3. Materials and Methods
The designed experiment aimed to record and analyze the online buying experience of consumers. Moreover, it investigated how consumer habits and experiences influence future purchase decisions. For this purpose, an online store with electronic products classified into subcategories (six) was created in WordPress [
35], a popular CMS combined with the Woo Commerce plugin [
36], and participants were assigned a specific scenario (task) to execute. Inside the e-shop pages, a set of advertisements was specially constructed and placed to examine how participants reacted to these marketing stimuli. Advertisements were interactive and had various forms ranging from fixed banners of various sizes, to animated banners, gif banners, and pop-up advertisements (
Figure 1,
Figure 2 and
Figure 3).
In each of these sub-categories, the constructed advertisements were placed (banners and pop ups of different types and sizes or relevance) to serve the research goals. Starting from the home page of the store, there was an advertisement in the form of an animated banner (160 × 600) on the right side of the page featuring irrelevant content. In the “Cell Phones & Tablets” category, we placed two advertisements with dimensions of 336 × 280. The first one included a product of the e-shop with a direct link to it, bypassing the typical navigation path. The second one was an advertisement of irrelevant content, and upon clicking, it led to a new website. An advertisement was considered irrelevant when the depicted product was not offered in the e-shop or it was not relevant to the rest of the e-shop products. The first pop-up with relevant content was placed in the “Books” subcategory as it included a 20% discount coupon. In addition, this category displayed an irrelevant advertisement with a dimension of 160 × 600. In the subcategory “Sports & Fitness”, the second pop-up of the store with irrelevant content appeared after three seconds. The time delay was added to simulate realistic online store conditions but also to observe the real reactions of our users in case they were disoriented and distracted. The second and last 50% discount coupon was placed in the online shopping cart just below the order submission button. In the “Gaming” category, two 336 × 280 advertisements are placed with relevant content (products included in the e-shop) and the same applied to the “Wearable, Drones & Hi-tech” category, with one relevant advertisement of 336 × 280 dimension added on the right side. Finally, the “TV & Multimedia” contained an irrelevant advertisement with a size of 160 × 600.
Throughout the scenario execution, participant data were recorded using the eye-tracking equipment including gaze sequences, fixations in areas of interest, and points/areas with particularly high visual traffic. All the above data were subsequently analyzed using statistical tools and descriptive statistics to draw conclusions.
In addition to the eye-tracker data, the study aimed at examining consumers’ visual behavior and relating it to their personality traits. After studying and analyzing the existing personality tests, we chose HEXACO-60, the six-factor personality test [
28], for the analysis of the personality of each user. Then, with the statistical analysis, we studied how the dominant personality traits of each user affected their ability to recognize specific advertisements, and the personal reaction participants had for each advertisement individually. In addition, we examined the role of advertisement content relevance (i.e., whether and to what extent the relevance of an advertisement affects its recognition).
3.2. Experiment Protocol, Metrics, and Instruments
As far as the research process is concerned, upon entering the laboratory space, participants were informed of the safety measures they had to follow and were sited in front of the eye-tracker for it to be calibrated. In terms of questionnaires, three different questionnaires were used: a personality questionnaire, a pre-test, and a post-test questionnaire. To save time, the personality questionnaire was answered earlier by each participant as it included 60 questions [
28]. Participants then were directed to the pre-test questionnaire—an online form—the purpose of which was to obtain demographic information with questions about age, gender, etc., as well as the general perception of the participant regarding e-marketing, shopping habits (products, type of sales, etc.), and digital advertising. Data collected with the pre-test questionnaire allowed us to statistically correlate the questions with the recorded behavior of the sample during task execution, and this was completed with the analysis of the personality test. The questionnaire was constructed (phrased and structured) so as not to cause suspicion about what would follow as we did not want to influence the results of the experimental process and bias users.
After submitting the pre-test questionnaire, users went through the calibration process and were then directed to the home page of the online shop and were given the scenario (task) to be executed; they could explore on their own until they had placed the required number of products in their shopping cart. The script was given on a separate printed form and the instructions were as follows:
Welcome to our online store.
Your mission is to an order that will contain at least two items from each product category of the store place in your cart.
In case you find online discount coupons, use them before placing your order.
