Keywords

1 Introduction

As over the course of time original equipment manufacturers (OEMs) in the automotive industry have assimilated their technical properties, design has become the most important factor leading a person to the decision of buying a certain vehicle. [2, 5, 7, 9] Consequently OEMs - especially within the premium sector – put huge efforts on the topic of exterior styling.

In order to match a customer’s liking, OEMs conduct so-called car and concept clinics. [8] These clinics are head-to-head comparisons of competing vehicles wherein target-market consumers are recruited to evaluate the upcoming models and indicate preferences. The clinics’ results are of high interest as OEMs take high risks by introducing a new vehicle concept or a newly designed model without a quantified consumer feedback, [6] however, contemporary car clinics generate great expenses as real prototypes are needed. Furthermore, these required prototypes are available quite late in the development process in order to implement distinct changes potentially causing high investments. [6] As a result, OEMs investigate intensively how to integrate and advance customers’ feedback within the development process at an earlier stage. [17]

Considering these facts car clinics with virtual models would help to solve these problems and would bring a huge benefit regarding timing, logistics and costs as they only require a digital model of a new concept or design instead of an actual, real prototype. [3, 8] However, until today – to our best knowledge - only the scientific research by A. Erdmann [8] and M. Söderman [15] have covered the concrete question whether or not virtual car clinics could replace real car clinics. Within his doctoral thesis Erdmann concluded that virtual vehicle models would provide the same output as real vehicle models but it has to be stated that a mostly qualitative questionnaire was applied which was not assured to be capable of differentiating. [8] In line with Erdmann’s thesis Söderman also only operated with qualitative questions, he, however, concluded that real and virtual prototypes do not provide the same results. [15] According to these diverging conclusions and lack of standardized instruments further investigation on this topic is required.

Analyzing the scientific approaches of how the customer’s perception of automotive design is constituted, sportiness is always a recurring aspect. [12, 14, 16] Referring to A. Oehme attractiveness and dynamics contribute significantly to the overall judgment on aesthetics. [14] Furthermore, several premium OEMs maintain sportiness as one of their central brand characteristics, e.g. Audi, BMW and Porsche. Taking these findings into account perceived ‘Sportiness’ of a vehicle exterieur was chosen as the central aspect for further investigation.

Hence, the following empirical study was designed and executed to address the question whether the perception of the overall sportiness of a vehicle exterior can show the same results for virtual as well as for real vehicle models. Due to the fact that there has been no standardized instrument available yet to describe and to measure the concept of a vehicle exterior’s sportiness [1, 5] this had to be developed and validated first.

2 Development of the Instrument

2.1 Method

In order to develop the standardized instrument a three-part approach was chosen. Firstly, a survey was conducted to gather all items which are related to a vehicle exterior’s sportiness. In the second step, raw data was edited and finally an extensive empirical study was executed to analyze the data and to identify the latent construct with its relevant factors concerning the perception of a vehicle exterior’s sportiness.

The approach’s first step was collecting the items. In this context, the main goal was to carve out all attributes which are exclusively linked to the sportiness of vehicles and not beyond as the item-set influences the final results of the exploratory factor analysis [10]. This means the resulting item-set as well as the factorial structure differ depending on the input of variables. [1] For this purpose the data collection was divided into three perspectives. The first part resembled the perspective of an original equipment manufacturer, the second one of the automotive press and the third one of the customer. Relating to the automotive press 300 articles were scanned for expressions considering descriptions of exterior’s sportiness. The perspective of the OEMs was simulated by reviewing 60 OEM model pamphlets for sporty items. At last the customer point of view had to be analyzed what was done by an online panel with 115 test subjects.

After data collection the selection was enlarged by items gathered from scientific literature concerning existing questionnaires about related aspects of automotive forms [1, 1014, 16]. The final item pool resulted in an entity of 150 items.

In order to detect very similar or even redundant attributes, four experts were instructed to perform card sorting independently of each other. The final item-set consisted of a total of 103 items which were the basis for the following study.

Considering the perception of a vehicle exterior’s sportiness the main goal of the following investigation was the identification of the latent construct (Fig. 1). In this context an online-survey was designed and conducted wherein the subjects were asked to assess one stimulus on the basis of the item-set preliminarily defined. The link was distributed via e-mail and social media. By clicking on the link the subjects were randomly directed to one of the different stimuli and asked to assess the exterior using a five point likert-scale for each of the items. All in all 108 items were integrated into the survey in random order including additional antonyms. The stimuli presented were nine images of vehicle exteriors which can be classified into the categories van, station wagon and sports car: Renault Kangoo, VW Caddy, Mazda Biante, Mercedes-Benz C-class T-model, Volvo V70, VW Golf Variant, AUDI R8, BMW i8, Mercedes-Benz SLS AMG. These vehicle categories were chosen as they were supposed to differentiate distinctly considering the vehicle’s sportiness. Regarding participation during three weeks of activation 301 subjects attended the online panel. 79.1 % out of them were male and the average age was 32.79 years.

