Estimate the Market Share from the Search Engine Hit Counts

Robert Viseur

Faculté Polytechnique (Université de Mons), Rue de Houdain, 9, 7000 Mons, Belgium

Keywords: Marketing, Advertising, Market Share, Media Share, Share of Voice, Word-of-Mouth, Management of

Innovation, Search Engine, API, Webometrics.

Abstract: The knowledge of the competitive environment (and, in particular, market share) is an important factor in

the management of innovation. This type of information is not always accessible to small and medium

enterprises. In addition, some sectors are changing rapidly under the pressure of technological change. We

propose in this research a method for estimating the market share based on media share, based on the hit

counts returned by search engines for each brand. We show the potential of this approach with a real

example (the automotive industry) and discuss the limitations associated with the operating mode of search

engines.

1 INTRODUCTION

In an article published in 2005 entitled “When

search engines occupy the media space...“, Olivier

Andrieu, a French specialist in commercial search

engines, noted that the Google search engine

exceeded its competitors (Yahoo, MSN and, on the

French market, Exalead) not only in terms of market

share but also in terms of media coverage. Google

produced from two to three times more adverts than

its competitors. The measure of the number of

adverts was based on the press review of a major

news website devoted to commercial search engines.

In the management of innovation, market share is

an important parameter for the understanding of the

competitive environment (Lambin, 1998); (Porter,

1992). Information on market share is generally

provided by panels of consumers and retailers. In the

absence of official or professional statistics, the

company must buy or edit this information. Access

to this information presents several challenges.

Firstly, the definition of market share can vary

depending on the data source (Lambin, 1998). Does

it speak of market shares in volume or revenue,

absolute or relative market share, etc.? Secondly, the

VSEs (Very Small Enterprises) and SMEs (Small

and Medium Enterprises) do not always have the

resources to collect, acquire and operate this type of

data. Thirdly, some markets have experienced rapid

and significant turbulence under the influence of

new technologies (Millier, 1997). Historical data

may therefore be quickly outdated.

The need for updated information on the

competitive position of a company (and, in

particular, the market shares of its competitors) and

the existence of a possible link between market

share and media coverage of the company concerned

motivated us to study the relationship between

media share and market share, as well as the

feasibility of deriving a simple and affordable way

of estimating market share. Market share is a well-

known concept, which means “the percentage of

sales held by each competitor in the market” (Kotler

and Dubois, 2000; p255). The concept of "media

share" that is exploited here will be defined by

analogy with market share, as “the percentage of

published documents citing a company compared to

the number of documents about companies in the

relevant market”. It is here estimated from hit counts

related to the results of a commercial search engine.

Our research is applied in the automotive market.

Our research is divided into three parts. Firstly

we will describe the state of the art. We shall

examine the link between advertising and market

share, and deepen the definition of media share. We

will then document the potential biases in the

method, including those induced by practical

operating conditions of commercial search engines.

Secondly we will present our methodology. Thirdly

we will present our results and discuss them,

particularly with regard to results from comparable

approaches.

112

Viseur R..

Estimate the Market Share from the Search Engine Hit Counts.

DOI: 10.5220/0004595101120117

In Proceedings of the 2nd International Conference on Data Technologies and Applications (DATA-2013), pages 112-117

ISBN: 978-989-8565-67-9

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

2 STATE OF THE ART

2.1 Advertising and Market Shares

Lambin (1998) defines marketing communications

as “the set of signals issued by the company towards

its audiences” (p. 615). The company has a set of

tools to facilitate the matching of supply and

demand. Advertising is part of the communications

mix, alongside sales promotion, public relations,

salesforce and direct marketing. It is used by

organizations to develop their reputation or that of

their products, services or ideas (Kotler, 1999).

The relationship between advertising and market

share has long been a subject for study in marketing.

The existence of a causal link between advertising

on the one hand and sales on the other hand is thus

assumed (Eagle et al., 2005). However its

importance according to the market, the company

size or the type of product is still discussed.

