The article is taken from the Journal of Retailing 85 (2, 2009, pages 145 to 158). The initial problem is pointed by the authors (Praveen Aggarwal and Rajiv Vaidyanathan from the University of Minnesota and Alladi Venkatesh from the University of California) is that with the wast amount of information available on the internet, there is no scientific method to analyse a brand’s reputation, based on what people (bloggers, reveiwers, retailers etc.) say about it. In this article called “Using Lexical Semantic Analysis to Derive Online Brand Positions : An Application to Retail Marketing Research”, the three researchers propose a “simple-to-use method that managers can utilize to assess their brand’s positionning“. More after the break.

Praveen Aggarval is PhD at the Marketing Department of the University of Minnestota (Labovitz School of Business & Economics, named after Joel Labovitz, businessman and patron from Duluth, MN), and Rajiv Vaidyanathan is Marketing Professor in the very same business school. Together with Alladi Venkatesh, Professor of Management and IT-expert from the University of California in Irvine, they stated the following problem : how can a product manager monitor, in a scientific way and based on the web content, the positioning of a brand?

Positioning a brand is, in marketing, the process of creating an identity and an image to a particular product or brand in order to differenciate it from its competitors. As we said in a previous article about the evolution of the brand-customer relationship, the web plays an increasing role in todays marketing, but this implies certain risks, which can be contra-productive. Without saying that product managers are losing control of their brands and products, they certainly have to deal with the multiplication of information sources about their products : blogs, customer reviews, retailer comments, comparative websites, tests etc. The retail industry is particularly competitive, that’s why branding is absolutely essential for managers (national brands as well as private brands).

In order to overcome these difficulties, the three specialists based their method on the semantic web (Berners-Lee, Hendler & Lassila 2001 ; Leuf 2006). Even if computers can’t understand text information, an analysis of the co-occurence of an adjective and a noun (brand, product) clearly give an insight of the web-users’ perception. To that extent, “the goal is to get a sense of the evaluation of brands in subjective sentences (…) by examining the association between the brand and various carefully selected adjectives or descriptors“. The aim therefore is to draw conclusions about the “brand’s online persona” based on a brand name + a certain adjective + a context.

To get the results, the three researchers basically used a beta version of Google’s API (Application Program Interface), which can be seen as the open-source part Google’s research program. The results returned (I leave the technical details aside, but the principle is to count pages that contain the brand name, and then those that contain the brand-adjective association) allow the draw a perceptual map, where differents brands can be compared. The analysis of P&G detergents reveal, for example, that Dreft and Ivory Snow (often associated with “baby”, rarely with “power”) seem to be perceived differently than Bold andEra (opposite results). The exception is the brand Cheer, to which both word are often associated. For Cheer and Era, we can note that the can design is identical, only te colour and the marketing “message” changes (typography, slogan etc.), which is apparently sufficient to differentiate the product images.

Further analysis is possible, for example on perfumes. The marketing researchers chose this particular product because of the importance of image and brand personality, “carefully crafted by marketers to position brands in a competitive marketplace“. Here, they point out the marketing efforts of brand managers. The method is particularly usefull for these marketers because of the “longitudinal analysis of a brand’s personality” that such monitoring permits. The conclusion states that their method does not account for negative qualifiers, what is quite suprising. I indeed wonder how such a study can be conducted if the tool does not differentiate “reliable”, “somewhat reliable” and “not reliable” ! But the authors ensure that “previous research has shown that the effects (of the polarity of the statements) are marginal (Pang et al. 2002)“. Very surprising…

Anyway, the method that the researchers developped enables marketing executives to track the positionning of their brands, particularly taking into consideration the dynamic nature of the web. In a moving environment, longitudinal tracking seems to become more and more important, and consumer understanding can be improved. Coped with consumer interaction, these means are strategic tools for web-based marketing.