A framework for choosing the right
affiliate partner
By
Jan Willem van Dijk
University of Groningen
Faculty of Economics and Business
Summary
Online advertising has grown significantly the last few years. Within online marketing affiliate marketing is an important pillar. Affiliate marketing is a performance based approach to online advertising. Essentially every possible website could become an advertiser’s affiliate partner.
Websites promoting advertisers are being compensated only when they actually generated a sale for the advertiser. More affiliating websites would give advertisers more advertising appearance . However, more websites does actually cost more time and effort to manage and not every affiliating website has a positive effect on the advertiser’s brand image. A framework for an advertiser that would help them select those websites that have positive advertising potential gives advertisers a great advantage. To be able to create such a framework research has been conducted to analyse the relationships between website characteristics and advertising potential. Website characteristics that are measured to analyse their effect on advertising potential are targeting, congruence, familiarity, traffic and ad obtrusiveness. The influence on brand image and advertising effectiveness are used to determine advertising potential.
The results from the analyses show that there are some significant relationships between some website characteristics and either brand image or advertising effectiveness. Website-advertiser congruence has a significant relationship with brand image. Websites that are congruent with the advertiser positively influence brand image. Traffic has a significant relationship with advertising effectiveness. Websites that have high traffic will more likely show higher advertising effectiveness. Ad obtrusiveness has a significant positive relationship with advertising effectiveness. Websites with more and larger advertisements on their website have higher advertising effectiveness. There has not been measured a significant relationship between the website characteristics targeting and familiarity with either brand image or advertising effectiveness.
Table of content
Introduction ... 3 Affiliate marketing ... 4 Hidden costs ... 6 Research question ... 6 Literature ... 9 Brand image ... 9 Advertising effectiveness ... 10 Website characteristics... 11 Website familiarity ... 11 Website traffic ... 12 Website-advertiser congruence ... 14 Targeting ... 15 Advertisement obtrusiveness ... 17 Conceptual model ... 19 Methods ... 20 Advertiser ... 20 Survey/Dataset ... 20 Procedure ... 21 Plan of analysis ... 25 Descriptive ... 25Multiple regression analysis ... 25
Cronbachs Alpha, multicollinearity and residuals ... 26
Introduction
The total spending worldwide on online advertising is growing significantly. In 2012 the total spending on online advertising was $99,2 billion, a growth of 16.2% compared to 2011 spending of $85,3 billion (GroupM.com). In the same year, total media ad spending worldwide only showed a growth of 4.9%. This pattern is foreseen to continue with expected growth in online ad spending to be on average 15% and total media ad spending on average only 4.5% (Strategyanalysts.com). Due to this larger growth rate online advertising will become more and more important.
However, there has not always been a growth within online advertising. After the internet bubble burst, online advertising spending dropped significantly. In 2002 the total amount in the US spend on online advertising was about 25 % less than it was in 2000 (Emarketer.com). This downfall was not only shown by the decreased spending, but the average performance of the online campaigns showed a dramatic downturn as well (Fulgoni, 2009). This decreasing performance was very well shown by a strongly decreased click-through-rate (CTR). CTR is a key performance indicator within online marketing showing advertising effectiveness (Chandon et al, 2003). This downfall of effectiveness and spendings in 2002 made advertisers and publishers analyse the existing advertising models within online advertising. This analysis created a better understanding about them and their possibilities (Papatla et al, 2002).
subscription to a newsletter on the advertiser’s website (Fulgoni, 2009). Therefore, the CPA model is even more performance based than the CPC model. These pricing models increase advertising effectiveness because publishers now become forced to target their website visitors with better fitting advertisements. For example, publishers that create content related to sport will increase their advertising income when they show sport-related banners. Because when sport-related banners are shown on a website that publishes sport-related content, visitors are more likely to click on the banners because there is a match between visitor, content and advertisement. This match also increases the probability that a click results in an action. The increased number of clicks and actions directly result in an increase of the publishers’ income, because they get compensation per click or action (Spilker-Attig, 2010).
Affiliate marketing
sale there is an automatic check if the visitor has an affiliate cookie. In this case the purchasing visitor has an affiliate cookie from Nu.nl. Now Nu.nl will receive a commission from Zalando.
Clicks and sales are tracked by a third party: the affiliate network. The affiliate network has the infrastructure to track the performance of the campaigns for both the advertiser and publisher. The affiliate network is set between the publishers and advertisers as an objective ‘scorekeeper’. Their numbers are leading and binding for both the advertiser and the publisher. There is a third party involved because the tracking infrastructure could not be created by every publisher or advertiser individually. Affiliate networks also provide the standard agreements on for example cookie time and tracking methods. Also an objective third party decreases the possibility on conflicts regarding the tracked numbers (Zanox.com).
For a publisher becoming an affiliate partner is in general a simple process. They can visit an advertiser’s website and check if the advertiser uses affiliate marketing. If so, the publisher can register by going through the terms and conditions and if interested fill in the registration form. Now the advertiser only has to accept them as their affiliate partner (Novak et al, 2000b). Publishers select only those advertisers they want to promote on their website.
With affiliate marketing a publisher only gets a commission if there has actually been established either a sale or lead through a click on banners on their website. This makes affiliate marketing a pure performance-based payment model (Truong, 2010). This characteristic leads to an important difference with traditional, non-performance-based advertising models where costs are made without the guarantee of results (Zeff and Aronson, 1997). Because of this typical characteristic of affiliate marketing advertisers can, in principle, have an endless amount of publishers advertising for them. However, publishers only have a limited amount of advertising space and therefore they should select those advertisers that will provide the highest advertising income.
been successfully advertising for the advertiser as well. Due to these opportunity costs advertisers lack the opportunity to reach niche target groups. This leads to a missed opportunity for advertisers to extract value from the long tail (Anderson, 2006). With affiliate marketing it appears that no opportunity costs are involved. Therefore much more publishers could advertise for the advertiser. This creates a larger advertising potential for the advertiser.
