What Is the AIDA Model and How Can You Apply It in Quantitative Research?

What Is the AIDA Model

The AIDA model describes the process from the moment a customer becomes aware of a marketing message to the point at which they take action. AIDA stands for Attention, Interest, Desire, and Action. In quantitative research, you can use AIDA to explain how advertising content, product information, or media experiences affect purchase intention, registration intention, or service-use intention.

Unlike a single scale, AIDA is a conceptual framework made up of several latent variables. Each component needs to be measured with a group of items in the questionnaire. After collecting the data, you can assess reliability with Cronbach's Alpha, run EFA, and then test the relationships with linear regression or PLS-SEM.

The origins of AIDA are commonly linked to the response process of a message recipient in marketing. When applying it to a current research topic, state the research context clearly and avoid claiming that every customer passes through four stages in one fixed order. AIDA is most suitable when your topic concerns consumer responses to advertising, digital content, brands, or sales information.

You can treat AIDA as a marketing behavior model and combine it with an appropriate underlying theory. For example, research on technology acceptance may refer to the TAM model when it needs to explain perceived usefulness and perceived ease of use. Those concepts should not be added to AIDA simply because they appear related.

Components of the Model

The four AIDA components should be separated into distinct concepts when you design the questionnaire. The table below shows the practical meaning of each component and examples of outcomes it may help explain.

Attention usually appears at the beginning of the process. An advertisement may be displayed several times without creating meaningful attention. Your items should therefore ask whether respondents noticed the content, image, brand, or main message.

Interest goes beyond seeing an advertisement. Respondents may indicate that they want to read more, learn about product features, or follow related information. This stage is suitable for measuring how appealing the content is to viewers.

Desire reflects the shift from learning about something to wanting it. Consumers may judge that the product fits their needs, is worth owning, or provides a benefit they are looking for.

Action is the component closest to a business outcome. You still need to define whether Action in your study means action intention or behavior that has already occurred. If the questionnaire is administered before purchase, “purchase intention” or “intention to act” is usually the more accurate wording.

When building the model, you can treat the four components as independent variables or as sequential mediators. The choice depends on your research question. If you need to distinguish a conceptual model from hypotheses, read What is a hypothesis before writing Chapters 2 and 3.

The Original Model and Common Extensions

The original AIDA model focuses on the response process to a message: attracting attention, creating interest, arousing desire, and encouraging action. In quantitative research, many studies use AIDA as a framework for classifying variables rather than mechanically testing whether every response occurs in a fixed sequence.

One extension is to divide Attention into awareness and recall. The two aspects may be related, but they are not identical. Someone may recognize an advertisement as soon as they see it but fail to remember the brand later.

Another approach is to extend Interest with information quality, interaction, or content fit. In digital media, views, clicks, or interaction time can be treated as supplementary behavioral data. These platform metrics should not automatically replace perception scales in the questionnaire because they measure different aspects.

Desire and Action can also be placed in a model containing a mediator or moderator. For example, brand trust may moderate the relationship between Desire and purchase intention. Each additional variable still needs a theoretical basis and its own scale. You can refer to the way factors are grouped in the SERVQUAL model, but do not mix service-quality components into AIDA when your topic does not study services.

Check three points when choosing an extended model. First, does the new variable answer your research question. Second, can you collect data for that variable from the correct survey respondents. Third, will the number of variables make the questionnaire too long for your available time and sample size.

If your topic focuses on a competitive environment, AIDA can be placed alongside other frameworks such as the SWOT model or Porter's five forces model. SWOT and Porter analyze the strategic context, while AIDA focuses on customer responses to a message. They do different jobs.

Common Scales for Each Concept

AIDA does not provide a ready-made set of Vietnamese statements that can be copied into every questionnaire. You need to find a scale that fits your context, translate and adjust the wording, and then check it with pilot data. Likert is commonly used to record respondents' level of agreement (Likert, 1932).

The table is a framework for developing items, not a validated scale for your study. When you include it in your thesis, replace the source column with the specific material you have read and cited. If the origin is unclear, describe the items as adapted to your research context instead of calling them an original scale.

Each concept should have multiple items so that you can assess internal consistency. Avoid writing four identical questions and assigning four different codes. Attention should focus on recognition or attention, Interest on the need to learn more, Desire on wanting the product, and Action on intention or behavior.

Keep the subject and object consistent throughout the questionnaire. If your topic examines advertising on TikTok, state clearly that respondents are evaluating TikTok advertising. If the respondents are customers who have already purchased, Action can measure behavior that has occurred. If they have not purchased, the items should measure future intention.

After collecting the data, check Cronbach's Alpha and Corrected Item-Total Correlation. Cronbach's Alpha at 0.7 or above is generally considered acceptable (Nunnally, 1978). For a new scale or exploratory research, Alpha from 0.6 may be considered (Hair et al., 2010). Corrected Item-Total Correlation at 0.3 or above is a commonly used threshold (Nunnally and Bernstein, 1994).

Applying the Model to Your Study

Suppose your topic is: “The impact of social media advertising on cosmetics purchase intention among students in Ho Chi Minh City.” You can use AIDA as a sequential model in which Attention affects Interest, Interest affects Desire, and Desire affects Action.

The conceptual model can be presented as follows:

Attention → Interest → Desire → Action

You can write four hypotheses from this model:

• H1: Attention has a positive effect on consumers' Interest in social media cosmetics advertising.
• H2: Interest has a positive effect on consumers' Desire for the advertised cosmetics product.
• H3: Desire has a positive effect on Action, measured by cosmetics purchase intention.
• H4: Attention has an indirect effect on Action through Interest and Desire.

