SERVPERF Model: Components, Scales, and Application

What is the SERVPERF model

The SERVPERF model measures service quality through customers' actual perceptions after they have used a service. When you use this model, the questionnaire focuses on how customers evaluate the company's actual service performance, rather than measuring expectations before use and perceptions after use as separate components.

The practical issue is how you turn the concept of service quality into observable variables that can be measured with a Likert scale. For example, customers may rate service speed, information accuracy, transaction safety, or employees' attentiveness on a scale from 1 to 5. You then test the data with Cronbach's Alpha, EFA, regression, or PLS-SEM, depending on your research model.

SERVPERF is often chosen when you want a shorter questionnaire and want to measure customers' direct evaluations. By contrast, the SERVQUAL model separates service quality into the gap between expectations and perceptions. Both approaches can address the same research problem, but they produce different questionnaire structures and different ways of interpreting the data.

Components of the model

SERVPERF commonly uses five service quality dimensions inherited from the SERVQUAL structure. Check your specific context before keeping all five dimensions unchanged, because customers of a bank, hospital, university, or e-commerce platform may understand the same item differently.

These five components are a starting point for designing the model. Do not treat them as five variables that must remain unchanged after data collection. If an item has a low corrected item-total correlation, a low loading, or clear cross-loading, keeping it simply to preserve the original number of items can weaken the results.

At the design stage, write an operational definition for each component. Reliability should be described through observable behavior, such as fulfilling commitments or processing transactions accurately. A definition that is too broad can cause the items to overlap with responsiveness or assurance.

You can also review SERVQUAL to compare how these five components are used in service quality research. If your topic needs to explain the causes of service gaps, the 5-gap service quality model addresses a different research question from SERVPERF.

The original model and common extensions

The core structure of SERVPERF measures perceptions of service performance. Therefore, you do not need a separate group of items measuring expectations if your study follows SERVPERF logic only. The questionnaire can begin with screening questions, continue with items for the five components, and then measure an outcome such as satisfaction or continuance intention.

A common extension treats service quality as an independent variable and adds satisfaction as a mediator. In that model, the five components can affect satisfaction, while satisfaction affects loyalty. Another option is to combine the five components into a higher-order service quality construct, but this requires clearer measurement-model decisions and stronger theoretical reasoning.

You need to distinguish a service quality measurement model from a model measuring service outcomes. SERVPERF answers how customers evaluate service delivery. Variables such as satisfaction, recommending the service to others, repurchase intention, or loyalty are outcomes that require separate scales.

If your research concerns overall satisfaction, you can review how a satisfaction index is developed in What is ACSI. Do not bring the entire structure of another index into SERVPERF simply because both concern customers. Each construct needs its own definition, measurement source, and position in the model.

Another extension adjusts the items to the industry. In banking, “the transaction is processed accurately” may be appropriate. In education, “support staff answer procedural questions on time” may be more suitable. In delivery services, “the order arrives on time” may matter more than “the service area is clean.” Adjusting the wording should preserve the construct's meaning rather than turning one item into a question about another construct.

Common scales for each construct

The five SERVPERF components are commonly referenced from the SERVQUAL structure. The foundational source for the service quality dimensions is the scale developed by Parasuraman et al., cited as (Parasuraman et al., 1988). This source is suitable for explaining the five components and providing an initial basis for your item pool.

This table is a reference framework, not a validated translation for every industry. Replace general wording with situations that respondents have actually experienced. For example, the item “employees are willing to help customers” should be rewritten so respondents know whether they are evaluating counter staff, call-center staff, or delivery staff.

For a Likert scale, you can use response levels from “strongly disagree” to “strongly agree,” following the tradition of attitude measurement established by Likert (Likert, 1932). Keep the direction consistent across all items. If you include reverse-coded items, mark them clearly in the data file and recode them before running Cronbach's Alpha.

After collecting the data, check Cronbach's Alpha and corrected item-total correlation before running EFA. Cronbach's Alpha from 0.7 upward is commonly considered acceptable (Nunnally, 1978), while a corrected item-total correlation from 0.3 upward is a commonly used reference point (Nunnally and Bernstein, 1994). For a new or exploratory scale, an Alpha from 0.6 may be considered according to (Hair et al., 2010), but you need to explain the context rather than automatically retaining every item.

