Keywords

1 Introduction

Banner advertisements (ads) have been one of the most popular forms of online marketing. Surveys reported that banner ads revenue has been growing during the past years and will continue growing in the coming years [1]. Nonetheless, the effectiveness of banner ads is an ongoing debate. On the one hand, it is argued that banner ads have little effect, because consumers intentionally avoid viewing ads. This phenomenon is known as banner blindness [2]. Eye-tracking studies reveal that most banner ads are rarely directly attended to, but instead are processed at a pre-attentive level [3]. Banner ads may also increase users’ perceived workload, cause distraction, and are detrimental to primary task performance [3]. On the other hand, there is evidence showing that exposure to banner ads enhances brand awareness [2], attitudes towards brands [4], and repeat purchases [5].

The effectiveness of banner ads is usually measured in relation to attitudes, ad recognition, ad recall, click-through rates, and purchase intention. Prior research found that viewers’ attention is among the factors that determine the effectiveness of banner ads. For instance, levels of attention paid to an ad influences brand awareness. The longer viewers pay attention to banner ads, the better they remember and recognize the ads later [4]. Attention to online ads is also positively associated with recall and purchase intention [6]. Using eye-tracking technology, earlier studies have reported that even at a minimal level of attention (i.e., to the extent that viewers may not realize having seen the banner ads), viewers’ attitudes toward the exposed ads were still affected [4, 7]. This is because at low attention, information could be unconsciously learnt, which could subsequently change viewers’ preferences and attitudes [8]. In addition to the informational appeal of banner ads, emotional responses are also important in boosting advertising effectiveness [9]. Research suggests that emotional ads (i.e., ads that intentionally induce emotions) increase click-through rates [10] and enhance brand favorability [11]. In traditional media such as television, brand awareness is successfully built on emotional rather than rational appeals [12], as viewer engagement begins with “a conscious or more likely unconscious, emotional response” [13]. Also, emotional ads are more effective for certain product categories more so than others; e.g. emotional ads work better for hedonic than for utilitarian products and for products with low rather than high involvement [14].

Given the importance of attention and emotion in driving the effectiveness of online advertising this study investigates how these two factors (i.e., emotion and attention) influence the effectiveness of hedonic/indulgent food banner ads in an adult population. Hedonic or indulgent foods are defined as products that are attractive due to their tastiness yet cognitively unfavorable due to their high caloric content and/or low nutritional quality. Results in this study show that the longer web users paid attention to a product ad, the more likely they recognized the product later and that the impact of emotion on purchase intention varies at different levels of attention. In particular, at low attention, happy consumers were more likely to buy hedonic foods than less happy individuals. At higher attention levels, the opposite was observed.

There are several motivations for undertaking this study. First, very few food studies investigating the effects of food advertising have been undertaken with an adult population and the findings have been inconsistent [15]. Second, the potential interaction between emotion and attention and its consequent effect on the effectiveness of food banner ads have not been yet investigated. Third, most studies use upstream brand outcomes including attitude, recall, and recognition, or click-through rates as metrics of banner ad effectiveness [2]. However, as recall and click-through rates may not reflect ad effectiveness in low attention situations [2, 4], this study considers the outcomes of brand recognition and purchase intention in measuring ad effectiveness. Fourth, rather than relying on self-reported measures, this study assesses attention and emotions objectively using eye tracking and automatic facial expression detection technologies, a mixed-methods approach that has been underutilized in the context of online advertising.