The process is completed upon order submission.
At this point, eye-tracking stopped, and users were asked to fill-in the post-test questionnaire. The purpose of the post-test questionnaire was to examine the attitude of consumers towards advertisements, the interference of privacy caused by advertisements, and the degree of annoyance in the first part. In the second part of the questionnaire, we uploaded each advertisement used in the store and invited users to recognize the advertisements according to the characteristics that interested them: i.e., location, content, color, type of advertisements, or if they did not recognize it at all (could not recall having seen the advertisement during the test). The questionnaire closed with a question about the discount coupons and the overall feeling regarding the advertisements of the specific store. The purpose was to scientifically confirm the reason why users recognized advertisements and examine whether there was a correlation with their personality.
Table 1 depicts the internal consistency of the HEXACO-60 scales. As seen in the table, the internal consistencies’ reliabilities ranged from 0.615 to 0.862 in the sample. The highest reliability of the HEXACO-60 scales is in the dimension of honesty–humility, and the lowest reliability is in the dimension of agreeableness. The results for Cronbach’s alpha reliability levels are consistent with other existing research on personality traits [
27,
28].
Table 2 shows correlations among the HEXACO-60 scales for the sample of 31 participants. As seen in the table, there is a statistically significant correlation between honesty–humility and agreeableness (r (29) = 0.410), between honesty–humility and openness to experience (r (29) = 0.417), and between agreeableness and conscientiousness (r (29) = 0.367). Furthermore, the highest correlation was r (29) = 0.699 between agreeableness and openness to experience. Finally, there is a negative correlation between emotionality and extraversion (r (29) = −0.424). As expected, we find low and negative correlated scores of honesty–humility with the other six factors as its variance is not presented well. Despite that, in honesty–humility and agreeableness, we expected a weak correlation between them as agreeableness is constructed differently from the standardized five factor models, as it lacks the aspects of anger and hostility, which explains the lowest Cronbach’s alpha level (α = 0.615) for agreeableness and suggests that the participants lacked indicative responses [
26,
27].
Table 3 presents the results of the logistic regression analysis performed for every recognition factor (location, product, content, color, banner type, do not remember at all) as a dependent variable to examine whether there are strong correlations between recognition factors and personality traits. Based on the results, only Product6 (Relevant3) and conscientiousness are statistically dependent (
p-value = 0.040). For the other factors, the results indicate that there is no statistically significant relationship between HEXACO’s personality traits and recognition factors for products. Further analysis also shows that only Content1 (Irrelevant_TV) and honesty–humility are statistically dependent (
p-value = 0.040), while the results for other factors indicate that there is no statistically significant relation between HEXACO’s personality traits and recognition factors for content. Regarding color, the results show that Color2 (Irrelevant2) and honesty–humility are statistically dependent (
p-value = 0.035) and Color3 (Animated-Irrelevant) and extraversion are statistically dependent (
p-value = 0.034). For the other factors, the results indicate that there is no statistically significant relationship between HEXACO’s personality traits and color. Regarding the banner type, Banner3 (Animated) and honesty–humility are statistically dependent (
p-value = 0.018) and Banner4 (Irrelevant1) and openness to experience are statistically dependent (
p-value = 0.045). Finally, we included the option of whether participants could not remember a particular advertisement at all. We found that Not at All2 (Irrelevant2) and honesty–humility are statistically dependent (
p-value = 0.007) and Not at All3 (Animated-Irrelevant) and honesty–humility are statistically dependent (
p-value = 0.038); finally, Not at All12 (Promo Cart) and conscientiousness are statistically dependent (
p-value = 0.050). For the other factors, the results indicate that there is no statistically significant relation between HEXACO’s personality traits and the recognition factor of not being able to remember the advertisement.
Eye-Tracking Results
To allow further analysis from the eye tracking data, areas of interest (AOIs) were created so each advertisement could be analyzed individually based on certain metrics. In our research, we focused on the metrics of time to first fixation (in seconds) and visit duration with the representative heatmaps.