Applying the model to your study

Suppose your topic is “Factors affecting customer satisfaction with delivery services in Ho Chi Minh City.” You could build a model with five independent variables: tangibles, reliability, responsiveness, assurance, and empathy. The dependent variable is customer satisfaction.

The conceptual model can be written as:

Tangibles, Reliability, Responsiveness, Assurance, Empathy → Customer satisfaction

Your hypotheses should state the expected direction of the effects clearly:

• H1: Tangibles have a positive effect on customer satisfaction.
• H2: Reliability has a positive effect on customer satisfaction.
• H3: Responsiveness has a positive effect on customer satisfaction.
• H4: Assurance has a positive effect on customer satisfaction.
• H5: Empathy has a positive effect on customer satisfaction.

You can combine the model into five hypotheses if your proposal requires it, but do not stop at H1 through H4 simply because of the numbering. Each hypothesis should connect one independent variable to one specific dependent variable, have a rationale from the literature, and match the questionnaire.

How to write this in your thesis

You can use the following paragraph and replace the bracketed sections: “Based on (Parasuraman et al., 1988), this study proposes that service quality comprises [number] components, including [names of the components]. The study tests the effects of these components on [name of the dependent variable] in the context of [research subjects and study area].”

In the methods chapter, state the survey population, the condition that respondents have used the service, the scale, the coding procedure, and the criteria for removing invalid responses. For sample size, you can refer to 5 to 10 observations per item (Hair et al., 2010). This is a reference basis, not a substitute for an argument about your model and data quality.

Testing the model with SPSS or SmartPLS

If your SERVPERF model uses reflective items and your goal is to test reliability, EFA, and regression, SPSS is easy to follow. Start with Analyze > Scale > Reliability Analysis to run Cronbach's Alpha, then use Analyze > Dimension Reduction > Factor to run EFA. In the Reliability Analysis dialog, select Statistics and check Scale if item deleted to see how Alpha changes when each item is removed.

For EFA, check the KMO and Bartlett's Test, Total Variance Explained, and Rotated Component Matrix tables. A KMO from 0.5 upward and a statistically significant Bartlett's test below 0.05 are reference points from (Kaiser, 1974). The eigenvalue-above-1 rule comes from (Kaiser, 1960), while total variance explained from 50% upward and factor loading from 0.5 upward are commonly referenced according to (Hair et al., 2010).

After EFA, run regression through Analyze > Regression > Linear. Put satisfaction in Dependent and the five components in Independent(s), then review the Model Summary, ANOVA, and Coefficients tables. Check VIF, the sign of B or Beta, t, and Sig.. If the model has several explanatory variables, regression sample size can be considered using the rule 50 + 8m, where m is the number of independent variables, according to (Tabachnick and Fidell, 2013).

SmartPLS is suitable when you want to test the measurement model and structural model at the same time, or when your model includes a higher-order latent construct, mediator, or moderator. In SmartPLS 4, import the .csv data, create constructs, drag indicators onto the constructs, connect the hypothesized paths, and then run PLS-SEM Algorithm and Bootstrapping.

For the measurement model, read Outer Loadings, Construct Reliability and Validity, and Discriminant Validity. Outer loading from 0.7 upward can be referenced from (Chin, 1998) or (Hair et al., 2022), CR from 0.7 and AVE from 0.5 upward from (Fornell and Larcker, 1981), and HTMT below 0.85, or below 0.90 for closely related constructs, from (Henseler et al., 2015). For the structural model, review Path Coefficients, R-Square, f-Square, Collinearity Statistics, and Predictive Relevance.

For PLS-SEM, R² values of 0.75, 0.50, and 0.25 can be interpreted as substantial, moderate, and weak, respectively, according to (Hair et al., 2011). VIF below 5 is referenced by (Hair et al., 2019). When running Bootstrapping, use 5,000 bootstrap subsamples according to (Hair et al., 2022), then report the path coefficient, t-value, p-value, and confidence interval.