Hidden costs
However, there are over 100 million websites on the net. And although it appears like there are no immediate opportunity costs for the advertiser involved with publishers becoming affiliate partners, there are actually some hidden costs of establishing/having an affiliate partnership with a publisher (Papatla et al, 2002). There are costs involved for the advertiser with the tracking of the affiliate partners; clicks, sales, everything needs to be tracked. The tracking is performed by an affiliate network. The advertiser pays the affiliate network for this service.
Other costs involved for the advertiser are the costs related to misbehaviour by their publishers. Within affiliate marketing there are possibilities for publishers to gain commission illegally. The costs for the advertiser are twofold. The costs involved with controlling their publishers on this behaviour and the costs related to actual misbehaviour by publishers that were not discovered (Tribal.com, 2010).
Next to tracking and controlling costs there are the extra human resources necessary for all affiliate partners. With a larger number of affiliate partners for the advertiser, less time per partner is available. An increase in the number of partners will lead to a decreasing amount of resources available to spend on the key account partners. This might cause long-term non-monetary costs as well (Papatla, 2002). With non-performance-based advertising models there are only a few advertising partners which are carefully selected. With so few partners there is great influence on how they promote the advertiser. With affiliate marketing there is no such careful selection of partners and therefore less influence on how they advertise. This loss of control could have a negative influence on the brand image (McCormick, 2010).
Research question
websites could also do some serious damage in regard to the advertiser’s brand image (Shamdasani et al, 2001).
Since not every aspiring publisher has an expected positive value for an advertiser, it is important for an advertiser to have a clear selection method that can be used to determine which publishers should become their affiliate partners and which publishers should be rejected. If an advertiser would have a framework that could be used to determine the publisher’s expected advertising value, considering both sales and brand image, affiliate marketing will likely be an effective marketing source for them. Without this framework affiliate marketing could have a severe negative influence on brand image and/or lead to a loss of valuable company resources used on publishers without any potential.
So far there has not been performed any research that resulted in such a framework. Therefore the research question is designed to create as a final result such a framework. Which results in the research question: ‘what website characteristics can best be used by an advertiser to predict the potential of a publisher aspiring to become an affiliate partner?’ Especially for affiliate marketing such a framework would be useful because only with affiliate marketing there are so many publishers that could be used as an advertising source. To determine a website’s potential website characteristics could be used to predict the potential affiliating website’s advertising value. Website characteristic’s examples are: size and target group. How these website characteristics influence both the advertisers’ brand image and advertising effectiveness will determine their advertising potential.
So far, research has been performed on several individual website characteristics and their influence on advertising potential. However there has not been conducted a research yet that resulted in a complete framework that can be used to value a website’s advertising potential. Research on the combined influence will create insights on the relative importance of each individual characteristic as well. To determine the advertising potential a website has, both the influence the characteristics have on advertising effectiveness and brand image will be researched. Advertising potential is the total influence on total expected sales for the advertiser. The combined scores of advertising effectiveness and brand image will determine overall advertising potential.
if the aspiring website will at least reach a certain minimum result regarding the important key performance indicators: impressions, click through rate and conversion rate (Chandon et al, 2003). The probability of reaching this minimum will be referred to as advertising effectiveness. A website that reaches a high score on advertising effectiveness is to be expected to generate many sales/leads.
Brand image is the other important variable that determines expected advertising potential. Most customers will prefer to purchase a product from a brand with a positive brand image above a unknown brand or a brand with a negative brand image (Keller, 2003). The expected influence of an aspiring partner on the advertiser’s brand image therefore determines advertising potential as well. A high score on brand image for an aspiring partner has a positive influence on the brand image for the advertiser and therefore enlarges the aspiring partner’s advertising potential. A negative score on brand image will decrease the website’s total advertising potential.
Literature
In 1994 the first banner ads were shown. Since then internet advertising has experienced a massive growth. Ever since internet advertising became popular, research has been conducted to determine the influence of several factors on advertising potential. Research on internet advertising potential has usually been divided in two different forms of measuring potential. One form of research has been mainly conducted on the influence a banner has on brand image, because banner advertisement, in this form, is seen as a way of marketing communication. Research indicates that online advertising has been found to be effectively influencing brand image (Wang et al, 2009).The second form measures the immediate results from the banners, the advertising effectiveness. Click through rate (CTR) is predominantly used to measure results this way (Briggs et al, 1997). Main difference between those two variables is that brand image has a long term effect and advertising effectiveness is more about direct response.
This research will be conducted on both brand image and advertising effectiveness. So, how will several website characteristics influence a long term advertiser’s brand image and how do these characteristics influence short term immediate advertising effectiveness as well. In order to create a complete framework that could be used to determine a website’s potential it is important to place emphasis not only on one of those aspects but on both of them. Short term and long term are both important to determine the advertising potential. To clarify brand image and advertising effectiveness they are now introduced first.
Brand image
Brand image is how consumers perceive the advertiser’s brand. This perception is based on associations consumers have regarding the brand in their memory. Brand associations are the information nodes that link to the brand node in memory and they contain the meaning of the brand for consumers. A brand can be classified in three different categories: attributes, benefits and attitudes. These associations combined form the meaning of the brand for the consumer. The favourability, strength and uniqueness of these associations form the dimensions that distinguish brands from each other. These dimensions determine the brand equity that brand image creates for the advertiser (Keller et al, 2003).
Brand image particularly is of importance in service industries. Here is the parallel with internet companies. The intangibility within services, which makes brand equity so important, is also of big importance within the internet. For instance trust issues are big within the internet. This intangible asset is largely influenced by brand equity. Studies show this too; brand equity is a key determinant of competitive success for internet firms (Trueman et al, 2012).
Brand image protection within online marketing is very important as well because the boundaries to create an online media vehicle are so much lower than creating a traditional media vehicle. Creating a website is way easier than setting up a television channel or a journal (Shamdasani et al, 2001). And since the boundaries to create an online media vehicle are so much lower, there are much more low-quality media vehicles online. Within online advertising there is a greater risk of a bad media vehicle negatively influencing the advertiser’s brand image.
Since brand image is a key determinant for internet firms, it is important to know what the influence is, that websites have on the companies they are advertising for. Especially since there are so many websites that could threaten their brand image. If some website characteristics would likely predict a negative influence on the advertiser’s brand equity, then these websites should not be used to promote the advertisers. And website characteristics that will likely improve the advertiser’s brand image should therefore be used as an advertising source.
