer loading from 0.7 upward, Composite Reliability from 0.7 upward, and AVE from 0.5 upward are thresholds that can be referenced to Chin (1998), Fornell and Larcker (1981), and Hair et al. (2022). HTMT below 0.85, or below 0.90 for closely related constructs, is presented by Henseler et al. (2015).

Illustrative output

The table below is illustrative output, not the result of a real study. The column names simulate how SPSS displays results in the Reliability Statistics and Item-Total Statistics tables.

In this example, all indicators exceed the Corrected Item-Total Correlation threshold of 0.3. You still need to examine the item wording and the number of indicators in each dimension. If an indicator has a low correlation, check data-entry errors, reverse coding, and how respondents understood the statement before deleting it.

How to present the scale in Chapter 3

In Chapter 3, present the source of the scale, the definition of each construct, the expected number of indicators, the Likert scale, and the coding method. A short table can include the indicator code, statement, reference source, and adaptation note. If you translated or rewrote the items, state clearly that this is an adapted version for your research context.

The methods section should also state the planned validation criteria. You may include Cronbach's Alpha, Corrected Item-Total Correlation, KMO, Bartlett's test, factor loading, total variance explained, or the corresponding SmartPLS indicators. Do not put actual results in Chapter 3 before you have run the data. Validation results belong in Chapter 4.

How to write this in your thesis

You can adapt the following paragraph: “The service quality scale was adapted from [author, year] and comprised [number] indicators across [number] dimensions. After comparing it with the context of [industry or service], the research team adjusted the wording of several indicators and used a [5 or 7]-point Likert scale, where [1] represented [lowest level] and [5 or 7] represented [highest level].”

When you have the actual results, continue: “The validation results showed that the scales had Cronbach's Alpha values ranging from [lowest value] to [highest value], and the retained indicators had Corrected Item-Total Correlation values of at least [value]. The deleted indicators were [indicator codes], for the following reason: [reason based on the criterion and content].” Replace every bracketed part with figures from your own output.

If you use SmartPLS, state clearly that you assessed the measurement model first and then tested the relationships in the structural model. Do not include CFI, TLI, or RMSEA in a SmartPLS report. These are fit indices associated with CB-SEM under Hu and Bentler (1999), not core criteria for PLS-SEM.

Common mistakes

The first mistake is taking a set of indicators from the Internet without knowing its source. When someone asks why an item measures service quality, you have no basis for explaining it. Save the author, year, context, structure, and reason for adaptation as soon as you build the questionnaire.

The second mistake is treating a high Cronbach's Alpha as sufficient. An Alpha that is too high can sometimes indicate that the indicators repeat nearly the same wording. Review the content, number of indicators, and factor structure instead of chasing an attractive number.

The third mistake is deleting indicators repeatedly to force EFA to meet a threshold. Each deletion needs a statistical or content-based reason, recorded for every run. If you delete many indicators, the original structure may no longer represent the construct.

The fourth mistake is combining all dimensions in one Cronbach's Alpha run. If the model has five dimensions, test each theoretical group first. Putting every indicator into one Alpha can produce results that are difficult to interpret.

The final mistake is using the wrong terms. An indicator is an individual statement in a scale, while a questionnaire is the complete data-collection instrument. Keeping these concepts distinct makes Chapter 3 easier to defend.

Frequently asked questions

Is it mandatory to use SERVQUAL for a service quality scale?

No. SERVQUAL by Parasuraman et al. (1988) is a widely used foundational source, but you need to demonstrate that it fits the service type, customers, and research model. If you use another source, state its source, structure, and reason for selection.

How many indicators make a service quality scale acceptable?

Each study has its own appropriate level. The number depends on the construct definition, number of dimensions, and scale source. Hair et al. (2010) mention a rule of 5 to 10 observations per indicator when considering sample size, but this rule does not determine how many indicators the scale must contain.

What should I do if Cronbach's Alpha is low?

First, check reverse-coded items, data-entry errors, and coding. Then review Corrected Item-Total Correlation, the item wording, and Alpha if the item is deleted. Cronbach's Alpha from 0.7 upward is commonly referenced under Nunnally (1978), while a level from 0.6 may be considered for an exploratory scale under Hair et al. (2010).

Can I translate a foreign scale word for word?

You can use a translation as a basis, but adjust the wording to the Vietnamese context and respondent population. In the thesis, state the source, adaptation process, and questionnaire-checking process. A translation that has not been validated should not be called the official version of the scale.

Should I use SPSS or SmartPLS to validate the scale?

SPSS is suitable when you want to run Cronbach's Alpha, EFA, and familiar regression analyses. SmartPLS is suitable when the model includes latent variables and you need to assess outer loading, CR, AVE, HTMT, and path coefficient relationships. Choose according to the model, hypotheses, and lecturer's requirements, rather than choosing a program because it produces more tables.

Open your .sav or .csv file again, make a source list for each group of indicators, check the coding, and record which indicators need adaptation before running a trial analysis. If you need to run the analysis on your own SPSS or SmartPLS data, see DoThesis M4 data analysis.