H4 must be tested through the indirect effect. Do not conclude that it is supported simply because H1, H2, and H3 have small p-values. With bootstrapping, the indirect effect is considered significant when the confidence interval does not contain 0 (Preacher and Hayes, 2008).

Define control variables only when you have a clear methodological reason, such as gender, age, income, or social media use frequency. Adding many control variables just to make the model look complete creates additional demands for the questionnaire, sample size, and interpretation of the results.

How to write this in your thesis: “Based on the consumer response process to a marketing message, this study proposes that Attention affects Interest, Interest affects Desire, and Desire affects Action. These relationships are tested using survey data from [research respondents] in the context of [research context].”

Use variable names consistently in the diagram, questionnaire, and data file. If the questionnaire uses “Purchase Intention,” do not change it to “Action” in Chapter 4 without explanation. This inconsistency often leaves readers unsure whether you measured actual behavior or only intention.

Testing the Model with SPSS or SmartPLS

SPSS is suitable when you want to assess reliability, run EFA, create composite variables, and use linear regression. The usual sequence is Analyze > Scale > Reliability Analysis for Cronbach's Alpha, followed by Analyze > Dimension Reduction > Factor for EFA, and then Analyze > Regression > Linear to test direct relationships.

In EFA, KMO at 0.5 or above and a statistically significant Bartlett test below 0.05 are commonly used conditions (Kaiser, 1974). Factor loading at 0.5 or above and total variance explained at 50% or above are common reference thresholds (Hair et al., 2010). Review the Rotated Component Matrix, Communalities, and Total Variance Explained, rather than looking only at KMO.

If the model has several latent variables and you want to assess the measurement model and structural model at the same time, SmartPLS may be more suitable. In SmartPLS 4, import the .csv file, create the constructs Attention, Interest, Desire, and Action, draw the arrows according to the hypotheses, and then run PLS Algorithm and Bootstrapping.

For PLS-SEM, you can report outer loading, Composite Reliability, AVE, HTMT, VIF, path coefficient, t-value, p-value, R², and Q². Outer loading at 0.7 or above is a commonly used reference threshold (Chin, 1998). CR at 0.7 or above and AVE at 0.5 or above are commonly used to assess reliability and convergent validity (Fornell and Larcker, 1981). HTMT should be below 0.85, or below 0.90 for closely related concepts (Henseler et al., 2015). VIF below 5 is a common reporting threshold (Hair et al., 2019).

When running Bootstrapping, set the number of subsamples according to the procedure you choose and state it clearly in the methods section. Current PLS-SEM guidance commonly uses 5,000 bootstrap subsamples (Hair et al., 2022). R² shows how much of the dependent variable is explained by the predictor variables. In PLS-SEM, R² values of 0.75, 0.50, and 0.25 are respectively interpreted as substantial, moderate, and weak according to the reference levels (Hair et al., 2011).

Do not include CFI, TLI, or RMSEA simply because you saw these indices in SEM materials. They are suitable for CB-SEM, while SmartPLS requires the appropriate PLS-SEM indicators. If you use AMOS, explain clearly why you selected CB-SEM and follow the corresponding procedure.

Common Errors When Using This Model

The first error is treating AIDA as four single items. Attention, Interest, Desire, and Action are four concepts, and each concept usually needs a group of items. If each concept has only one question, it will be difficult to assess reliability and measurement validity.

The second error is calling purchase intention actual action. A respondent's statement that they intend to buy does not prove that they have bought the product. In Chapter 3, record the dependent variable accurately as “purchase intention” when your data come only from a questionnaire administered at one point in time.

The third error is creating too many hypotheses without a logical basis. The AIDA model does not automatically prove that every marketing variable affects every component. Each arrow needs to be explained through theoretical reasoning and the research context.

The fourth error is deleting items only to increase Alpha. An item with low Corrected Item-Total Correlation may need review, but you should read the item, check reverse coding, and review the data before deleting it. The first pass of EFA seldom produces a tidy table. Removing items, rerunning the analysis, and recording each round is normal, but you need to explain the reason.

The final error is mixing AIDA with other models without distinguishing their functions. You can refer to Maslow's hierarchy of needs to explain motivation, but Maslow and AIDA are not the same type of model. State which model provides the underlying framework and which model supplies additional variables.

Frequently asked questions

Is the AIDA model a theory?

AIDA is commonly used as a model or framework for describing the marketing response process. In your thesis, state its role in the study accurately and use the word “theory” only when your underlying literature actually presents AIDA in that way.

How many factors does the AIDA model have?

AIDA has four components: Attention, Interest, Desire, and Action. When designing the questionnaire, you can code them as ATT, INT, DES, and ACT, but the codes are only conventions in the data file.

Can the AIDA model be used in quantitative research?

It can be used when the study examines customer responses to advertising, brand content, or product information. You need to convert each component into a latent variable, design the items, collect data, and test the model with SPSS or SmartPLS.

Should you run the AIDA model in SPSS or SmartPLS?

SPSS is suitable when the model is simple, the scales have been combined into variables, and you want to run EFA and regression. SmartPLS is suitable when you have several latent variables, want to assess the measurement model and structural model together, or need to test an indirect effect with bootstrapping.

Where can you find scales for the AIDA model?

Choose an original study or previous study with a context close to your topic, then check the definition and measurement of each concept. Do not copy a set of questions from the internet without knowing its source. The Vietnamese translation also needs to be adapted to the respondents and checked with pilot data.

Open your data file now, check the four groups of codes ATT, INT, DES, and ACT, and compare each item with the hypotheses before running the analysis. If you need to run the model on your own .sav or .csv file, see M4 data analysis with SPSS and SmartPLS.