What Is the TAM Model? Components, Scales, and Application

What is the TAM model

You may have a topic about intention to use a banking app, online learning, an e-wallet, or management software, but still be unsure how to connect the variables into a model. The TAM model, short for Technology Acceptance Model, is a theoretical framework commonly used to explain why users accept or reject a technology.

The original TAM model was proposed by Davis in 1989. Its two central variables are Perceived Usefulness, usually abbreviated as PU, and Perceived Ease of Use, abbreviated as PEOU. In the model's logic, users who perceive a technology as more useful and easier to use tend to develop a more positive attitude, which then increases their intention to use it.

The model originates from (Davis, 1989). When you apply it in a thesis, you need to distinguish the theoretical model from your own research model. TAM provides the structure and logic for relationships between constructs. You still need to choose the context, extended variables, respondents, and measurement scales that fit your study.

If you need a separate explanation of this concept, you can read what is TAM technology acceptance. This article focuses on turning TAM into a model that can be measured and tested with questionnaire data.

Components of the model

The original TAM is usually presented as a sequence from perceptions to behavioral intention. Some diagrams also place actual use at the end of the model. Do not add every variable to your questionnaire simply because it appears in the literature. Identify the variables that your research question genuinely requires.

PU answers the question, “What is useful about this technology for me?” PEOU answers the question, “How easy will this technology be for me to use?” The two questions are related but not identical. An application may be useful but difficult to operate, or easy to use without creating a clear benefit.

ATT is often included when you want to explain the psychological process from perception to intention. However, many recent studies use a shorter version and test the direct effects of PU and PEOU on intention to use. Whether you retain or remove ATT should depend on your research objective, theoretical foundation, and sample-size limits.

When you draw the diagram, each arrow must represent the correct hypothesis. For example, PEOU may affect PU because users often evaluate an easy-to-use system as more useful. PU and ATT may both affect BI. Before designing the questionnaire, identify which variable is independent, which is a mediator, and which is dependent.

The original model and common extensions

The original TAM is compact and easy to test. It does not, however, fully explain contextual factors such as trust, perceived risk, social influence, facilitating conditions, or technology-task fit. Many studies therefore add external variables and test whether they affect PU, PEOU, ATT, or BI.

A common extension is to add trust and perceived risk in studies of digital banking, e-commerce, or online payments. Explain why these variables relate to your research context instead of adding them simply to increase the number of hypotheses.

Another option is UTAUT or UTAUT2 when your topic emphasizes social influence, facilitating conditions, hedonic motivation, price value, or habit. UTAUT was proposed by (Venkatesh et al., 2003), while UTAUT2 extended the model to consumer contexts through (Venkatesh et al., 2012). TAM and UTAUT may both appear in your literature review, but choose one main logic so that the model does not become conceptually overlapping.

You may also compare TAM with TPB or other models when building the theoretical foundation. TPB by (Ajzen, 1991) focuses on attitude, subjective norm, and perceived behavioral control. These variables do not automatically become TAM variables. If you combine the models, explain the theoretical basis for each relationship and identify which model each variable comes from.

A good research model does not need a large number of variables. The committee will usually ask three questions: why did you choose this variable, why does that arrow exist, and where did the scale come from? Each additional variable increases the number of hypotheses, indicators, sample-size requirements, and explanations needed in Chapter 4.

If you want to compare TAM with other models, see the TAM technology acceptance model, as well as the articles on innovation diffusion, Expectation Confirmation Model, or DeLone and McLean IS Success. Each model answers a different question, so do not combine variables simply because the topic titles all involve technology.

Common measurement scales for each construct

TAM scales usually use several indicators for each construct. You can consult the original scale structure and then adjust the wording to fit the Vietnamese context. The statements below are examples of wording, not a validated translation for every topic.

You should not copy the English statements directly into a Vietnamese questionnaire. Replace “this technology” with the specific object, such as “Bank X's digital banking application” or “online learning platform Y.” The replacement must preserve the construct's meaning and must not shift the item from perceived evaluation to service-quality evaluation.

For a Likert questionnaire, keep the number of points and coding direction consistent. The Likert scale was used to measure attitudes in the original study by (Likert, 1932). If you choose a 5-point scale, you can define 1 as strongly disagree and 5 as strongly agree. Place the instructions immediately before the group of items so respondents do not misunderstand them.