2.2 Concept of Share of Media

The concept of “media share” will be defined by

analogy with market share, as “the percentage of

published documents citing companies compared to

the number of documents about companies in the

relevant market”.

This concept is differentiated from “share of

voice”, “social share of voice” and “word-of-

mouth”. Share of voice is the share of advertising

expenditure of the company compared to total

advertising spending in the relevant market. Social

share of voice refers to the importance of a term

compared to a set of terms on social medias

(Emerson and al., 2012). This notion is closer to

word-of-mouth, which means existing interpersonal

communication, especially between the consumer

and the environment, face to face or, increasingly,

on the Web (Kotler and Dubois, 2000).

The concept of media share is clearly

differentiated from share of voice (there is no

question of budget). It is, however, closer to social

share of voice. Its scope is broader than just social

networking. Share of media can be related to word-

of-mouth, as it includes its effects. Share of media

also includes the communication about the brand in

the press or communications operations relayed by

the company itself (e.g. press releases).

2.3 Possible Biases within the proposed

Method

2.3.1 Reliability of Hit Counts by

Commercial Search Engines

Using results from a search engine involves a

decision on how to interact with it. For example,

Google (google.fr) offers two ways of accessing its

search engine: manually, through its WUI (Web

User Interface), and through its API (Application

Programming Interface) (code.google.com). The

latter has the apparent advantage of permitting the

automation of queries, thus facilitating the repetition

of various tests over time. However this opportunity

comes up in practice against large differences in

results observed between the results from the Web

interface and those derived from the API, with the

same search parameters (same keywords, same

geographical area or language). This would not be

crippling to the researcher if Google did not prohibit

the execution of automated queries on the WUI,

sparking a spontaneous use of the API in this

particular context of use (McCown and Nelson,

2007).

The use of quantitative data from the search

engines is part of a research field called webometrics

(Thelwall et al., 2005). The behaviour of APIs has

been studied in the literature. McCown and Nelson

(2007) found significant differences in hit counts

between the WUI and API of MSN, Yahoo! and

Google. They confirmed the results obtained by

Mayr and Tosques (2005) with the Google API.

However, problems are encountered beyond the

simple use of search engines API. Thus recovery is

weak between the results of different search engines

(Véronis, 2006). Despite the global reach of major

commercial search engines, geographic bias may

exist. In addition, the hit counts of search engines

are not stable over time, while complex queries do

not always give results consistent with the theory of

sets (Boolean logic) (Viseur, 2012a).

The importance of the problems varies from one

tool to another. Early in 2012, Bing presented a

more predictable behaviour than its competitor

Google, when using both API or complex queries

(Viseur, 2012). Due to the speed of technological

developments it is nevertheless necessary to

periodically reassess these conclusions. Using the

Web user interface (WUI) for the collection of

search results using simple queries globally poses no

problem whatsoever on Google (google.com) or

Bing (bing.com).

EstimatetheMarketSharefromtheSearchEngineHitCounts

113

2.3.2 Representativeness of Web Users

Several media can be selected for a campaign: press,

television, billboard, radio and cinema (Kotler and

Dubois, 2000). A sixth media was added in the

nineties: the Internet. It has gradually grown in

importance. In 2005, more than one out of two

French people were connected (Roustan and al.,

2005).

However, are consumers on the Web

representative of all consumers in the market ? In

practice, they tend to be.

According to Médiamétrie

(www.mediametrie.fr), the number of connected

French people has increased in size from 27.21 to

38.27 millions between December 2005 and

December 2010. And, according to the IAB

(www.iabfrance.com), Internet has become the

preferred medium of French people just behind

television. The public has also logged fewer

specificities.

Furthermore, the Internet significantly affects the

purchasing process, fostering opportunistic

behaviour and encouraging cross information

upstream of the buying process. Valuing the

opinions of peers is becoming more important and is

a “tool of redistribution of power between

consumers and producers-distributors” (Roustan et

al., 2005; p12). More than just a channel of

purchase, Internet currently appears to be an

information channel strengthening consumer

expertise. Even if not buying online, the user

willingly prepares his/her purchase through

information on the Internet. Search engines are a key

access point to information behind the websites of

famous stores.

