Analysing formal visual elements of corporate logotypes using computational aesthetics

Kitsopoulou, Vasiliki (2018) Analysing formal visual elements of corporate logotypes using computational aesthetics. Doctoral thesis, University of East Anglia.

[thumbnail of Vasiliki_Kitsopoulou_PhD_thesis_-_6524036.pdf]
Download (5MB) | Preview


The marketing mix contains a significant proportion of elements that derive their appeal and effectiveness from visuals. This thesis proposes the application of quantitative measures from the literature on computational aesthetics to evaluate and study the formal characteristics of corporate visuals in the form of logotypes (logos). It is argued that the proposed approach has a number of advantages in terms of efficiency, consistency and accuracy over existing approaches in marketing that rely on subjective assessments. The proposed approach is grounded on a critical review of a diverse literature that encompasses Marketing, Art History and Philosophy, and, Visual Science and Psychology. The computational aesthetic measures are framed within the construct of Henderson and Cote (1998) and van der Lans et al. (2009), in order to analyse brand logo design elements along with their effect on consumers. The thesis is underpinned by three empirical studies.
The first study uses an extensive set of 107 computational aesthetic measures to quantify the design elements in a sample of 215 professionally designed logotypes drawn from the World Intellectual Property Organization Global Brand Database. The study uses for the first time an array of different measures for evaluating design elements related to colour that include hue, saturation, and colourfulness. The metrics capture both global design features of logos along with features related to visual segments. The metrics are linked to logo elaborateness, naturalness and harmony, using the theoretical framework of Henderson and Cote (1998). The results show that measures have a very diverse behaviour across metrics and typically follow highly non-normal distributions. Factor analysis indicates that the categorisation of the measurements in three factors is a reasonable representation of the data with some correspondence to the dimensions of elaborateness, naturalness and harmony.
The second study demonstrates that the proposed computational aesthetic measures can be used to approximate the subjective evaluation of logo designs provided by experts.  
Specifically, eight design elements for the sample of 215 logos, corresponding to harmony, elaborateness and naturalness, are evaluated by three experts. The results show for the first time that computational aesthetic measures related to colour along with other measures are useful in approximating subjective expert reviews. Unlike previous literature, this research combines both standard statistical methods for modelling and inference, along with more recent techniques from machine learning. Linear regression analysis suggests that the objective computational measures contain useful information for predicting proxy subjective expert reviews for logos. Model accuracy is substantially improved using neural network regression analysis based on Radial Basis Functions.
The last study examines the role of consumer personality traits as moderators of the effect of perceived logo dynamism on consumer attitude towards the logo. One hundred and twenty-two participants were asked to evaluate elements of logo design (visual appearance, complexity, informativeness, familiarity, novelty, dynamism and engagement), their attitude towards the brand and their personality traits (sensation seeking, risk taking propensity, nostalgia and need for cognition). The estimates extracted were shown to vary significantly in terms of central tendency and dispersion and mostly follow non-normal distributions. Following Cian et al. (2014) the moderated mediator model by Preacher and Hayes (2008) is applied to test the suitability of personality traits as moderators of the effect of logo dynamism on attitudes towards the logo. The personality traits used as moderators are Need for Cognition and Risk-Taking Propensity, whereas Engagement was used as a Mediator. This is the first study to employ personality traits as moderators in such a study using this methodology. The results offer limited support of the role of personality traits as moderators in this relationship. Therefore, the study strengthens the case for the development of objective measures of visual characteristics.

The working hypothesis in the thesis is that, with the help of computational aesthetic measures, marketing visuals such as corporate logos, can afford themselves to a consistent quantitative approach which can prove to be important for researchers and practitioners alike. By being able to group and measure the aesthetic differences, similarities and emerging patterns, access is gained to a new family of metrics, which can be applied to any type of logo across time, product, industry or culture.

Item Type: Thesis (Doctoral)
Faculty \ School: Faculty of Social Sciences > Norwich Business School
Depositing User: Jackie Webb
Date Deposited: 06 Dec 2018 12:58
Last Modified: 24 Jan 2021 01:38


Downloads per month over past year

Actions (login required)

View Item View Item