Advertising effectiveness
Within affiliate marketing it is not only about driving new traffic towards the advertiser’s website. Affiliate marketing is pure performance based, so a click is not enough yet. The website visitors that reached the advertiser’s website should convert as well. Some website will mainly send public that may have low purchase intentions. Where other websites will have more visitors that are actually planning to buy products (Naylor et al, 2012). For instance, a price-comparison website that sends traffic towards the advertiser will likely send a group that will have an above average conversion rate. Because people that visit a price comparison website are in general within the lasts steps of the buying process. Therefore CR (conversion rate) is an important measurement to determine advertising effectiveness as well. Conversion rate shows the average number of website visitors that actually made a purchase. Comparing conversion rate between the different websites gives a valuable insight in these website’s potential (Lougney et al, 2008).
To test the immediate advertising effectiveness it is important to make use of both CTR and CR to have a complete overview of the website advertising potential. CTR and CR can be easily combined since CR follows upon CTR.
Website characteristics
In order to answer the research question the relevant publisher characteristics should be identified. These characteristics will be used to determine the advertising potential of an aspiring website for the advertiser. In the upcoming section relevant characteristics have been determined and how these characteristics will influence advertising effectiveness and brand image.
Website familiarity
Website familiarity in an E-commerce environment refers to how well a user knows the website and the website procedures (Gefen, 2000). Yoon (2002) and Siau et al (2003) showed with their research that website familiarity largely increases trust. This enlarged trust leads to higher perceived website quality and intention to reuse the site. Website quality and intention to reuse influence both advertising effectiveness and brand image (Mccoy et al, 2008). Therefore website familiarity is an important characteristic to determine a website’s advertising potential. It is important to notice that familiarity in general increases trust, however this is not always the case (Mccoy et al, 2008).
Within online advertising source credibility is an important characteristic as well. Source credibility influences the effectiveness of banner advertising. And because banner advertising is the main form of online advertising source credibility has a major influence on website advertising potential (Choi et al, 1999). The correlation between source and ad effectiveness has been demonstrated through many researches within traditional media (Gotlieb et al, 1991, Goldsmith et al, 2000). Research conducted by Brunner and Kumar (2000) shows a positive correlation between a person’s attitude regarding the website and the effectiveness of the online ads as well. The hosting website’s reputation persuades the consumer to visit the ‘endorsed’ advertising retailer.
Source familiarity influences not only advertising effectiveness, but the advertisers’ brand image as well. Advertising gives companies the possibility to link other brands, people, places etcetera to their brands. These links influence brand image (Keller, 2003). Fuchs stated already in 1964 that the prestige of a magazine ‘rubbed off’ on advertisers. For a website as media vehicle this will likely be the case as well. The media vehicle reputation, in this case the website, therefore has an influence on the advertiser’s brand image. A website with a good reputation will ‘rubb off’ positively, resulting in a higher brand image for the advertiser. Therefore source familiarity influences brand image.
Within the online environment established and respected websites can serve as a pseudo endorser for the advertising retailer. Through this endorsement the retailers reputation is influenced by the advertising website. Familiarity with the website increases the impact the advertising has on the brand image. Advertising on a non-familiar website will have less impact than advertising on a familiar website has. In general this familiarity will have a positive effect on brand image (Romaniuk et al, 2012).
Hypothesis 1: Website familiarity leads to higher advertising effectiveness.
Hypothesis 2: Website familiarity has a positive effect on the advertiser’s brand image.
Website traffic
A second website characteristic that influences advertising potential is the amount of traffic an affiliating website has. Traffic correlates strongly with website familiarity. Most well-known websites will have a high amount of traffic as well. This however is not always the case. For example, some websites have high familiarity because of their offline presence. Elsevier.nl is a website that is has high familiarity because of their offline magazine, but they only have a limited number of visitors on their website and therefore their result on website traffic would be low.
is counted as an additional page view. If a user navigates to a different page and then returns to the original page, a second page view is recorded as well (Google.com). Traffic is a characteristic used to determine expected advertising potential. Within traditional media magazine advertisers often use the number of subscriptions to determine the expected value of advertising in the specific magazine. With the traditional online pricing method, Cost Per Mille, pricing is based on the number of page views, so based on the amount of traffic as well. More page views means a higher price. This indicates the tie between the number of page views and the advertising value.
More page views can indicate a website either has more unique visitors or the average visitor surfs to more different pages on the website. If a website for instance has 1000 page views this could either be a 1000 unique visitors visiting only one page on the website or one visitor generating a 1000 page views by looking into many different pages on the website. Therefore, the amount of traffic indicates the combination of reach and the possibility for ad repetition. This wider reach and ad repetition influence both advertising effectiveness and brand image (Sumner, 2001).
Product awareness already increases significantly after only one exposure to an advertisement. More reach within the traffic therefore leads to more brand awareness (Briggs and Hollis, 1997). Higher brand awareness usually positively influences trust. And this enlarged trust has a positive effect on brand image. When a website has many page views, which indicates a wide reach, then this wide reach will help the retailer to increase brand awareness. Therefore is brand image positively correlated with website traffic.
However, a website with many pageviews could indicate that a visitor visits many different pages on this website as well. In this case, many pageviews indicate the possibility for high ad repetition. Research by Chatterjee et al (2003) showed that this ad repetition leads to building brand awareness. Even though unaware of the presence of the advertising banner, visitors become aware of the advertised brand. This leads to a higher expected reaction in the future on this brand. Although, when the advertising banner is shown too often this might have a negative effect on perceived brand image (Mccoy, 2009). A visitor might get the feeling it is spammed with an advertiser.
shows a positive effect as well, the likeliness of an enormous decrease of CTR is low (Laroche et al, 2006). A website with a large reach will influence advertising effectiveness positively as well. This large reach means there are many different website visitors. Therefore this website has many potential customers that could reach the retailer through the affiliating website.
A high amount of page views indicates a wide reach, high average website usage (many page views per Unique Visitor) or a combination of both. This will have a positive influence, as described above, on advertising effectiveness. Therefore the following hypothesis:
Hypothesis 3: More traffic leads to higher advertising effectiveness.