After collecting data, the usual process includes checking missing data, coding variables, running Cronbach's Alpha, running EFA if you use SPSS, and then testing the measurement model and structural model if you use SmartPLS. Cronbach's Alpha at 0.7 or above is commonly considered acceptable according to (Nunnally, 1978). For exploratory research, Alpha from 0.6 may be considered according to (Hair et al., 2010), but explain the context instead of automatically retaining every low-reliability scale.

Applying TAM to your topic

Suppose your topic is “Factors affecting students' intention to use e-wallets in Ho Chi Minh City.” You could use PU, PEOU, ATT, and BI as the four main constructs. Your respondents would be students who know about or have encountered e-wallets. If you include actual use in the model, you need clear items measuring usage frequency or behavior.

An illustrative model may include the following relationships:

• PEOU positively affects PU.
• PEOU positively affects ATT.
• PU positively affects ATT.
• ATT positively affects BI.

The four hypotheses could then be written as follows:

H1: Perceived Ease of Use has a positive effect on students' Perceived Usefulness of e-wallets.

H2: Perceived Ease of Use has a positive effect on students' attitude toward using e-wallets.

H3: Perceived Usefulness has a positive effect on students' attitude toward using e-wallets.

H4: Attitude toward Use has a positive effect on intention to use e-wallets.

This is an illustrative model, not the result of an actual study. For your own study, replace the object, location, unit of analysis, and variables with specific content. If the topic concerns online learning, replace “e-wallets” with the learning platform and make the indicators fit learning activities.

You also need to identify which relationships are direct and which are mediated. If you want to test ATT as a mediator between PU and BI, state how you will test the indirect effect in the methods chapter. With bootstrapping, an indirect effect is considered significant when its confidence interval does not contain 0 according to (Preacher and Hayes, 2008). In SmartPLS, you can read indirect effects after running bootstrapping.

A paragraph that you can adapt for your thesis is: “Based on the TAM model of (Davis, 1989), this study proposes that Perceived Ease of Use and Perceived Usefulness affect the attitude of [survey respondents] toward [technology]. Attitude is assumed to have a positive effect on intention to use [technology]. Hypotheses H1 to H4 are developed to test these relationships.”

Draw the model with consistent rectangles or ellipses, and use construct names that match the codes in your data file. If the diagram says “Perceived Usefulness” while the file uses PU, introduce both names the first time the construct appears in the results. This allows the committee to connect the diagram, questionnaire, and output.

How to test the model with SPSS or SmartPLS

SPSS is suitable when you need to check data, produce descriptive statistics, run Cronbach's Alpha, EFA, correlations, and linear regression. You can start with a .sav or .csv file, then check variable names, missing values, and reverse-coded variables before running the analysis. In SPSS, Cronbach's Alpha is located at Analyze > Scale > Reliability Analysis.

If you use EFA, the usual path is Analyze > Dimension Reduction > Factor. Check the KMO and Bartlett's Test, Total Variance Explained, and Rotated Component Matrix tables. KMO from 0.5 and Bartlett's test with statistical significance can be referenced to (Kaiser, 1974). Factor loading from 0.5 and total variance explained from 50% can be referenced to (Hair et al., 2010). A first EFA run that comes out clean is unusual. If an item loads strongly on multiple factors or fails to reach an acceptable loading, review the theoretical meaning, record the reason for removing it, and rerun the analysis in a controlled way.

After EFA, you can use regression to test direct relationships. In SPSS, select Analyze > Regression > Linear, place the dependent variable in Dependent, and place the predictor variables in Independent(s). Read the Model Summary, ANOVA, and Coefficients tables. Report Beta, p-value or Sig., and R Square, and check multicollinearity. For regression, a sample size of n at or above 50 + 8m can be referenced to (Tabachnick and Fidell, 2013).

SmartPLS is suitable when the model contains latent variables and many indicators, and you want to test the measurement model and structural model together. In SmartPLS 4, create a project, import the .csv file, drag the indicators to constructs, connect the constructs with arrows, and run PLS-SEM Algorithm. Then run Bootstrapping, with the number of bootstrap subsamples set to 5,000 according to (Hair et al., 2022).