Fig. 1.
figure 1

Screenshot of the online panel

2.2 Analysis

In the context of data analysis the first calculations concerned the specific item values. Relating to these computations, items were excluded which did not differentiate or measure the same property as the remaining item-set. Before starting the exploratory factor analysis the criteria KMO, MSA and Bartlett’s test on sphericity had to be checked. [4] As all quality criteria were achieved the dimensional reduction could be performed.

After confirming the data set to be valid the most suitable approach of exploration had to be determined. In line with determining the general methodology the main axis analysis was chosen as it is best for tracing relations between items. [4]

Before fixing the rotation methodology the number of factors had to be defined. Summing up the different approaches – Kaiser-Guttman, Screeplot, MAP and Parallel-Analysis according to Horn – three to seven latent factors could be anticipated.

Regarding the rotation technique orthogonal methods were not applied because they did not fulfill the requirement of an easy structure. So the most common oblique rotation – the Direct-Quartimin method was used. [4]

The structure chosen explains 65.6 % of the variance and consists of six depending factors which are relying on 33 items. Regarding the item clusters the next step was defining suitable factor labels which clearly describe what is measured – see Fig. 2 with the original German items and their translation.

Fig. 2.
figure 2

Construct of a vehicle exterior’s sportiness (originally in German)

The six depending factors were labeled as ‘attractiveness’, ‘aggression’, ‘modernity’, ‘perfection’, ‘premium’ and ‘functionality’. Regarding other approaches to describe general automotive design their factors are closely related to the concept of sportiness, e.g. ‘attractiveness’, ‘emotionality’, ‘innovation’, ‘quality’ and ‘dynamics’. [12, 14]

The next step implies the questionnaire’s derivation from the online panel’s methodology. In order to stay consistent the five point likert-scales and the final item-set were adopted. Regarding the questionnaire’s analysis an unweighted approach which results in one value per factor has been chosen. These values are based on the calculation of the factors’ arithmetic means which depend on the associated items. This approach provides a subdivided image of a vehicle’s exterior sportiness which needs no more interpretation because the higher a rating the higher the perception of the vehicle exterior’s sportiness.

3 Assessing a Vehicle Exterior’s Sportiness in a Virtual Car Clinic

After defining an instrument for measuring the subjective perception of a vehicle exterior’s sportiness the following research concentrates on the question whether the evaluation of virtual and real vehicles provides the same results based on the developed questionnaire.

3.1 Method

The survey was designed against the background that the conditions match in the real and virtual environment. In order to grant identical stimuli all vehicles used in the current study were by Audi. This decision was based on the fact that the required virtual holistic vehicle models are only available at Audi in-house. In accordance to comparability the vehicles chosen were of the same concept. Finally, two current Sport Utility Vehicles (SUVs) were selected, labeled as model A and model B.

Regarding the survey’s design a within-subject design was set. Each test subject had to assess all four vehicles, two real and two virtual models. Based on a middle-sized effect the a priori calculation provided a minimum sample size of 36.

As the clinic’s location a spacious hall on neutral terrain, outside Audi’s premises was chosen. The room was lined by molleton and was divided into two areas according to reality and virtual reality. The virtual models were visualized on a scale of one to one on a 2D back-projection screen which also served as camouflage between the two areas. The back-projection was chosen as it avoids shading by test subjects. A 2D projection was applied based on the determination that the entire exterior should be visible at any time. This constraint led to a minimum virtual distance of the subject to the screen which measured about two meters. According to Audi’s experts these conditions supersede 3D visualization as the 3D effect gets lost. A head mounted display (HMD) was also neglected as HMDs can increase the probability of cyber-sickness and unforeseeable effects might arise as most of the subjects have no experience with HMDs. [8]

In order to grant similar conditions in reality and in virtual reality the virtual vehicle presentations were based on pre-rendered high definition images which replicated the subjects’ trajectory when walking around the real vehicle on a specified distance. This correspondence was realized by the fact that the test subjects were asked to follow a taped line on the floor while examining the real vehicle. This oval line around the real vehicle with a constant distance of two meters represented the virtual trajectory. In order to completely define the virtual trajectory the eye height of a 50. percentile man was set. Figure 3 shows the survey’s setting from above.

Fig. 3.
figure 3

Setting of the study (‘S’ represents the subject)

In line with immersion an application was implemented which enabled the test subjects to ‘walk’ around the virtual vehicle. The interaction of the test subjects with the virtual model was carried out by the arrow keys of a keyboard provided. Pressing the arrow key to the left let to the virtual vehicle turning to the right which conveyed the operator to walk clockwise. On the one hand the application allowed to rotate the vehicle fluently and on the other hand subjects could stop at any point. This should simulate the natural exterior examination (see Fig. 4).