Website-advertiser congruence
The third website characteristic to determine website advertising potential is website-advertiser congruence. Congruency between website and advertiser exists when there is a match between the products that the website is advertising for and the website’s visitors. For instance: congruence exists when a website about shoes is advertising for a shoe company, because a shoe website will probably attract visitors interested in shoes (Shamdasani et al, 2001). Website-advertiser congruence could be a content match, but a target-group match as well. For instance for an advertiser selling women shoes; fashionista.nl, a fashion website, has a match because of content. Libelle.nl, a website that specifically targets women, is a relevant website as well because their target group is women. Research has been conducted that shows the influence a website has as a vehicle on banner effectiveness (Choi and Rifon, 2002).
Within traditional media the effect of the match between content and advertisement has been researched. And the research shows that media content has a significant effect on advertising effectiveness. When the same advertisement is shown to an equal public but within different context the ad generates a different effect (Chaiken and Stangor, 1987). When the content of the ad fits well with the context the public is more likely to accept the message. And the people who are interested in certain content are more likely to be interested into those content related advertisements. A person looking in a fashion magazine will be more likely interested in a shoe ad than a person reading a football magazine (Aaker and Brown, 1972).
negative impact on advertising effectiveness in combination with congruency is obtrusiveness. When the ad’s obtrusiveness increases, the match between website advertising and content might have a negative effect on advertising effectiveness probably due to privacy issues (Goldfarb and Tucker, 2011). But, since this effect only appears in the case of highly obtrusive ads, the majority of time a match between content and advertising has a positive effect on advertising effectiveness.
Website-advertiser congruence is both a match between the target group of the website and the advertisement and a match between the website content and the advertisement visitor as well. Kusumoto (2002) says that ‘If brand user’s sense of value and the magazine readers’ sense of values are similar, the readers will read the brand advertisements with a positive attitude, resulting in high probability brand message absorption’. Therefore this match between target-group and advertisement influences brand image as well. Research (Cho and Kim, 2012) shows that congruence correlates positively with the advertiser’s brand image. When website-advertiser congruence is large enough, advertising can even be a service towards website visitors. Through specific advertising these websites can make consumers aware of products that are highly relevant to them (Rappaport, 2007). This positively influences brand image as well.
If there is website-advertiser congruence, involvement grows according to Malthouse (2007). More involvement enlarges the effect on brand image. This however means that not only positive but negative associations have a higher impact as well. So if a website’s content is congruent but the consumer has a negative association towards the website, this website congruence will have a negative effect on brand image. Congruence will influence brand image negatively only if there are negative associations with website. These negative associations will probably not appear often enough to diminish the overall positive effect congruence has.
Hypothesis 4: website-advertiser congruence has a positive effect on advertising effectiveness.
Hypothesis 5: website-advertiser congruence has a positive effect on the advertiser’s brand image.
Targeting
Potential customers have different preferences and interests. These customers can be segmented in different groups of interest. Market segmentation according to Kotler (2005) is: ‘Market segmentations is the subdividing of a market into homogenous subsets of customers.’ If the market gets divided more these groups can be targeted very specific and will likely have a higher response if approached with their interests. Research has shown that banner advertising effectiveness is positively influenced when targeting is involved. The most effective targeting forms used are behavioural targeting and geographic location targeting (Gallagher et al, 1997).
Expected advertising effectiveness will increase when the affiliating website has a specific target group, compared to an affiliating website without a specific target group. However, research shows that targeting only increases advertising effectiveness when there is a match between the target group and the advertised product. Congruence therefore has a moderating effect on targeting. A website with a specific target group most likely has relative high advertising effectiveness, however only, when their audience fits with the advertiser. There will likely be higher advertising effectiveness because more visitors will click on the banners because the banners’ content matches with the target group’s interests. When there is no fit between the audience and the advertiser, the fact that this audience is a specific target group will not positively influence the expected advertising effectiveness. Targeting will improve advertising effectiveness if there is website content and advertiser congruence. Without this congruence targeting will decrease expected advertising effectiveness (Bergemann and Bonatti, 2011).
For instance geographic targeting: a website, NU.nl, that targets Dutch people and an advertiser focussing on the Dutch market have congruent geographic target groups. Advertising effectiveness will be higher than if the website would not have been specifically targeting the Dutch and instead would have had a worldwide target group, for instance CNN.com. Here targeting increases the effectiveness of advertising. When there is website content advertiser congruence, targeting increases the advertising effectiveness. However targeting decreases advertising effectiveness when there is no website-advertiser congruence. If the website would target Dutch people and the advertiser targets French people, targeting decreases advertising effectiveness.
to advertisers that use media that does not target a specific target group, another brand image is created. Therefore a potential affiliate website that has a high targeting degree might have a higher potential for the advertiser. However only when there is congruence between advertiser and the potential affiliate’s website; advertiser-website congruence is necessary to increase advertising potential when there is a high targeting degree. When there is no advertiser-website congruence there probably will not be a positive influence on the advertiser’s brand image. Congruence has therefore a moderating effect.
So, when there is congruence, targeting will probably influence brand image and advertising effectiveness positively. Without congruence a higher degree of targeting will have a negative impact on both brand image as advertising effectiveness.
Hypothesis 6a: Targeting increases advertising effectiveness if there is advertiser-website congruence.
Hypothesis 6b: Targeting decreases advertising effectiveness if there is no advertiser-website congruence.
Hypothesis 7a: Targeting has a positive effect on brand image if there is advertiser-website congruence.
Hypothesis 7b: Targeting has a negative effect on brand image if there is no advertiser-website congruence.