2 Literature Review and Theoretical Foundation

2.1 The Customer Journey

Any advertisement is a “complex piece of communication”, and for it to be effective, the audience’s attention is required [16]. Visual attention is a multilevel selection process where some sensory information is processed deeper than others [17, 18]. This biological mechanism is essential to offset the limitation of human brain in simultaneously processing all the visual information received through the eyes [17]. Visual attention can be measured by visual fixation, a condition where the eyes remain relatively stationary over an object, mainly for optimal visual processing [19]. Eye fixation is obtained from an eye-tracker, which is a device that records where viewers look, how long they look, and other eye-related parameters such as blinking and pupil diameter [20]. Different measures have been used to quantify attention in online advertising such as percentage of clicks [6], number of fixations [3, 4, 21], total dwell time [21], and total fixation duration [4, 7]. Using different measures of attention may lead to different results as these metrics may reflect different aspects of attention. For instance, [4] found that the number of fixations was not significantly related to recognition but th e total fixation duration was. Total fixation duration (total time the eyes remain fixated) is usually adopted as a proxy for cognitive effort and the extent of cognitive processing of information [22]. Although eye tracking technology has been predominantly employed in recent years to explore the relation between attention and food choice/preference, a clear link has not yet been established. Earlier studies have reported on a non-significant difference in attention between savory and non-savory foods [23], while attention was shown to be automatically directed to tasty foods [24]. In an attempt to better attribute attention and food preferences, a model predicting food choice based on several eye-tracking parameters has previously been proposed [25]. Product categories used in the model ranged from healthy items (e.g., apple, salad) to beverages (e.g., soft drink, beer); hedonic snacks, however, were not included in the categories. The findings support the assumption that food decisions, especially decisions related to hedonic/indulgent items, could be a complex process [26] due to the “affectively attractive but cognitively unfavorable” [27] nature of the products.

Several perspectives have been put argued to explain the influence of emotions on food choice and food consumption, including: affect transfer; affective evaluation; affect regulation; and self-regulation. Affect transfer states that in the low thinking condition, objects associated with positive valence induce positive attitude and vice versa [28]. This psychological effect is often observed in food advertising targeting children [29]. When individuals are engaged in more effortful thinking, two mechanisms describe the influence of affect on food choice: affective evaluation and affect regulation [30]. Affective evaluation, or emotion congruence as a broader term, states that a person’s attitude or judgement is influenced in a direction which is congruent with his or her emotional state [28, 30]. Individuals who experience positive emotions form more positive attitude toward a stimuli and are more ready to engage in activities than those who are in negative emotional state [31]. Another model of explaining the effect of emotions under thinking conditions is affect regulation. This model posits that people in a pleasant state avoid actions that potentially destroy their state while people with negative affect tend to engage in actions that can instantly uplift their mood. If emotion changing cues in the environment are salient, affect regulation is likely to occur [32]. Among product categories, hedonic foods such as snacks high in sugar demonstrate a particularly prominent emotion changing property [30]. Last but not least, self-regulation explains food intake behaviors by assuming that self-control over eating functions on limited resources [31] which could be depleted by other cognitive processes. Negative emotions are believed to diminish such resources while positive emotions aid in their replenishment [33].

Given the multitude of theoretical perspectives put forth attempting to explain the effects of emotion on advertised audiences, the lack of sufficient empirical research in this area, and the mixed findings on the effects of food advertisements on adults, a priori hypotheses cannot be put forth. Instead, the following propositions regarding the effects of banner ads on web users are presented to guide this study.

  • P1. Emotional valence is positively associated with brand recognition

  • P2. Attention is positively associated with brand recognition

  • P3. Attention is positively associated with purchase intention

  • P4. The effect of attention on the relationship between emotional valence and purchase intention will be negative.

3 Methodology

3.1 Research Design

A lab experiment was conducted at a large Canadian university to gather data on web users’ visual attention, emotional response, and purchase intention following exposure to banner ads. Anonymous data collection was completed in one week. Participants received a 20$CAN Amazon gift card for taking part in the study. A mixed-methods approach yielded data collected by means of eye-tracking technology, facial expression emotion detection software, and a short online survey upon completion of the experiment. PROC GLIMMIX in SAS Statistical Software (SAS Institute Inc., Cary, North Carolina) was used in the data analysis, i.e. to test the effect of emotional valence, total visual fixation duration, and their interaction on brand recognition and purchase intention. The study was approved by the Ethics Committee of the authors’ institution.

3.2 Participants

A convenience sample of 48 adults, screened for normal or corrected-to-normal vision was recruited for this study. A single source bias test was conducted, where data were randomly rearranged (i.e., paired) so that each participant provided responses to only the independent variable or the dependent one [34]. The test results showed that brand attitude correlates with purchase intention for the full sample (n = 828; b = 0.6647) to a highly similar level as when using the paired sample (n = 414; b = 0.6382). Thus, only a minimal difference between the correlations of the two data sets exists (d = 0.0265), hence single source bias is not a threat in this study. Of the 48 participants recruited, 19 had incomplete fixation and valence data. Hence, the data analysis was performed on responses from 29 participants.