2.3.3 Impact of Negative Word-of-Mouth

However the operation of search engines can cause

unexpected effects in the way the reputation of

young companies grows. The online store

DecorMyEyes (decormyeyes.com), which sells

glasses, boasted at the end of 2010 of how it had

really commercially taken off after creating a

negative buzz around its brand. Negative opinions of

disgruntled customers (and intentionally mistreated

by the company) have been generating traffic and

backlinks to the company website, helping the

company to improve its Web positioning (Segal,

2010).

The atypical communication strategy adopted by

DecorMyEyes is an opportunistic exploitation of a

current weakness of search engines. They do not

actually measure the semantic orientation

of the

information published in the context of a quote link.

They are mainly based on syntactic relevance and

backlinks (seen as citations), thus following the

principles of the PageRank algorithm (Brin and

Page, 1998); (Duffez and Andrieu, 2004).

This shortcoming of search engines is a possible

bias in the estimation returned by the search engine.

A company would be heavily criticized and

associated with a bigger estimated market share.

This bias is compounded by the fact that a

dissatisfied customer tends to speak more than a

satisfied customer. The customer who received an

effective solution speaks positively of his experience

to at least five people around him (Kotler and

Dubois, 2000). By contrast, an angry customer will

talk about his misadventure to 11 people, heavily

affecting the reputation of the company (Kotler,

1999). A communication policy based on a negative

buzz may thus jeopardize the company's sales over

the long term. The consumers' behaviour will

depend on the final processing of the complaint.

Indeed, the loyalty of a customer depends on his/her

satisfaction (Kotler and Dubois, 2000).

This is particularly the case in competitive

markets (Lambin, 1998). A claim, reflecting

dissatisfaction, does not necessarily lead to customer

defection, if it is addressed appropriately.

3 METHODOLOGY

We propose to study the relationship between

market share and the media share. The media share

will be estimated from the number of results (hit

counts) returned by a search engine for a considered

brand.

We carried out our study in the automotive

market, because its sales figures are readily available

and provided by recognized national organizations

(FEBIAC in Belgium, CCFA in France, etc.). We

will use the figures for 2010. Less volatile and well

documented, this market seems appropriate for

validating the usefulness of our approach.

The hit counts estimated by search engines have

reliability problems (Viseur, 2012a). Our

experimentation showed the Bing search engine

provided the most consistent results, and we used the

number of results estimated by the Microsoft Bing

search engine. The operator “loc:” was used for

A sentence is characterized by a positive semantic orientation

when it has positive associations, negative otherwise (Turney,

2002).

DATA2013-2ndInternationalConferenceonDataManagementTechnologiesandApplications

114

geographic targeting (“FR”, “DE”, “GB”). The

measurements on the search engine were made in

late 2011, after preliminary tests in 2010. Note that

the calculation of the media share may require a

reformulation of queries (e.g. Volkswagen is often

called “VW”, Opel cars are sold under the Vauxhall

brand in the UK, etc.) or the elimination of certain

brands (e.g. “smart” is a common term in English).

We will verify the following hypothesis: “Market

share and media share are proportionate”.

We will not test the hypothesis that the market

share is negatively influenced by negative

communication about a brand because of the

practical difficulty in assessing the importance of

positive or negative communication. Turney's

method (2002) requires the execution of a large

number of automated queries, and relies on

linguistic tricks that are difficult to generalize in this

case (use of a specific vocabulary for product

reviews on the Internet, for example).

4 RESULTS

The correlation (Pearson coefficient) between the

market share and media share of 25 French brands

stands at 0.73 (see Table 1). The correlation can be

considered as strong.