Advertisement obtrusiveness
The 5th and last website characteristic that influences advertising potential is ad obtrusiveness. Ad obtrusiveness is the degree that the ad is visible and stands out in comparison to the website content (Goldfarb and Tucker, 2011). This visibility can for example be influenced by banner size and media effects (animation effects). Ad obtrusiveness influences both advertising effectiveness and brand image.
obtrusiveness has on direct actions, advertising effectiveness is less significant. However Spalding et al (2009) show a positive correlation between ad obtrusiveness and advertising effectiveness as well. A higher degree of obtrusiveness influences brand image as well, as was pointed out above (Spalding et al, 2009). Increasing ad obtrusiveness influences brand image even more than advertising effectiveness according to this research. Increasing the ad obtrusiveness has a major impact on brand image. However, research by Goldfarb et al (2011) shows that increasing ad obtrusiveness can influence brand image negatively as well. Especially when there is high ad obtrusiveness and website-advertiser congruence, website visitors can feel privacy issues for customers can occur. Within the research by Goldfarb et al (2001) pop-ups were included. But, within affiliate marketing pop-ups are in general forbidden within the affiliate programs and therefore these results from Goldfarb et al (2001) are less relevant. Other research shows that it is important for advertisers to take care that they do not annoy, irritate or aggravate potential customers with their advertisements (Lodish, 2011). Higher ad obtrusiveness increases the possibility to annoy, irritate or aggravate customers. Lee et al (2010) found that within online marketing ads are already more intrusive than within traditional marketing. And this intrusiveness leads customers to not intend to return to the site. When on a website advertisements become more obtrusive this could lead to a negative reaction which influences brand image (Goldfarb et al, 2001). This however, will only be the case when the ad obtrusiveness is too high. In general we expect ad obtrusiveness to have a positive impact on the brand image. For advertising effectiveness we expect ad obtrusiveness to have a positive impact as well.
Hypothesis 8:Ad obtrusiveness increases advertising effectiveness
Congruence Targeting Familiarity Obtrusiveness Traffic Congruence Targeting Familiarity Obtrusiveness
Conceptual model
These hypotheses combined lead to the following model that can be used to predict the advertising potential an aspiring affiliate website has.
Model 1: conceptual model
Methods
In this chapter the research design used to answer the research question will be outlined. According to Malhotra et al (2007) three sorts of research designs can be used. Exploratory research which can give insights in a not clearly defined problem. Descriptive research that can be used to describe market characteristics. And causal research that has as the main objective to determine cause and effect relationships. Within this research we will make use of the causal research design since this research is about the effect website characteristics have on advertising potential.
To discover how the selected website characteristics influence advertising potential research will be performed on both their influence on advertising effectiveness and brand image. For researching advertising effectiveness an existing database is used with actual numbers from the affiliate program from a fashion web shop. The data used for researching brand image comes from a survey. The data from the survey and the existing data from the web shop will both be used for quantitative research. The collected numerical data will be used to analyse the relationship between the website characteristics and the website advertising potential using a statistical method (Aliaga & Gunderson, 2002).
Advertiser
Within this research the fashion web shop Zalando was used as the topic of research. Zalando has been active within the Dutch market since 2010. Since then, Zalando has become the biggest online fashion web shop on the Dutch market. Their brand has been strongly advertised both online and on television. Therefore, Zalando is now a well-known brand within The Netherlands which makes Zalando a useful brand to investigate brand image. Ever since Zalando became active within the Dutch market, they have had an affiliate program. This program is exclusively accessible through the affiliate network Zanox. Right now the Zalando affiliate program is one of the largest affiliate programs within the Dutch market. Using the data from the Zalando affiliate program will therefore be useful for researching advertising effectiveness. The Zalando data used comes from the Zanox system.
Survey/Dataset
gather respondents (Malhotra & Birks, 2007). The survey has been distributed through the usage of direct mailing and social media (Facebook). The target population was men and women above 16 living in The Netherlands. In total, 117 respondents filled in the survey completely.
The data used for the research on advertising effectiveness was extracted from the affiliate network Zanox. They exclusively run the Zalando affiliate program in The Netherlands. Therefore they track all clicks and sales from all affiliate partners Zalando has in The Netherlands. For this research we used their data from February 2013. The data from the Zanox system only gives an overview on affiliate partners that generated at least one sale. Therefore only affiliate partners that generated at least one sale are within the data. To use only those partners that have some substantial number of clicks, only the affiliate partners with over 200 clicks were included within the data (Haans et al, 2013). Therefore, every Zalando affiliate partner with at least 1 sale and at least 200 clicks within this period was included within the data. A total of 200 affiliate partners met this requirement in February 2013.
Procedure
The survey, used to research the influence the website characteristics have on brand image, contained 15 screenshots and after every screenshot there were 7 questions. These questions were equal for every screenshot. In addition, questions about age, gender, online experience and education were added. These additional questions are used to describe the respondents. The target group for this research were men and women above 16 from The Netherlands, therefore the survey was held in Dutch. The screenshots showed websites with Zalando advertisements. The advertisements at the screenshots have different sizes and positions for every screenshot. The website characteristic traffic was not included because it is not possible to test the influence traffic has on brand image with a survey. Every website that was used for a screenshot was valued by the researcher on the website characteristics being researched. This value could go up from one until five.
Congruence: a higher level of congruence between the website’s subject and target group with the Zalando’s target group results in a high score on congruence. A website about football like www.vi.nl that has not a lot of congruence with fashion and a totally different target group as Zalando therefore receives a low value on congruence. A blogging website about fashion like com has the same target group as Zalando and is about fashion and is therefore highly congruent with Zalando. This website would have been valued highly on congruence.
has such a specific target group the website would receive a high value on targeting. A website as www.nu.nl which is a news site and has a very wide target group would therefore be valued low on targeting.
Obtrusiveness: the value on obtrusiveness depends on the size and number of Zalando advertisements on the website. Within the survey the screenshots from the websites have different number of advertisements and the advertisements are sized differently. Within the survey screenshots only show Zalando advertisements. The higher the combined appearance of the Zalando advertisements on the website is, the higher the website is valued on obtrusiveness. A website with less advertisements/smaller advertisements therefore receives a low score on obtrusiveness.
Familiarity: how well known the website is determines the value on familiarity. A website like www.telegraaf.nl is very well known and therefore receives a high value on familiarity. A website like www.fashiolista.com is less known and therefore receives a lower value on familiarity.
For instance the website Quest.nl received values for congruence (2) targeting (3) obtrusiveness (3) and familiarity (4). Quest is a website that has low congruence with Zalando, but the website is quite familiar. Combined the 15 screenshots have all kinds of different combinations on how they are valued on the website characteristics.