The two software packages do not follow the same procedure. SPSS is usually easier to explain when you use EFA and linear regression. SmartPLS is convenient for PLS-SEM. AMOS is suitable for CB-SEM, and fit indices such as CFI, TLI, and RMSEA must be kept separate from PLS-SEM logic. JASP and R can reduce software costs, but you must control the syntax, analysis packages, and table exports yourself. A service that runs SPSS may save operating time, but the greatest risk is being unable to explain why an item was removed when the committee asks.

Common mistakes when using this model

The first mistake is calling every service quality scale SERVPERF while still asking about both expectations and perceptions and then calculating the difference. That mixes two measurement logics. Decide before designing the questionnaire whether you are measuring actual perceptions or calculating the gap between expectations and perceptions.

The second mistake is copying items word for word from another industry. An item suitable for a hotel may not suit online banking. Keep the construct's meaning, but adjust the context, the service provider, and the behavior respondents can actually evaluate.

The third mistake is forcing the data to retain all five components. EFA may produce a structure different from the initial theoretical model. Record each item-removal round, the reason for removal, and the rerun results. Removing an item with statistical and theoretical justification can be acceptable. Removing an item only to obtain a more attractive p-value is difficult to defend.

The fourth mistake is using “service quality” and “satisfaction” as two names for the same construct. Service quality describes an evaluation of service delivery, while satisfaction is an overall evaluation after the experience. If the two scales contain almost identical items, discriminant validity will become a problem.

The final mistake is reporting only Alpha and p-value while ignoring the measurement structure. A sound SERVPERF model needs a measurement source, variable definitions, coding procedures, reliability results, EFA or a measurement model, and then hypothesis tests. The reader needs to see the path from the questionnaire to the conclusion, not just one coefficient table.

Frequently asked questions

How many components does the SERVPERF model have?

The SERVPERF framework is commonly implemented with five components: tangibles, reliability, responsiveness, assurance, and empathy. This structure is referenced from the service quality scale developed by Parasuraman et al. (Parasuraman et al., 1988). When applying it to a specific study, you need to retest the number of components and the grouping of items with actual data.

What is the difference between SERVPERF and SERVQUAL?

SERVPERF focuses on customers' perceptions of actual service performance. SERVQUAL measures both expectations and perceptions and commonly analyzes the gap between the two. As a result, SERVPERF usually produces a shorter questionnaire, while SERVQUAL requires two sets of responses for the same content.

Does the SERVPERF model need to measure expectations?

If you state that you are using SERVPERF based on actual perceptions, the questionnaire does not need a separate expectations section. You can still ask about overall satisfaction or continuance intention as outcome variables, but those variables must remain separate from the five service quality components.

Should you run the SERVPERF model in SPSS or SmartPLS?

SPSS is suitable when you want to run Cronbach's Alpha, EFA, and linear regression through a familiar procedure. SmartPLS is suitable when you are testing PLS-SEM, using a higher-order construct, or analyzing a model with many paths. Choose according to your research question and the reporting format you are required to use, not simply because one software package produces more tables.

How many SERVPERF items are sufficient?

There is no fixed number that applies to every industry. Base your decision on the construct definition, the measurement source, and content coverage. After data collection, check corrected item-total correlation from 0.3 according to (Nunnally and Bernstein, 1994), Alpha according to (Nunnally, 1978), and factor loading according to (Hair et al., 2010), then explain your decision to retain or remove each item.

What should you do if SERVPERF EFA fails?

First check missing data, reverse-coded items, incorrect coding, and variables with only one value. Then review KMO, Bartlett, communalities, loading, and cross-loading. You can remove items one at a time with a documented rationale, rerun the analysis, and record each round. If the structure remains unstable, review the wording, survey context, and construct definitions instead of trying multiple options until the output looks better.

Open your proposal and data file, finalize the five components, list the items, and state the hypotheses before running the tests. If you need to connect the model to the questionnaire and a real data-analysis workflow, see M3 and M4 of DoThesis.