3.3 Procedure

Prior to the experiment, participants reviewed and signed an informed consent form, and were briefed about the experimental task. Next, participants were seated in front of a computer. A nine-point calibration test was performed to assure the accuracy of gaze tracking. Thereafter, participants were asked to go to a Walmart microsite and search for recipes that have Oreo products as ingredients. The purpose was to mimic the natural browsing process, where Internet users are usually occupied with their primary tasks. The microsite has a large rotating banner (i.e., a banner that automatically runs ads on a loop) on the header and a vertical banner on the left. The banners display snack products from different brands (see Fig. 1). Sixteen distinct banner ads were used, with each ad corresponding to a different brand (Oreo being one of the 16 advertised brands). The advertised products are mainly chocolate and cookies. Therefore, participants naturally encountered up to 16 brands while performing their primary task. After finding the recipe, participants were directed to a short online survey. The survey asked participants to self-report their attitude towards the brand, their purchase intention from the brand, as well as a brand recognition test.

Fig. 1.
figure 1

The microsite used in the experiment

Full size image

3.4 Apparatus and Measures

Noldus Observer XT (Noldus, Wageningen, Netherlands) and CubeHX (Montréal, Canada) [35] was used to performed the synchronization between attention (eye-tracking) and emotions (facial expression analysis) data, which are described below.

Levels of attention were measured by a Tobii X60 eye-tracking device (Tobii Technology, Sweden). The device uses near-infrared light to illuminate participants’ eyes, causing a reflection pattern on the corneas and pupils. The device’s sensors capture images of the eyes and the reflection patterns, which are then processed by advanced image-processing algorithms [20]. To determine the level of attention, an area of interest (AOI) was created for each naturally encountered product image displayed in each ad during the task, and the total fixation duration on each AOI was measured.

Participants’ emotions were extracted using FaceReader 6 (Noldus, Netherlands). FaceReader is an automatic facial expression analysis software. It analyzes facial expressions by: identifying the face region in a video image; creating a 3D-model of the face; classifying the model using an artificial neural network that was trained with nearly 2000 manually annotated images [36]. For any facial expression, FaceReader outputs intensity values of six basic emotions (i.e., sadness, surprise, disgust, anger, fear and happiness) and a neutral state. The outputs are logged with timestamps, allowing them to be synchronized with events [37] (see Fig. 2). Events occur when eyes fixate on an AOI, and intensity values are averaged over the fixation period.

Fig. 2.
figure 2

Automatic facial expression analysis

Full size image

Emotional valence (the negative/positive or unpleasant/pleasant dimension of emotion) is then calculated as the intensity of happiness divided by the sum of the intensity values of sadness, disgust, surprise, anger and fear.

After participants completed their experimental task, to test for brand recognition, a list of snack products (see Fig. 3) was presented on screen and subjects chose items that they remembered seeing on the microsite. Recognition is coded as 1 (one) for correctly recognized products and 0 (zero) for products incorrectly recognized or for products that were fixated but not recognized.

Fig. 3.
figure 3

Brands’ products presented to participants during the brand recognition test

Full size image

Purchase Intention

A list of brands whose products appeared in banner ads was presented to participants, who were then asked if they had intention to buy any of them. Purchase intention was measured using a 2-item, 7-point semantic differential scale: “very improbable vs. very probable” and “possible vs. impossible” [38].

4 Results

4.1 Recognition

Results reveal that participants’ ability to recognize brands subsequently to an initial impression of the brand in a banner ad is positively affected by their emotional valence (β = 2.437, p < 0.05). As participants feel more positive when looking at a product in banner ads, they are more likely to recognize the product later. Results also show that the more attention paid to an online product ad, the higher the recognition probability (β = 0.565, p < 0.05). Thus, propositions 1 and 2 are supported (Table 1).