Figure 1: survey
The existing data used for this research from the Zanox network is used to research the relationship between the website characteristics and advertising effectiveness. The data from Zanox shows the clicks and sales for every affiliate partner from Zalando in February 2013. The measured number of sales per affiliate partner is used to determine advertising effectiveness. More sales naturally indicate higher advertising effectiveness. The number of sales is the best indicator for advertising effectiveness within affiliate marketing because the compensation for the affiliate partners is based on the number of sales as well. Therefore the number of sales is more effective in measuring effectiveness than the number of clicks because an affiliate partner could generate a large number of clicks but when this does not result in sales this affiliate partner does not have high advertising effectiveness (Chatterjee et al, 2003).
different. The websites within the Zanox data set do not only have Zalando advertisements on their websites like the websites used within the survey. The websites within the Zanox data set do have other advertisements on their websites as well. To value obtrusiveness these other advertisements are included to determine a website’s total obtrusiveness. Bigger advertisements and a larger number of advertisements on a website leads to a higher valuation on obtrusiveness. To value the websites within the database on traffic, the expected number of page views a website will likely generate is used. A website like www.startpagina.nl generates millions of page views every month and would therefore be valued high on traffic. A website like haarkapsels.org generates only a few thousand page views and will therefore be valued low on traffic. The complete valuation overview can be seen in appendix 2. Figure 2 shows an example from websites being valued.
For instance the website www.speurders.nl received the following scores: congruence (2), targeting (2), traffic (4), familiarity (5) and obtrusiveness (3). Speurders is a well-known site with a lot of traffic, therefore they receive high values on traffic and familiarity. But speurders.nl has a very wide target group and has little to do with fashion. Therefore speurders.nl is valued low on targeting and congruence. The amount of advertising on speurders.nl is about average which results in an average score on obtrusiveness.
Figure 2: website characteristic valuation
Website congruence targeting traffic familiarity obtrusiveness Aantal sales
Plan of analysis
Descriptive
The results from the survey and the existing data will be analyzed. Statistical analysis will be performed using the statistical program SPSS 17.0. The first step within the data analysis is to describe the results of the questions within the survey used to get information on the respondents. Their age, education, gender and online experience will be shortly described. After the description of the survey data an overview will be given on the Zanox data. A short description of the important facts from the Zanox data will be presented.
Multiple regression analysis
After the description part the relationship between the dependent variables (brand image and advertising effectiveness) and the independent variables (website characteristics) will be analyzed using the data from the survey and the Zanox dataset. In order to find the relationship between the dependent variables and the independent variables regression analyses will be performed on the data. Regression analyses are used because it is: ‘a powerful and flexible procedure for analyzing associative relationships between a metric-dependent and one or more independent variables’ (Malhotra & Birks, 2007). Because the relationship between the website characteristics and brand image and advertising effectiveness will be tested, a multiple regression analysis will be performed. The outcome from these regression analyses will determine the hypotheses’ outcome. With the hypotheses concerning targeting, a moderating effect from the variable website-advertiser congruence is expected. This relationship will be tested with a multiple regression analysis as well. The combined factor, centered targeting * centered congruence, will be added within the analysis to check if congruence has a moderating effect on targeting. In order to deal with multicollinearity the combined factor is built out of centered scores (Malhotra & Birks, 2007). This analysis will be performed on both survey data and the Zanox data. The following formulas are used within the analyses.
Brand image multiple linear regression analysis
Advertising effectiveness multiple linear regression analysis
Brand image: targeting website-advertiser congruence moderation
( ) Advertising effectiveness: targeting website-advertiser congruence moderation
( ) Explanations of the abbreviations in the formulas
BI = Brand Image AE = Advertising Effectiveness WB = Website Familiarity WC = Website-Advertiser Congruence TR = Traffic TA = Targeting AO = Advertising obtrusiveness
Several statistics from the analyses will be used. Adjusted R square will indicate the strength of the overall model as a predictor for the dependent variable. Beta will be used to determine the relationship effects. Sig. will be tested to determine the significance of these relationship effects. Cronbachs Alpha, multicollinearity and residuals
To determine the reliability from the questions used to determine brand image within the survey Cronbach’s Alpha will be tested. A Cronbach’s Alpha above 0.7 will indicate sufficient internal consistency.
For both the research on brand image and advertising effectiveness a test will be performed to determine if there are outliers which impact might be too large on the model’s predicting value. Depending on their impact, the outliers might be removed from the data to increase the model’s reliability (Malhotra & Birks, 2007).
Results
This chapter will give an overview of the results from the statistical tests described above. First an overview on the respondents is given using the descriptive data. After the descriptive results an overview on the test results on brand image will be shown. Then the results on the research on advertising effectiveness will be given. The last section will give a total overview on the results.
Descriptive results
First an overview will be provided concerning the respondents from the survey on brand image. This overview will provide information on the respondents demographics and on their experience with online shopping as well. This is followed by an overview concerning the affiliate websites from the actual Zanox data.
The total group of respondents from the survey consists out of 117 persons. 164 respondents started with the survey. Since filling in the survey took quite some time, it was expected some people would not complete the survey. From the group who completed the questionnaire 58 percent is male and 42 percent is female.
The average respondent was just above 31 years old. Most respondents were within the group 19-34. Which is likely explained due to using convenience sampling as the researcher is within this age group as well. Figure 3 shows an overview on age distribution.
Figure 3: age distribution
Figure 4: education distribution
Other questions asked were related to the respondents’ online experience and, more specifically, their experience with Zalando. Most respondents do have experience with online shopping. Of the respondents, 85 per cent has bought at least once something online. From the respondents, 38 per cent has even bought something at the Zalando shop. And every respondent knew the brand Zalando.
A total of 200 websites were included within the Zanox dataset. All these websites did have at least 200 clicks and one sale. The data was collected in February 2013. The total amount of sales these websites combined generated was 28.357. The average number therefore is 142 sales per website. The website that provided most sales was www.actiepagina.nl. They generated in total 6053 sales. A total number of six websites generated only one sale.
Analyses
This section will give insights in the results from the multiple linear regression analyses that were performed on both the data from the survey and the Zanox data. In this first paragraph we will look at the results from the analysis on the data from the survey.
