Technology Acceptance Model TAM: Components, Scales, and Testing

What is the TAM model

The Technology Acceptance Model, or TAM, explains why a person accepts or rejects a technology, system, or platform. You can use TAM when your topic examines the use of mobile banking, e-wallets, online learning, management software, healthcare applications, or a digital platform.

Davis proposed the model in 1989, based on two central perceptions: perceived usefulness and perceived ease of use. These two constructs affect attitude toward using, behavioral intention to use, and actual system use. The model's origins are presented in Davis's study of perceived usefulness, perceived ease of use, and user acceptance of information technology (Davis, 1989).

In a thesis, TAM fits questions such as whether perceived usefulness increases intention to use technology, whether ease of use affects attitude or intention, and whether intention leads to actual use. You can also read the background in the TAM model or what TAM technology acceptance means.

TAM does not automatically make a research topic a sound model. You still need to define which technology you are studying, who uses it, whether the dependent variable is intention or actual behavior, and which scale and questionnaire fit the study.

Components of the model

The original TAM is usually presented as a sequence from perceptions to attitude, intention, and behavior. Depending on your software and data, you can test the full sequence or focus only on the relationships specified in your research model.

You need to write the relationships between constructs as directional hypotheses. For example, PEOU may affect PU because technology that is easy to use can help users recognize its benefits more clearly. PU and PEOU may also affect ATU, while ATU and PU may affect BI. Keep a construct in the model because it serves your research objective and context, rather than adding every construct simply because it appears in the original TAM diagram.

If your study concerns an information system after users have experienced it, you can also consult the DeLone and McLean IS Success model. That model focuses on system quality, information quality, use, and benefits, so its research questions differ from TAM.

The original model and common extensions

Davis's original TAM focuses on PU, PEOU, attitude, intention, and actual use. Some studies remove attitude to make the model more concise and test the direct effects of PU and PEOU on BI. This is a methodological decision, not a mandatory rule. Explain why the structure fits your research context.

One extension adds external variables before PU and PEOU. Interface quality, user experience, trust, or social influence, for example, can be positioned as factors that explain users' perceptions. Each construct needs its own definition and scale. Avoid combining several ideas into a general construct called “technology factors”.

TAM is also often compared with other models. Ajzen's TPB explains behavior through attitude, subjective norm, and perceived behavioral control (Ajzen, 1991). UTAUT by Venkatesh et al. adds performance expectancy, effort expectancy, social influence, and facilitating conditions (Venkatesh et al., 2003). UTAUT2 extends the consumer context with constructs such as hedonic motivation, price value, and habit (Venkatesh et al., 2012).

You can consult innovation diffusion theory if your topic focuses on how a group adopts an innovation over time. When users have direct experience and your questions concern usefulness and ease of use, TAM is usually easier to implement.

Another extension combines TAM with ECM when the study examines continuance intention. The expectation-confirmation model, ECM, fits questions about whether users continue using a system after trying it. When combining the models, avoid overlap between satisfaction, perceived usefulness, and behavioral intention.

Common measurement scales for each construct

The table below is a guide for finding and adapting scales, not a validated questionnaire for every context. Check the original source, adapt the wording to your respondents, pilot the questionnaire, and test the scales again with your study data.

When writing indicators, use one sentence for one idea. For example, “The application helps me save time and reduce costs” contains two benefits. A respondent may agree with one and disagree with the other. You also need to identify the subject clearly. “I find the system easy to use” is more suitable than “The system is easy to use” when the whole questionnaire uses the first person.

A 5-point or 7-point Likert scale can be used, provided that you clearly label the points from strongly disagree to strongly agree and keep the coding direction consistent. Likert is the foundation of this type of scale (Likert, 1932). If you include reverse-worded indicators, mark them in the data file and reverse code them before running Cronbach's Alpha or EFA.

Do not copy a translated scale word for word from another thesis. A banking application, an online learning system, and internal software may use different names for their functions. Replace the technology name in the indicators, while preserving the theoretical meaning of PU, PEOU, ATU, or BI.

Applying TAM to your topic

Suppose your topic is “Factors affecting students' intention to use online learning applications in Ho Chi Minh City”. The model may include PEOU, PU, ATU, and BI. In this example, all four constructs are latent variables, and each is measured by multiple indicators on a Likert scale.

The conceptual model can position PEOU and PU as independent variables, ATU as a mediator, and BI as the dependent variable. A possible hypothesis structure is:

• H1: Perceived ease of use has a positive effect on perceived usefulness of the online learning application.
• H2: Perceived usefulness has a positive effect on attitude toward using the online learning application.
• H3: Perceived ease of use has a positive effect on attitude toward using the online learning application.
• H4: Attitude toward using has a positive effect on intention to use the online learning application.

The hypotheses above are illustrative model content, not conclusions about your data. When writing your proposal, add a theoretical argument for each relationship. During data collection, include screening questions about whether respondents have used the application if the topic requires actual experience.

You can also add social influence or trust, but then answer three questions: what theoretical basis supports the construct, is there a suitable scale, and does adding it change the research question. The more constructs you include, the longer the questionnaire becomes, and the sample-size requirement and interpretation burden also increase.

How to write this in your thesis

You can adapt the following paragraph: “Based on Davis's TAM (Davis, 1989), this study proposes that perceived ease of use and perceived usefulness affect [attitude/intention to use] among [study population] toward [technology]. The model includes [number] constructs and [number] hypotheses, tested using survey data collected from [population] in [location] during [time period].”