Table 4 and
Table 5 present the data collected from the eye-tracker for both relevant and irrelevant advertisements in terms of time to first fixation on each advertisement. The tracker data show that only 8 of the 31 users fixated on the animated banner advertisement with a mean time of 15.58 s and a maximum time of 59.04 s. Nevertheless, these eight users focused their attention to a great extent, as it seems from the mean, which is the highest mean fixation for all the irrelevant advertisements. As for the relevant advertisements, the first advertisement was placed in the “Wearables, Drones & Hi-Tech” category. The tracker data show that 23 of the 31 looked at the advertisement with a mean time of 8.04 s and a maximum time of 37.04 s (M = 8.04, SD = 9.38). For the relevant2 advertisement, the tracker data show that 24 of the 31 looked at the advertisement with mean time of 7.64 s and a maximum time of 29.65 s (Mean = 7.63, SD = 7.88). Similarly, for the relevant3 advertisement, 24 of the 31 participants fixated on the advertisement with a mean time of 9.94 s (Mean = 9.94, SD = 9.64). It is worth mentioning that advertisements with relevant content played a significant role in leading to the users’ first fixation, as they attracted the users’ attention longer and an increased number of participants fixated on it.
Figure 4,
Figure 5 and
Figure 6 depict the heatmaps from the eye-tracking data collected during the experiment. A heatmap is a graphical way to visualize user visual behavior (number and duration of fixations). The warm colors indicate areas that attracted the highest visual attention, and the cool colors point to areas with the lowest attention number of fixations.
Table 6 and
Table 7 present the eye-tracking results for the visit duration of relevant and irrelevant advertisements, respectively. According to these data, on the homepage, consumers focused less on the image at the beginning of the page and more on the different products. More precisely, the visualization shows that users fixated more on the products on the first and third row. This fact can be supported by the visit duration table, as participants—after fixating on the advertisement for the first time—usually did not go back to look at the advertisement again with N = 40 visits and mean = 0.51 s. In addition, the horizontal banner attracted their attention as well.
In the other categories, products seemed to gain more attention than the banners. Participants tended to look at the areas with product information (description, image, price, etc.) and pay less attention to banners. This is also evident from the heatmaps in
Figure 4 and
Figure 5, where the “reddish” areas depicting a high number of fixations are on product-related information and not on the advertisements. In particular, in the “Books” category, despite the higher number of revisits of N = 77, the mean time was 0.51 s. The same pattern was observed in the rest of the categories and could provide valuable insights for web page designers and marketing experts regarding advertisement placement. In the “Gaming” category, consumers paid more attention to banners compared to the banners placed in other categories. Another interesting fact is that despite the high count of revisits (N = 98 and mean = 0.57 s) on the promotional banner, participants—as mentioned earlier—could not recognize or recall the promo code banner. On the pop-up advertisement in the “Sports & Fitness” category, the close button attracted the most fixations (which is rather expected), while on the “Promo Pop-Up”, it was the promo code itself as well as the close button that shared the most fixations.
4. Discussion
At first, we attempted to investigate whether there is a relationship between personality traits and location, and the results indicate there is no statistical evidence to support this relation. Furthermore, we examined the relationship between personality traits and the product recognition factor. Only Product6–Relevant3 and conscientiousness with a p-value of 0.040 was found to be statistically dependent and influence recognizability, while for the other factors, there was no statistical significance in our sample. Conscientiousness is strongly correlated with a user’s pursuit of perfection and organizational skills. In this case, as it is correlated with a specific product relevant to the e-shop, it indicates the users’ self-awareness of the task at hand, and the effect it had on their purchase decision making.
Only Content1–Irrelevant_TV and honesty–humility with a p-value of 0.040 were found to be statistically dependent. For the rest of the advertisements, there was no statistical significance in our sample. In this advertisement, the content pictured Black Friday sales, which can relate to the consumer’s need to purchase products and services at the best price possible, as high honesty–humility scores indicate users with low tendencies of greediness.
We also examined the relationship between personality traits and the recognition factor of color. Color2–Irrelevant2 and honesty–humility with a
p-value of 0.035 and Color3–Animated (Irrelevant) and extraversion with a
p-value of 0.034 were found to be statistically dependent. As expected, with high scores of extraversion, users are easily distracted by visual stimuli due to their low attentional control and disengagement [
37,
38]. In the Irrelevant2 advertisement, the main color was blue, which partially confirms the previous findings about the color blue and its characteristics of competence, sociability, and liveliness, which are characteristics of high scores in HEXACO’s extraversion scale, as it was the only statistically significant advertisement with this color.