This correlation is higher when we distinguish

between the premium brands (Audi, BMW,

Mercedes, Alfa Romeo, Volvo, Lancia, Land Rover,

Porsche and Lexus) with a distinctive character

linked to exclusivity (Štrach and Everett, 2005), and

the general brands (others). The correlation (Pearson

coefficient) between market share and media share

for generalist brands amounted to 0.83, against 0.88

for premium brands.

Table 1: Correlation coefficient.

Germany France United-Kingdom

All brands

0.76 0.73 0.78

Generalist

carmakers only

0.75 0.83 0.89

Premium

carmakers only

0.75 0.88 0.56

The premium brands also have a greater media

share than their market share.

A similar study on German data gave

comparable results, with a correlation of 0.76

(Pearson coefficient) by taking the first 18

manufacturers, 0.75 taking only generalist carmakers

and 0.75 with only premium manufacturers.

The results are comparable with the UK, except

for premium brands. The correlation of 0.56 for

premium carmakers can be explained by the absence

of some popular brands in the UK (such as Jaguar),

for which the market shares were not disclosed.

5 RELATED WORKS

Uncles et al. (2010) studied the impact of word-of-

mouth on market share. They distinguished the

effects of the total volume of word-of-mouth

(WOM), the volume of positive word-of-mouth

(PWOM) and the volume of negative word-of-

mouth (NWOM). The authors confirm the existence

of a strong correlation between market share on the

one hand, and word-of-mouth (WOM) and positive

word-of-mouth (PWOM) on the other hand.

Media share is strongly influenced by word-of-

mouth on the Web. The results of the Uncles et al.,

(2010)'s study tend to confirm the validity of our

approach.

Xu et al., (2010) chose a similar approach to

ours. They studied the correlation between quarterly

market shares of cell phone manufacturers (Nokia

and Motorola) and the relative volume of Web

search estimated on the basis of Google Trends

(www.google.com/trends/). The goal is to detect

quickly, within a turbulent market, trends and

changes in the market shares of competing firms.

The authors concluded that there was a strong

correlation between market share and user activity

measured by the volume of queries about brands.

Our conclusion is confirmed by the work of Xu et al.

(2010). However their approach should provide a

higher reactivity. On the other hand, it presupposes

the availability of temporal data for market shares.

6 DISCUSSION AND

PERSPECTIVES

The correlation between market share and media

share appears strong in the automotive sector. It

increases again by sorting brands, distinguishing

between generalist brands and premium brands. A

precise definition of the reference market is

important for the proper functioning of the method.

The results confirm that our approach provides an

opportunity for estimating the market share and the

balance of power between companies based on the

media share.

This preliminary study also confirmed the

limitations associated with the use of the Google

search engine as a tool for webometrics, especially

EstimatetheMarketSharefromtheSearchEngineHitCounts

115

when the API is used. This made the production and

the communication of a study to update and deepen

the potential biases of search engines API within

webometrics studies: see Viseur, 2012a.

The fact that media share of premium brands

seems to be greater than their market share could be

explained by the additional communication about

premium brands, which are able to generate passion

and elicit comments in the official press or

participatory media (forums, blogs, etc.). Albert et

al. (2012) indeed confirm that the brand passion, i.e.

“the strong positive feeling towards a brand”, is

linked to unique and prestigious brands, and in turn

leads to greater positive word-of-mouth.

Several points could be studied further.

Firstly, the study could be deepened for the

automotive sector. The method could be applied to

models of cars, in order to assess if it could also be

used to estimate the market share of products and

services (and not just brands).

Secondly, although this first experiment was

conducted on a mature market, our aim is to obtain a

method applicable to more turbulent markets. The

smartphone market is one of those markets. The

application of the method presents practical

difficulties related to the bias caused by the

accumulation of historical data in commercial search

engines. Several approaches could improve this

method. They include the use of news search

engines, the use of Twitter timeline content, the use

of the Google “daterange” operator or the use of a

custom search engine with a specific index (see e.g.

Viseur, 2012b).

Thirdly, the impact of negative opinions on the

market shares of brands could be further explored.