In Chapter 3, distinguish clearly between the source of the scale, the translation process, wording adjustments, the initial number of indicators, and the criteria for removing an indicator. In Chapter 4, present the results in this order: sample characteristics, scale reliability and validity, the structural model, hypothesis testing, and discussion.

How to test the model with SPSS or SmartPLS

SPSS is suitable when you need to clean data, produce descriptive statistics, run Cronbach's Alpha, conduct EFA, and estimate linear regression. Run EFA through Analyze > Dimension Reduction > Factor, then check KMO and Bartlett's Test, Total Variance Explained, and Rotated Component Matrix. For Cronbach's Alpha, open Analyze > Scale > Reliability Analysis, move the indicators for each construct into the Items box, select the Statistics button, and choose Scale if item deleted.

SmartPLS is suitable when your model contains latent variables, multiple relationships between constructs, a mediator, or a moderator. In SmartPLS 4, create a project, import the .csv file, drag the constructs and indicators into the model, connect the arrows, and run PLS-SEM Algorithm. Then run Bootstrapping to check the path coefficient, standard deviation, t-value, and confidence interval. Separate measurement model assessment from structural model assessment according to the two-step logic (Anderson and Gerbing, 1988).

For the measurement model, check Outer Loadings, Cronbach's Alpha, rho_A, Composite Reliability, AVE, and Discriminant Validity. Outer loading from 0.7 upward is commonly used in PLS-SEM (Chin, 1998). CR from 0.7 and AVE from 0.5 upward are commonly referenced thresholds (Fornell and Larcker, 1981). HTMT below 0.85, or below 0.90 for closely related constructs, is proposed for assessing discriminant validity (Henseler et al., 2015).

For the structural model, check VIF, Path Coefficients, R Square, f Square, and Q Square. VIF below 5 is a commonly reported threshold in PLS-SEM (Hair et al., 2019). R² values of 0.75, 0.50, and 0.25 can be interpreted as substantial, moderate, and weak, respectively, depending on the PLS-SEM context (Hair et al., 2011). Q² greater than 0 indicates meaningful predictive relevance under the cross-validation logic (Stone, 1974). When running bootstrapping, 5,000 subsamples is an option stated in current PLS-SEM guidance (Hair et al., 2022).

If you use AMOS or another CB-SEM tool, CFI, TLI, RMSEA, and SRMR are among the fit indices to examine. Do not place CFI, TLI, or RMSEA in a SmartPLS results section as though they belonged to the same family of indices. CFI and TLI from 0.90 upward, and RMSEA and SRMR no greater than 0.08, belong to the CB-SEM context (Hu and Bentler, 1999).

Common mistakes when using this model

The first mistake is calling every technology topic TAM without including PU or PEOU. If your model no longer contains the two central perceptions, explain whether it is an extended TAM, a simplified TAM, or a combined model. The model name in Chapter 1, the research diagram, and the methods section must be consistent.

The second mistake is taking a scale from a similar thesis and describing it as the original scale. Trace the construct back to its source, check the original context, and state clearly what you translated or adapted. A scale used in a previous study still needs to be tested again on the current sample.

The third mistake is creating hypotheses after looking at the p-value. That approach adjusts the model to the results rather than to the research question. You may remove an indicator for measurement reasons, but record the criterion, the run in which it was removed, and the effect on scale reliability or validity.

The fourth mistake is concluding that a relationship “has an effect” only because the coefficient is positive. In SPSS or SmartPLS, check the p-value or confidence interval, the direction of the coefficient, effect size, and the model's explanatory power. A positive coefficient without statistical significance is not enough to accept a hypothesis.

Finally, do not confuse intention to use with actual use. A respondent may intend to use an application without having used it, or may have used it without intending to continue. These two constructs need different screening questions and indicators.

Frequently asked questions

What is the Technology Acceptance Model TAM?

TAM explains technology acceptance through perceived usefulness, perceived ease of use, attitude, intention, and actual use. Davis proposed the model in his study of user acceptance of information technology (Davis, 1989).

Must the TAM model include attitude?

There is no single structure that is mandatory for every topic. Some studies retain ATU to explain the process from perceptions to intention, while others test the direct effects of PU and PEOU on BI. State the reason for your choice based on your objective, context, and theoretical literature.

Should you run TAM in SPSS or SmartPLS?

If you need Cronbach's Alpha, EFA, and regression with composite variables, SPSS can meet those needs. If the model contains latent variables, several simultaneous relationships, a mediator, or a moderator, SmartPLS lets you assess the measurement model and structural model within the same workflow.

Where should you obtain TAM scales, and how many indicators are enough?

Start with Davis's scales, then adapt the context and wording to the specific technology (Davis, 1989). The number of indicators depends on the construct and the source scale. Do not choose a number simply to make the questionnaire shorter. After collecting data, check scale reliability and validity.

What should you do if no TAM hypotheses are accepted?

First check coding, missing data, reverse-worded indicators, sample size, scale quality, and the way you ran the model. If the procedure is correct, nonsignificant results are still findings that should be reported honestly. Discuss the user context, technology characteristics, and study limitations instead of changing the data to obtain a p-value.

Open your data file, create a mapping table between PU, PEOU, ATU, BI, and their indicators, and check each hypothesis before choosing SPSS or SmartPLS. If you need to run the analysis on your own .sav or .csv file, see M4 quantitative data analysis.