Brand image
Cronbachs Alpha was calculated using the input from the survey. The total number of responses on every question relating brand image = the number of respondents * the number of screenshots = 117 * 15 = 1755. The total number of items within this test is seven because with every screenshot seven questions were asked regarding brand image. The outcome from this test shows Cronbachs Alpha is 0.918 which indicates that there is high internal consistency between the questions.
A multiple linear regression analysis was performed on the data from the survey. The relationship between the independent variables (targeting, familiarity, congruence and obtrusiveness) and the dependent variable (brand image) was analysed using the multiple linear regression model. In table 1 a summary of the results from the linear regression analysis is presented. Adjusted R² has a score of 0.661. A score on adjusted R² above 0.5 suggests a strong fit between the independent variables and the dependent variable (Malthotra & Birks, 2007). Therefore, with a score of 0.661, we can conclude that the independent variables within the model can give a reasonable well explanation on the brand image value.
Dependent variable R R² Adjusted R² Brand image 0.871 0.758 0.661 Table 1: multiple linear regression analysis: targeting, familiarity, congruence and obtrusiveness
Table 2 shows the regression coefficients. With the usage of these statistics we will now answer the hypotheses on brand image.
Independent variable Standardized beta Significance
Targeting -0.132 0.598
Familiarity 0.141 0.565
Congruence 0.985 0.000
Obtrusiveness 0.163 0.395 Table 2: multiple linear regression analysis brand image
Hypothesis 2: Website familiarity has a positive effect on the advertiser’s brand image.
The regression analysis shows a standardized beta of 0.141 with a statistical significance of 0.565. Based on the analysis using the data from the survey there is no significant relationship between the independent variable familiarity and the dependent variable brand image (with p=>.05). Therefore the hypothesis that website familiarity is positively influencing brand image should be rejected.
The regression analysis shows a standardized beta of 0.985 with a statistical significance of 0.000. Based on the analysis using the data from the survey there is a significant relationship between the independent variable website-congruence and the dependent variable brand image (with p=<.05). Therefore the hypothesis that website-advertiser congruence is positively influencing brand image should be accepted.
Hypothesis 9: Ad obtrusiveness positively influences brand image.
The linear regression analysis shows a standardized beta of 0.163 with a statistical significance of 0.395 Based on the analysis using the data from the survey there is no significant relationship between the independent variable obtrusiveness and the dependent variable brand image (with p=>.05). Therefore the hypothesis that obtrusiveness is positively influencing brand image should be rejected.
Hypothesis 7a: Targeting has a positive effect on brand image if there is advertiser-website congruence.
Hypothesis 7b: Targeting has a negative effect on brand image if there is no advertiser-website congruence.
The linear regression analysis shows a standardized beta of -0.132 with a statistical significance of 0.598. Based on the analysis using the data from the survey there is no significant relationship between the independent variable targeting and the dependent variable brand image (with p=>.05). However, according to the hypotheses congruence has a moderating effect on targeting. In table 3 a summary of the results from the linear regression analysis is presented. The centered variables are used because otherwise the VIF scores would be too high. Adjusted R² has a score of 0.626. Adjusted R² has not improved compared to the analysis performed without the added variable targeting*congruence.
Dependent variable R R² Adjusted R² Brand image 0.871 0.759 0.626
The results from the analysis can be seen in table 4 and show that congruence has no significant moderating effect on targeting.
Independent variable Standardized beta Significance Congruence Centered .955 .003
Targeting Centered -.111 .691
Congruence_Targeting_Centered -0.51 .831
Obtrusiveness Centered .169 .408
Familiarity Centered .168 .560 Table 4: multiple linear regression analysis brand image
An analysis to check on outliers was performed as well. The results from this analysis show that there are no observations that have a score of more than 3 standard deviations away from the predicted score (Malhotra & Birks, 2007).
An analysis on multicollinearity was performed as well. The results are presented in table 5.
Independent variable VIF
Targeting 2.427
Familiarity 2.307
Congruence 1.345
Obtrusiveness 1.388
Table 5: multicollinearity multiple linear regression analysis brand image
No independent variable has a score above 10 and therefore we can conclude that the independent variables will not likely measure the same relationship.
Advertising effectiveness
Table 6 shows a summary of the results from the linear regression analysis performed on the data without the outliers. Adjusted R² has a score of 0.060. A score on adjusted R² beneath 0.1 suggests almost no fit between the independent variables together and the dependent variable (Malhotra & Birks, 2007). Therefore, with a score of 0.060, we can conclude that the independent variables together do not work well as a predictor on advertising effectiveness .
Dependent variable R R² Adjusted R² Advertising
effectiveness
0.289 0.084 0.060
Table 6: multiple linear regression analysis: targeting, familiarity, targeting, congruence and obtrusiveness Table 7 now shows the regression coefficients. With the usage of these statistics we will now answer the hypotheses on advertising effectiveness.
Independent variable Standardized beta Significance
Targeting 0.053 0.523
Familiarity 0.009 0.941
Congruence -0.096 0.210
Obtrusiveness 0.155 0.033
Traffic 0.218 0.098
Table 7: multiple linear regression analysis dependent variable: advertising effectiveness
Hypothesis 1: Website familiarity leads to higher advertising effectiveness.
The regression analysis shows a standardized beta of 0.009 with a statistical significance of 0.941. Based on the analysis using the data from the Zanox data there is no significant relationship between the independent variable familiarity and the dependent variable advertising effectiveness (with p=>.05). Therefore the hypothesis that website familiarity is positively influencing advertising effectiveness should be rejected.
Hypothesis 4: website-advertiser congruence has a positive effect on advertising effectiveness.
Hypothesis 3: More traffic leads to higher advertising effectiveness.
The linear regression analysis shows a standardized beta of 0.218 with a statistical significance of 0.098. Based on the analysis using the Zanox data there is a significant relationship between the independent variable traffic and the dependent variable advertising effectiveness (with p=>.05). The relationship is significant because the performed reression analysis test is two-tailed, both positive and negative regression is tested. However, according to the hypothesis the relationship between traffic and advertising effectiveness should only tested one-tailed, a positive relationship. Therefore, p=>.05 two-tailed becomes p=>.1 one-tailed. Therefore the hypothesis that traffic is positively influencing advertising effectiveness should be accepted.
