The 5-Gap Service Quality Model: Structure and Application

What is the 5-gap service quality model

The 5-gap service quality model is a theoretical framework for explaining why customers rate a service below or above their initial expectations. Parasuraman, Zeithaml, and Berry developed the model through their research on service quality, and it was later associated with the SERVQUAL scale in (Parasuraman et al., 1988).

The model traces service quality from customer expectations and management perceptions to service standards, service delivery, and the information communicated externally. The five gaps help you explain the causes, while SERVQUAL is commonly used to measure customers' perceptions of service quality.

You can read a shorter overview in What are the 5 gaps of service quality, or place this model within the Research models category when choosing an approach for your topic.

You need to separate two layers in your research. The 5-gap model explains gaps in service management, while SERVQUAL is a scale commonly used to measure service quality through customer perceptions and expectations. You can therefore use the model to build your argument, but your analysis must still identify the independent variables, dependent variable, and items clearly.

Components of the model

The five gaps are not five independent variables that must all be entered into SPSS. They are five points at which service quality may deteriorate. When you move to your thesis, select the parts that can be measured with a questionnaire and that fit your respondents.

Gap 5 is usually the part closest to customers' evaluation of service quality. If you survey service users, your questionnaire will often focus on actual perceptions, expectations, or service quality dimensions. The first four gaps can explain management causes, but they require suitable data from managers, employees, or process documents.

The five gaps are also not the same as five fixed measurement dimensions. The commonly used SERVQUAL dimensions are Reliability, Responsiveness, Assurance, Empathy, and Tangibles. These are scale dimensions, whereas the five gaps form a theoretical structure. Mixing the two concepts in your theoretical framework can lead the committee to ask which model you are actually testing.

The original model and common extensions

The original model developed by Parasuraman et al. focused on how service quality is formed. SERVQUAL was later used widely to measure the gap between customer expectations and perceptions. You can read the separate article on the SERVQUAL model if your study measures service quality with specific items.

In many theses, students use the five gaps as a theoretical foundation but test only a shortened version. For example, the model may use perceived service quality as the dependent variable, with Reliability, Responsiveness, Assurance, Empathy, and Tangibles as independent variables. This approach fits customer data better because you can collect the variables in one questionnaire and run regression or PLS-SEM.

Another extension measures actual perceptions directly instead of asking respondents to rate both expectations and perceptions. This logic is commonly associated with SERVPERF. Compare it with the article on the SERVPERF model before deciding whether your study should use a gap score or perception scores only.

When you present the original model and its extensions, separate three points. First, the 5-gap model explains the sources of service quality problems. Second, SERVQUAL provides a way to measure quality dimensions. Third, your empirical model is the version adapted to your industry, location, and respondents.

You should not enter all five gaps into a regression model simply because the topic title contains the number 5. If you do not have items measuring Gaps 1 through 4, you cannot draw direct conclusions about them. In that situation, state that the 5-gap model provides the theoretical basis, while your quantitative model focuses on perceived service quality and measurable dimensions.

Common scales for each concept

The scale most often referenced is SERVQUAL by Parasuraman, Zeithaml, and Berry. Their work measures service quality through dimensions such as reliability, responsiveness, assurance, empathy, and tangibles (Parasuraman et al., 1988).

These items are starting points for adaptation, not validated wording for every industry. Replace “provider” with the hospital, bank, university, e-commerce platform, or other context in your study. If you include reverse-coded items, mark them in your codebook immediately so that you do not enter them incorrectly when running the analysis.

For a Likert questionnaire, keep the number of response points and the wording consistent. The Likert scale originated from Likert's technique for measuring attitudes (Likert, 1932). In SPSS, each item should have a short name, such as REL1, REL2, RES1, or EMP1. Put the full wording in Variable Label rather than using long variable names with spaces.

After data collection, the usual sequence is data checking, Cronbach's Alpha, EFA when the model requires exploration of the structure, and then regression or PLS-SEM. Cronbach's Alpha from 0.7 upward is commonly considered acceptable (Nunnally, 1978). Corrected Item-Total Correlation from 0.3 upward is a common reference point (Nunnally and Bernstein, 1994). For a new or exploratory scale, Alpha from 0.6 may be considered under (Hair et al., 2010), but you must explain the context instead of automatically retaining every item.