Banner3–Animated and honesty–humility with a p-value of 0.018 and Banner4–Irrelevant1 (Avon) and openness to experience with a p-value of 0.045 were found to be statistically dependent. For the remainder of the advertisements, there was no statistical significance in our sample. Based on previous eye-tracking research, animated banners are ignored by users as a type (i.e., animation); however, we found enough evidence to correlate it with a personality trait to confirm or deny the impact of a trait to visual stimuli and animation.
Finally, we examined the relationship between personality traits and participants’ ability to recognize or remember the advertisements at all. Not at All2–Irrelevant2 and honesty–humility with a p-value of 0.007, Not at All3–Animated–Irrelevant and honesty–humility with p-value = 0.038, and finally Not at All12–Promo Cart and conscientiousness with a p-value of 0.050 were found to be statistically dependent. For the remainder of the advertisements, there was no statistical significance in our sample. Yet again, honesty–humility plays a significant role in consumers’ recall ability and relevance. In this case, users that could not recognize irrelevant advertisements were strongly correlated with honesty–humility traits. Thus, users with high scores in extraversion, emotionality, and conscientiousness did not recognize relevant advertisements.
6. Conclusions
Visual attention, relevance, and recognition factors are intertwined when attempting to predict and model human behavior and—in our case—the effect of online advertisements on purchasing behavior. In our study, we implemented the framework of the HEXACO personality traits model to examine whether there is evidence of specific traits that affect recognition and visual behavioral patterns and how eye-tracking analysis can support our findings. To conclude, personality traits and especially honesty–humility can act as a predictive force for some important aspects in consumers’ banner advertisement recognition ability. Despite that, we must consider the internal validity of our sample and how it may have affected our results.
Previous researchers have attempted to investigate the issues of effectiveness, recall, and visual attention of banner advertisement while considering the impact of personality traits [
15,
16,
37,
38]. Although there are differences in the personality models, as discussed in previous chapters, it is of crucial importance to examine how other studies have approached this issue. Kobayashi et al., in [
39], implement the Big Five model to predict users’ receptiveness and how BFM is influenced by different creative designs in web advertisements and cognitive bias. The results indicate that high openness to experience and extraversion have a strong correlation with click-through rates, while there were no significant results to suggest that cognitive bias was affected by BFM. Souiden et al., in [
40], examine the effect of extroversion and introversion in consumer attitudes to general and online advertising. Based on their findings, consumers having higher scores in extraversion shows a strong correlation with attitudes towards advertising in general and online, while the same pattern is not indicated for introverts. In their efforts to signify the importance of personality traits in online advertisement, Açan et al. [
41] studied the intercorrelations between Big Five personality traits, perception towards online advertisement, and purchase intentions using established model scales. With a study sample group of 278 participants, the results indicate a low correlation between purchase intentions and personality traits, while purchase intentions are strongly correlated with consumer perceptions.
Despite HEXACO’s increasing popularity in recent academic research [
42], there are limited studies that investigate its impact in the marketing or advertising sector. Our research signifies the importance of the HEXACO model in banner advertisement and its unique structure, as our preliminary analysis and results address the predictive factor of the honesty–humility personality trait in banner recognition and advertisement recall.
In the future, we intend to enrich our sample with a variety of users (age, background, etc.), as the present research was mainly addressed to graduate students. An interesting point would be to study how expert and novel users perceive advertisement messages and how their expertise or familiarity affects their ability to recognize specific details of an advertisement based on their personality traits. Moreover, the present research focuses on the content of the advertisements (relevant and irrelevant) in correlation with the personality traits. The next step is to focus our attention on specific types of advertisement appeals to examine how the advertised content is perceived and influences consumers based on their personality traits.
Regarding future work, neuromarketing is a promising and rising scientific field that has inspired the current article. To further extend our research, our aim is to include emotional and facial expression recognition analysis. Understanding how each individual human reacts to emotional feedback and how our personalities affect our approach and the way we conceptualize brand advertisement and decision making is a growing field of research in the scientific community.