For the measurement model, read Outer Loadings, Reliability, Convergent Validity, and Discriminant Validity. CR from 0.7 and AVE from 0.5 can be referenced to (Fornell and Larcker, 1981). HTMT below 0.85, or below 0.90 for conceptually close constructs, was proposed by (Henseler et al., 2015). For the structural model, read Path Coefficients, R Square, f Square, Construct Crossvalidated Redundancy or Q², and Inner VIF Values. VIF below 5 is reported according to (Hair et al., 2019), while R² can be interpreted using the 0.75, 0.50, and 0.25 bands from (Hair et al., 2011).

SmartPLS does not use CFI, TLI, and RMSEA for evaluation in the same way as CB-SEM. These indices belong to CB-SEM, and the thresholds are stated in (Hu and Bentler, 1999). If you use SmartPLS but copy AMOS's Model Fit table into Chapter 4, the committee may immediately ask why the method and fit indices do not match.

When reporting results, separate the two stages: evaluate the measurement model first, then evaluate the structural model. You can present one table for outer loading, Alpha, rho_A, CR, AVE, and HTMT, followed by another table for path coefficients, t-value, p-value, R², and Q². Do not write that the “model is acceptable” when only one index meets a threshold and the remaining indices have not been checked.

Common mistakes when using this model

The first mistake is treating TAM as a fixed list of variables. TAM is a foundation, while the research model must fit the specific respondents and technology. A study of intention to use a healthcare application may need to consider trust or privacy, whereas a study of internal enterprise software may need to focus on task fit.

The second mistake is proposing a hypothesis without a corresponding scale. If H3 concerns attitude, the questionnaire must include a group of indicators measuring ATT. If the model includes a mediator, plan the indirect-effect test from the methods chapter.

The third mistake is translating PU and PEOU into statements that are too general. “The application is good” may combine usefulness, ease of use, and service quality. Each indicator should reflect one main idea and identify its subject and object clearly.

The fourth mistake is removing an item solely to increase Alpha. SPSS may display the Cronbach's Alpha if Item Deleted column, but that number is only a suggestion. Check the corrected item-total correlation, theoretical meaning, coding, and whether respondents can understand the item.

The fifth mistake is claiming an effect after looking only at a correlation. Correlation indicates the degree of association between variables, while regression or a path coefficient is used to test relationships in the model according to your analysis design. The final conclusion should be based on the coefficient, p-value or confidence interval, and direction of the effect.

The final mistake is taking a model from another study and changing only the technology name. Check the context, respondents, scale wording, and time of use. A scale suitable for enterprise employees may not fit students using an e-wallet.

Frequently asked questions

How many variables does the TAM model have?

The original TAM usually centers on PU, PEOU, attitude, intention to use, and possibly actual use. Your research model may contain fewer or more variables depending on the objective, but every additional variable needs a theoretical basis and a suitable scale.

Should I use SPSS or SmartPLS for the TAM model?

You can use SPSS if your objective is data cleaning, Cronbach's Alpha, EFA, correlation, and regression. SmartPLS is suitable when you want to test latent variables, the measurement model, the structural model, mediation effects, or several relationships at the same time.

Is attitude required in a TAM model?

It is not required for every topic. You can use a shorter model with PU, PEOU, and BI if you have a basis for removing ATT. Explain this choice in the literature review and state clearly how your research model differs from the original model.

Where do TAM scales come from?

Start with the original source (Davis, 1989), then check studies conducted in the same technology context and with similar respondents. When translating or adapting the items, state clearly that the scale is a reference scale adapted for [research context]. Do not present it as the original translation if you do not have evidence to confirm that.

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

First check the coding direction, missing data, reverse-coded variables, and corrected item-total correlation. Then review the indicator wording and the Cronbach's Alpha if Item Deleted column. Remove an item only when you have both statistical and theoretical reasons, and record each processing round so you can explain it in the thesis.

Can TAM be used for continued-use intention?

It can be used, but you need to determine whether the research question concerns initial acceptance or continued use. For continued use, you may need a different theoretical logic or additional variables that fit post-use experience. Do not rename BI as “continued-use intention” if the entire scale still measures first-time use intention.

Open your data file now and list the constructs, indicator codes, and scale sources before drawing the TAM diagram. When you need to check the model, scales, and run the analysis on your own .sav or .csv file, you can use DoThesis M4 data analysis.