Fig. 4.
figure 4

Comparison of the subject’s perspective: real vs. virtual

The study itself started with the instruction of the test subjects whereupon it has to be stated that the survey was always performed by a single person and they were not informed about the research’s purpose. In order to avoid any order dependencies the order of stimuli presentation was permuted for each subject. After the instruction the assessment of the perceived sportiness of the vehicle’s exterior was done through the previously described questionnaire. All in all the survey lasted about one hour per person.

3.2 Results

In the end 41 subjects attended the study. The average age was 37.4 years (SD = 15.8), 73.2 % were male, 19.5 % had already had experience with virtual reality and 48.8 % were employed in the automotive industry which included employees at an OEM as well as employees at service providers.

The main focus of the following analysis is on the conformities between reality and virtual reality. All results are based on the developed questionnaire measuring the perceived sportiness of a vehicle’s exterior and rely on the factors’ arithmetic means (see Fig. 5).

Fig. 5.
figure 5

Results of model A and model B: virtual vs. real

When comparing individual ratings aggregated on factor level high correlations can be found between real and virtual conditions (r = 0.675). Separating the models’ results for further investigation the analyses between virtual reality and reality show strong effects – see Table 1.

Table 1. Correlation between reality and virtual reality separated by vehicle and factors

In addition an ANOVA with repeated measures on the three factors ‘Vehicle’ (Model A vs. Model B), ‘Condition’ (Real vs. Virtual) and ‘Dimension’ (Attractiveness, Aggression, Modernity, Perfection, Premium, Functionality) was carried out to investigate the ratings’ absolute differences. In the following analysis only those effects will be discussed which are relevant to the current methodological contribution.

For the factor Vehicle a just significant main effect was found, F(1, 40) = 4.10, p = .050, \(\eta_{{\rm p}}^{2}=.093\). For the factor Condition the main effect also reached significance, F(1, 40) = 16.4, p < .001, \(\eta_{{\rm p}}^{2}=.29\). For the factor Dimension a significant effect was discovered as well, F(5, 200) = 114, p < .001, \(\eta_{{\rm p}}^{2}=.74\).

For the interaction between the factors Condition and Vehicle a small effect was detected on α = 5 % level, F(1, 40) = 4.15, p = .048, \(\eta_{{\rm p}}^{2}=.094\). As this interaction indicates that the ratings under virtual vs. real conditions differed between the two vehicles, a Post hoc test (Newman-Keuls) on this interaction was executed. Results show no significant differences in the ratings between virtual and real conditions for model B but for model A, p = .001. Referring to model A, virtual vs. real, the Post hoc test (Newman-Keuls) revealed significant differences for the dimensions Attractiveness (p = .008), Aggression (p < .001) and Premium (p = .003) but no significant differences for the dimensions Modernity, Perfection and Functionality. If differences occurred ratings under virtual condition were lower than under real condition.

4 Discussion and Outlook

In the current contribution an empirical study was designed and executed to address the question whether the perception of the overall sportiness of a vehicle’s exterior can show the same results for virtual as well as real vehicle models. In order to measure the subject’s perception of sportiness a questionnaire was developed and validated first. Its final concept describes six depending factors based on 33 items: ‘attractiveness’, ‘aggression’, ‘modernity’, ‘perfection’, ‘premium’ and ‘functionality’.

Within the empirical study the subjects had to evaluate two current Audi SUV models under real and virtual conditions. Results show that the overall ratings under real and virtual conditions correlate highly. Still absolute differences have been identified between the captured subjective ratings of real and virtual models, especially for one of the observed vehicles and under certain dimensions. If absolute differences occurred evaluations under real condition revealed to be higher than ratings than under virtual condition.

It could be assumed that the application of a virtual presentation lowers the intensity of a visual impression compared to the real world. In line with these findings it has to be hypothesized what might have influenced this. Firstly there is the survey’s setting with its 2D back-projection, the projection’s resolution and the pre-defined subject’s trajectories in reality and virtual reality. For all of these three parameters more extensive alternatives could have been applied, e.g. 3D projection, 4 k resolution, higher degree of freedom by real-time rendering and a higher level of immersion by a HMD. While absolute differences were dependent on stimulus set and results were diverging between model A and model B, this dependencies have not been fully understood yet. In order to grant assured knowledge the parameters listed need to be objects of further research.

All in all the results of the current study could show a high potential for the application of virtual car clinics especially for the assessment of perceived sportiness of a vehicle’s exterior in the future. This contribution also shows the need for a methodically developed questionnaire in order to define and measure a certain latent construct relevant for customers’ perception of a product – here the perception of a vehicle exterior’s sportiness. It could be shown that the developed instrument can show very similar results for virtual as well as real vehicle models.

Regarding virtual car clinics it can be assumed that the current approach could be applied to other areas of customer perception as well – given the existence of appropriate instruments/questionnaires to access the relevant constructs. In the future virtual car clinics can enable OEMs to advance customers’ feedback early within the development process. This shift is highly requested as it saves time and money and will allow faster customer centered redesign iterations in the future.