The impact of negative opinions on market share is

not easy to quantify. See the astonishing example of

the DecorMyEyes online store. Uncles et al. (2010)'s

works provide an initial insight into the effects of the

total volume of word-of-mouth (WOM), the volume

of positive word-of-mouth (PWOM) and the volume

of negative word-of-mouth (NWOM).

REFERENCES

Albert, N., Merunka, D., Valette-Florence, P., 2012, Brand

passion: Antecedents and consequences, Journal of

Business Research, available online 5 January 2012.

Andrieu, O., Duffez O., 2004. Google : Trucs de pros,

Editions Micro Application.

Andrieu, O., 2005. Quand les moteurs de recherche

occupent l'espace média..., Abondance.com (retrieved

April 10, 2013).

Brin, S., Page, L., 1998. The anatomy of a large-scale

hypertextual Web search engine, Seventh International

World-Wide Web Conference (WWW 1998), April 14-

18, 1998, Brisbane, Australia.

Cafferky, M. E., 1995. Let Your Customers Do the

Talking, Upstart Pub Co.

Eagle, L., Kitchen, P. J., Rose, L., 2005. Defending brand

advertising's share of voice: A mature market (s)

perspective, The Journal of Brand Management,

13(1), pp. 65-79.

Emerson, T., Ghosh, R., Smith, E., 2012. CASE STUDY:

Using the Social Share of Voice to Predict Events That

Are about to Happen, Practical Text Mining and

Statistical Analysis for Non-Structured Text Data

Applications, pp. 127-131.

Kotler, P., 1999. Le marketing selon Kotler, Editions

Village Mondial, Paris.

Kotler, P., Dubois, B., 2000. Marketing management,

Publi-union Editions, Paris.

Lambin, J.-J., 1998. Le marketing stratégique, Ediscience

International, Paris.

Mayr, P., Tosques, F., 2005. Google Web APIs: an

instrument for webometrics analyses, Proceedings of

the ISSI conference.

McCown, F., Nelson, M. L., 2007. Agreeing to Disagree:

Search Engines and their Public Interfaces, ACM IEEE

Joint Conference on Digital Libraries (JCDL 2007),

June 17-23, 2007. Vancouver, BC, Canada. p. 309-

318.

Millier, P., 1997, Stratégie marketing de l'innovation

technologique, Dunod.

Porter, M., 1992. L'avantage concurrentiel, InterEditions.

Roustan M., Lehuede, F., Hebel, P., 2005. Qu'est-ce

qu’Internet a changé aux modes d'achat des Français ?,

CREDOC (www.credoc.fr), Cahier de Recherche,

n°213, novembre 2005.

Segal, D., 2010. A Bully Finds a Pulpit on the Web, The

New York Times, 26 novembre 2010. URL:

http://www.nytimes.com/2010/11/28/business/28borke

r.html (retrieved April 10, 2013).

Štrach, P., Everett, A. M., 2005. Globalizing Luxury

Automobiles through Mergers: Three Brands at the

Crossroads», Working Paper, No. 5/2005.

Thelwall, M., Vaughan, L., Björneborn, L., 2005.

Webometrics, Annual Review of Information Science

and Technology, 39, 81-135.

Turney, P., 2002. Thumbs up or thumbs down ? Semantic

orientation applied to unsupervised classification of

reviews, Proceedings of the 40th Annual Meeting of

the Association for Computational Linguistics, p. 417-

424, Philadelphia.

Uncles, M. D., East, R., Lomax, W., 2010, Market share is

correlated with word-of-mouth volume, Australasian

Marketing Journal (AMJ), Volume 18, Issue 3, August

2010, pp. 145-150.

Véronis, J., 2006. Etude comparative de six moteurs de

recherche, Université de Provence, 23 février 2006.

Viseur, R., 2012a. Les moteurs de recherche commerciaux

sont-ils des outils de webométrie fiables ?, Actes du

30ème congrès InforSID, Montpellier (France), 29-31

mai 2012.

DATA2013-2ndInternationalConferenceonDataManagementTechnologiesandApplications

116