Hypothesis 8: Ad obtrusiveness positively influences advertising effectiveness.
The linear regression analysis shows a standardized beta of 0.155 with a statistical significance of 0.033. Based on the analysis using the Zanox data there is a significant relationship between the independent variable obtrusiveness and the dependent variable advertising effectiveness (with p=>.005). Therefore the hypothesis that obtrusiveness is positively influencing advertising effectiveness should be accepted.
Hypothesis 6a: Targeting increases advertising effectiveness if there is advertiser-website congruence.
Hypothesis 6b: Targeting decreases advertising effectiveness if there is no advertiser-website congruence.
The linear regression analysis shows a standardized beta of 0.053 with a statistical significance of 0.523. Based on the analysis using the Zanox data there is no significant relationship between the independent variable targeting and the dependent variable advertising effectiveness (with p=>.005). However, according to the hypotheses congruence has a moderating effect on targeting. In table 8 a summary of the results from the linear regression analysis is presented. The centered variables are used because otherwise the VIF scores would be too high. Adjusted R² has a score of 0.055. Adjusted R² has not improved compared to the analysis performed without the added variable targeting*congruence.
Dependent variable R R² Adjusted R² Brand image 0.290 0.084 0.055
The results from the analysis can be seen in table 9 and show that congruence has no significant moderating effect on targeting.
Independent variable Standardized beta Significance Congruence Centered -0.093 .227 Targeting Centered .044 .627 Congruence_Targeting_Centered -0.20 .801 Familiarity Centered .007 .956 Obtrusiveness Centered .155 .033 Traffic Centered -0.20 .801 Table 9: multiple linear regression analysis advertising effectiveness
The results from the analysis on multicollinearity are presented in table 10. Because there is no independent variable with a score above 10, it is not likely that the same relationships are measured.
Independent variable VIF
Congruence 1.212
Targeting 1.411
Traffic 3.575
Familiarity 3.192
Obtrusiveness 1.085 Table 10: multicollinearity advertising effectiveness
Hypotheses
Table 11 shows a summary of the results concerning the hypotheses
Hypothesis Accepted/rejected
Hypothesis 1: Website familiarity leads to higher advertising effectiveness.
Rejected
Hypothesis 2: Website familiarity has a positive effect on the advertiser’s brand image.
Rejected
Hypothesis 3: More traffic leads to higher advertising effectiveness. Accepted
Hypothesis 4: website-advertiser congruence has a positive effect on advertising effectiveness.
Rejected
Hypothesis 5: website-advertiser congruence has a positive effect on the advertiser’s brand image.
Hypothesis 6a: Targeting increases advertising effectiveness if there is advertiser-website congruence.
Rejected
Hypothesis 6b: Targeting decreases advertising effectiveness if there is no advertiser-website congruence.
Rejected
Hypothesis 7a: Targeting has a positive effect on brand image if there is advertiser-website congruence.
Rejected
Hypothesis 7b: Targeting has a negative effect on brand image if there is no advertiser-website congruence.
Rejected
Hypothesis 8:Ad obtrusiveness increases advertising effectiveness. Accepted
Hypothesis 9: Ad obtrusiveness positively influences brand image. Rejected
Conclusion
In this final chapter the results from the analyses are being discussed. Further, implications on the most important findings are explained. Limitations this research was confined to are given. And finally, directions for future research are included to improve comprehension on the subject.
Discussion
Online advertising has become a well-studied topic. The combination of affiliate marketing and online advertising is less studied yet. This research’s main objective was to examine the influence website characteristics have on a website’s advertising potential in order to create a framework that could be used to predict an aspiring affiliate website’s advertising potential. Determining advertising potential is essential for an advertiser with an affiliate program. Too many or wrong websites within an advertiser’s affiliate program can have a negative influence on the advertiser’s brand image or on the effectiveness of the affiliate program. Creating a framework that could help to determine a website’s advertising potential would be very interesting for advertisers.
This study used data that was extracted from a survey with 117 respondents and actual data from the affiliate program from an online fashion web shop, Zalando. Within the survey, respondents were shown several Zalando advertisements on different websites. Following the screenshot, questions relating to brand image were asked. The actual data from the Zalando affiliate program contained websites and the number of sales they generated for Zalando. The websites shown within the survey and the websites that were part of the data from the Zalando affiliate program were valued on their website characteristics. The combination of these scores and the outcome on brand image and the number of sales generated was used to analyze the relationship between these characteristics and brand image/advertising effectiveness.
nature of websites within the Zanox data. The majority of high congruent websites within the database consists out of non-professional fashion blogs/websites. There are many very professional websites within the database as well. These websites in general are not very congruent with Zalando. Above average performance from these non-congruent professional websites might be a reason that congruent, less professional websites had less scores on advertising effectiveness.
The website characteristic targeting has, according to the results, no significant relationships with both brand image and advertising effectiveness. Congruence was predicted to have a significant moderating influence on targeting. The absence of a moderating effect from congruence on targeting contradicts with the findings of Goldfarb and Tucker (2011). Their findings indicate a positive relationship between targeting and brand image. According to their findings, targeting would increase advertising effectiveness as well. Within their research targeting was purely focused on a fit between advertiser and website. Within our research targeting reflects more the degree of website target group specificity which might account for the absence of a relationship within this research. The analysis performed within this research shows a significant positive relationship between traffic and advertising effectiveness. More traffic generally results in more page views. Ad repetition could however decrease average click through rate which would indicate a negative influence from traffic (Robinson et al, 2007). But within this research the ad repetition coming from the higher amount of traffic has increased the number of actions derived from the advertising website. This is corresponding with the research from Yaveroglu et al (2008). Their findings indicate that higher ad repetition results in a higher expected reaction.
Ad obtrusiveness did show a significant positive relationship with advertising effectiveness. According to the research by Goldfarb and Tucker (2011) ad obtrusiveness should increase advertising effectiveness. However, within the research performed by Goldfarb and Tucker (2011) pop-up advertisements were part of the research as well. Pop-ups are much more obtrusive