Applying the model to your study

Suppose your topic is: “Factors affecting the digital banking service quality of customers in Ho Chi Minh City.” You can use the 5-gap model as the explanatory foundation and then build a shortened quantitative model containing the five SERVQUAL dimensions and overall service quality.

The conceptual model may include five independent variables, Reliability, Responsiveness, Assurance, Empathy, and Tangibles. The dependent variable is Service Quality. You can add Customer Satisfaction or Intention to Continue Use if your research questions and measurement sources support the extension clearly.

Four illustrative hypotheses could be written as follows:

• H1: Reliability has a positive effect on digital banking service quality.
• H2: Responsiveness has a positive effect on digital banking service quality.
• H3: Assurance has a positive effect on digital banking service quality.
• H4: Empathy has a positive effect on digital banking service quality.

If your model includes Tangibles, you can add H5. The hypothesis numbering must match the model diagram, hypothesis summary table, and results section. Do not label Reliability as H1 in the diagram and then refer to Responsiveness as H1 in Chapter 4.

You also need to decide whether you are testing the effects of service quality dimensions or measuring the gap between expectations and perceptions. If you use gap scores, describe the formula for creating the variable and how you handled the two response sets. If you use perception scores only, call the construct perceived service quality and avoid claiming that you directly measured all five gaps.

You could adapt the following sentence for your thesis: “Based on the 5-gap service quality model and the SERVQUAL scale of (Parasuraman et al., 1988), this study proposes to test the effects of [service quality dimensions] on [dependent variable] in the context of [industry or location].”

A sound model connects theory, scales, hypotheses, and analysis. If the Assurance construct has four items but your methods section describes three, fix that mismatch before running the model. Create a mapping table with the item code, construct, item wording, source, and related hypothesis.

Testing the model with SPSS or SmartPLS

SPSS is suitable when your model uses composite scores and your main goals are reliability testing, EFA, correlation, and linear regression. Start with Analyze > Scale > Reliability Analysis, enter the items for each construct in a separate run, and select Scale if item deleted through the Statistics button to see the effect of removing an item.

If you need EFA, use Analyze > Dimension Reduction > Factor. Select an appropriate extraction method and rotation method, then inspect KMO and Bartlett's Test, Total Variance Explained, and Rotated Component Matrix. KMO from 0.5 upward and a statistically significant Bartlett test are commonly cited criteria under (Kaiser, 1974). Factor loading from 0.5 upward is commonly referenced under (Hair et al., 2010). An Eigenvalue greater than 1 is an extraction rule associated with (Kaiser, 1960).

For regression, go to Analyze > Regression > Linear. Place Service Quality in Dependent and the SERVQUAL dimensions in Independent. Read Model Summary, ANOVA, and Coefficients. Check VIF, residuals, Durbin-Watson, and the signs of the coefficients. Durbin-Watson between 1 and 3 is a reference criterion under (Field, 2013).

SmartPLS is suitable when your model includes latent variables and multiple items, and you want to assess the measurement model and structural model together. Import your .csv file, create constructs, draw the paths between constructs, and run PLS-SEM Algorithm. Then run Bootstrapping to test the path coefficient. Assessing the measurement model before the structural model follows the guidance in (Hair et al., 2022).

In SmartPLS, the tables to read include Outer Loadings, Construct Reliability and Validity, Discriminant Validity, Path Coefficients, R Square, f Square, and the bootstrapping results. Outer loading from 0.7 upward, CR from 0.7 upward, and AVE from 0.5 upward are reference levels supported respectively by (Chin, 1998) and (Fornell and Larcker, 1981). HTMT below 0.85, or below 0.90 for closely related constructs, is proposed in (Henseler et al., 2015). VIF below 5 is reported under (Hair et al., 2019).

The table below is illustrative output, not the result of a real study.