Table 1. Regression results: emotional valence and attention on recognition
Full size table

4.2 Purchase Intention

Results demonstrated that attention (see Fig. 4) is positively related with purchase intention (β = 0.758, p < 0.001); hence, participants’ intent to buy increases as they pay more attention to the online banner ads. Furthermore, the effect of attention on the relationship between emotional valence and purchase intention was significant at one-tailed (p = 0.035). In particular, for fixations less than 1.41 s (based on 1.41 = 1.009/0.717), purchase intention is positively affected by emotion, whereby consumers are more likely to make a purchase when they experience a positive emotion (i.e. happiness). In contrast, when the visual fixation on the products is greater (fixation duration ≥ 1.41 s), consumers’ likelihood to buy the advertised brand’s product tends to decrease as emotional valence increases. The plot in Fig. 5 illustrates this interaction effect for total fixation durations of 200 ms and of 5 s. Thus, propositions 3 and 4 are supported (Table 2).

Table 2. Regression results: emotional valence and attention on purchase intention
Full size table
Fig. 4.
figure 4

Heatmap of user visual attention

Full size image
Fig. 5.
figure 5

Valence and purchase intention, by attention

Full size image

5 Discussion and Concluding Comments

This study examined the effects of attention and emotional valence and attention paid to banner ads on the user’s ability to subsequently recognize advertised brands’ products and their intention to purchase them. Forty-eight participants searched for a recipe on a designated website. Using eye tracking technology and automatic facial expression analysis to infer emotional valence, results from this study confirm the positive relation between attention and brand recognition. Also, an interaction effect of attention and emotions on brand recognition was not found. This means the impact of emotions on brand recognition is not dependent on attentional levels and vice versa. This result suggests the impact of emotions on brand recognition, and, more broadly, brand awareness, even at low levels of attention. In addition, at low level of attention, the intention to purchase indulgent snack food increases as positive valence increases. Simply put, consumers who experienced happiness during viewing of the banner ads expressed a greater intention to buy the advertised product. At higher levels of attention, the opposite was observed. When consumers feel positive but spend a longer time processing a banner ad, their likelihood to buy decreases. This may be attributed to the availability and utilization of cognitive resources during the processing of the advertisement, making the web user more skeptical and critical of the advertisement’s message, thereby lowering their willingness to purchase the advertised product.

This study also demonstrates how affect influences purchase intention of indulgent food varies at different levels of attention to banner ads. If visual attention is assumed to be “the amount of conscious thinking going on” ([12], p. 67), its effect could be interpreted in the context of low and high thinking conditions. In the low thinking/low attention situations, consumers’ intentions of purchase are in accord with their valence. Underlying mechanisms could be deduced from the affect primacy theory [39] or the implicit misattribution model [40]. The former states that affective responses are first reactions before any cognitive activities. Thus, without further cognitive activities take place, responses are likely to be in accord with the affect experienced. The implicit misattribution model assumes the automatic attribution of one’s emotional state to the target objects. For example, without thoughtful evaluation, consumers in a happy state may engage in a purchase decision because they unconsciously link sources of their positivity to the product. On the other hand, for higher thinking/higher attention situations, emotional valence is negatively associated with the purchase intention of hedonic foods. This could be explained from the theoretical lens of affect regulation (i.e., people tend to behave in a direction that improves their negative mood or maintains their positive mood) or the self-regulation hypothesis (i.e., consumption of hedonic food is the result of self-control failure due to negative emotions), although the latter is less likely to account for the behaviors in the context of food choice [30]. While self-regulation fails to explain the phenomenon in this context, it is plausible that varied mechanisms might be triggered for different food types.

For practitioners, the study highlights the importance of setting marketing goals when designing banner ads. An ad campaign for the purpose of creating brand awareness should be different from one created to elicit purchase intention. Also, as shown, varied metrics for measuring advertising effectiveness can yield different results. Thus, choosing the right measures to evaluate the impact of digital ads is critical.

This exploratory study paves the way for a broader research program in relation to disentangling the factors impacting the effectiveness of banner ads. Future research could involve a between-subject research design where the groups are exposed to distinct manipulations of the creative execution (i.e., the banner ad design) or test for ‘environmental’ factors ranging from the effect of webpage content congruency with banner ad content to the effect of purchase decision involvement on the current understanding as emergent from the results of this study. Additionally, qualitative data could have be collected so as to enrich the understanding of the mechanism explaining the effectiveness of banner ads. Lastly, other quantitative data collection methods may be utilized, including eye-fixation related potential [41], so as to delve deeper in the neurophysiological bases of the studied phenomena.