In the illustrative output, Assurance has a positive path coefficient and a p-value below 0.05, while Empathy has no statistical evidence at this level. When writing your actual results, replace these values with the numbers in your own output and report the bootstrap confidence interval if the software provides one. Do not infer conclusions about your data from the illustrative output.

For PLS-SEM, R² of 0.75, 0.50, and 0.25 are commonly described as substantial, moderate, and weak according to (Hair et al., 2011). The f² levels are small at 0.02, medium at 0.15, and large at 0.35 according to (Cohen, 1988). Q² greater than 0 indicates predictive relevance according to (Stone, 1974). CFI, TLI, and RMSEA belong to CB-SEM, so do not add them to a SmartPLS assessment out of habit.

Common mistakes when using this model

The first mistake is calling the five SERVQUAL dimensions the five gaps. Correct the terminology in your diagram, research objectives, and hypotheses. If your study surveys customers only, state clearly that your data do not cover the internal gaps between management, standards, and employees.

The second mistake is copying items from another study and claiming that the scale automatically fits your context. Banking, hospitals, education, and tourism use service terms differently. Review the wording, pilot the questionnaire, and record which items you adapted.

The third mistake is removing an item only because Alpha increases, without considering its content and theoretical structure. An item should be removed for a statistical reason, poor content fit, or a data-entry problem. Save each data version and record the round in which each item was removed.

The fourth mistake is using an expectation and perception gap score without explaining how it was calculated. Respondents may interpret the two sections differently, making the gap score difficult to explain. If your objective is to measure perceptions only, use the perception model and name it accurately.

The final mistake is including too many variables when the sample size is insufficient. For factor analysis, the rule of 5 to 10 observations per item is commonly cited under (Hair et al., 2010). This rule is only one design reference and does not replace consideration of the number of items, constructs, and data quality.

Frequently asked questions

Is the 5-gap service quality model the same as SERVQUAL?

They are close, but not interchangeable. The 5-gap model explains where service quality may deteriorate, while SERVQUAL is a commonly used scale for measuring service quality dimensions (Parasuraman et al., 1988). In your thesis, present the relationship between them and then identify the empirical model you are actually testing.

What gaps are included in the 5-gap service quality model?

The model includes the gap between customer expectations and management perceptions, management perceptions and standards, standards and actual service, actual service and communication, and finally customer expectations and perceptions. Every study has data for all five gaps.

Should I use SPSS or SmartPLS for this model?

SPSS is suitable when you aggregate the dimensions into scores and then run EFA and regression. SmartPLS is suitable when you want to retain the latent-variable structure and assess outer loading, CR, AVE, HTMT, and path coefficient in one workflow.

Choose according to your model, sample size, scale type, and supervisor's requirements. AMOS is suitable for CB-SEM but requires a different approach to model fit assessment. JASP and R can handle many analyses, but you need to prepare the workflow and output presentation yourself.

Do I need to include all 5 gaps in the questionnaire?

Include the gaps in your questionnaire only if your research questions require you to measure them and you have suitable respondents. In a customer survey, the more common measurement is perceived service quality, or expectations and perceptions for each item.

If you cannot measure Gaps 1 through 4, use the 5-gap model in the theoretical framework and state clearly that your quantitative model tests only the dimensions for which you have data.

How many items are enough for the SERVQUAL scale?

There is no fixed number that applies to every industry. The number depends on the scale source, context, construct content, and test results. Make sure each construct has enough items to represent its content, then assess Cronbach's Alpha, EFA, or the measurement model.

If Cronbach's Alpha is low, should I discard the whole scale?

You cannot conclude that immediately. Check Corrected Item-Total Correlation, item content, reverse-coded items, missing data, and coding procedures. Alpha from 0.7 is commonly referenced under (Nunnally, 1978), while an exploratory scale may be considered from 0.6 under (Hair et al., 2010). If you remove an item, record the reason and rerun the analysis in sequence.

How to write this in your thesis

Open your data file, create a mapping table connecting the five SERVQUAL dimensions, item codes, scale sources, and hypotheses, then choose SPSS or SmartPLS according to your actual model through M4 data analysis.