20-6-2016
Optimizing media strategies
How should an omni-channel strategy company with a given advertising
budget allocate it’s budget in such a way that it is most effective?
Derkjan Olivier
UNIVERSITY OF GRONINGEN
FACULTY OF ECONOMICS AND BUSINESS MSC. MARKETING INTELLIGENCE MASTER THESIS
1
Abstract
A lot of new advertising channels exist nowadays due to the online world. This study investigates which advertising channels are most effective for an onmi-channel company by looking at the contacts customers have with several advertising mediums. When investigating only the direct effect of all advertising mediums to actual purchase behaviour, it appears that folder advertising has the most influence. Further, the results of this study indicate that synergy effects of a combination folder advertising and the lagged effect of television performs slightly better than a combination of print and radio advertising. Finally, this study indicates that the demographic factors age and income are of great importance when selecting the most effective advertising medium.
1. Introduction
Retailing has changed very rapidly the last two decades due to online retailing and the digitalizing world nowadays. According to an OECD report of 2007, approximately 85 percent of the internet users in de EU take the internet as a medium to search information about products and services. Actually 45 percent purchase goods and services online (OECD, 2007). Business to customer (B2C) situations made use of this by applying an approach what is called, multichannel marketing. Neslin et al (2006) define multi-channel customer management as the design, deployment, coordination, and evaluation of channels to enhance customer value through effective customer acquisition, retention and development. This means that companies could be able to serve customers with more channels than only using brick-and-mortar stores. However, with the rise of the digital marketing and online retailing we shift to a new phase where the online world is even more of importance (Leeflang et al. 2014). Customers use mobile phones, tablets, social media to search for information about products. Research by Rigby 2011 suggest that there is a shift from a multichannel perspective to an omni-channel perspective, where omni-omni-channels consist of more omni-channels (e.g. mobile apps (Xu et al. 2014)) compared to a multi-channel strategy. Within this strategy, customers want to buy products online and are gaining some attention in finding information about products, services, etc. in stores (Verhoef, Neslin, and Vroomen 2007). This phenomena is called showrooming, which means that customers go to the stores to gain information and search online at their online devices for the best prizes. Another phenomena, called webrooming, is actually the other way around. Customers go online to find the information they need and finally purchase the products in stores. In addition, both methods could strengthen each other what is called cross-channel synergy (Verhoef, Neslin and Voormen, 2007).
2 To give a well-grounded answer to the question “How should an omni-channel strategy company with a given advertising budget allocate it’s budget in a way that it is most effective?” a few steps will be taken. First of all, this study will discuss the contribution and relevance of the subject. After that, the conceptual framework will be explained and the hypotheses will be derived. Next, the research mechanisms will be discussed and the hypotheses will be tested in which the following questions will be discussed:
How does traditional advertising affect actual purchase behaviour? How does internet advertising affect actual purchase behaviour ? Do synergy effects exist between the advertising mediums?
How does age affect the relationship between advertising mediums and actual purchase behaviour ?
How does income affect the relationship between advertising mediums and actual purchase behaviour ?
How does educational level affect the relationship between advertising mediums and actual purchase behaviour ?
3
2. Contribution and relevance
This study investigates whether there exist synergy effects between several advertising methods for an omni-channel strategy company. More precisely, the synergy effects of traditional advertising (i.e. print adv., radio adv., television adv.) and online advertising (web adv.). Research confirmed that there are some advertising synergies between multi offline media channels (Naik and Raman, 2003;Vakratsas and Ma, 2005). Naik and Peters (2009) found synergies between investing in offline and online media channels. According to Dinner, Van Heerde and Neslin (2014) considerable research is done about own advertising effects, however more research is needed to understand the synergy effects for an omni-channel company (Ha, 2008).
Table 1 (following page) shows all studies related to this study. It compares their main characteristics and the resulting findings. Many studies use the traditional mediums: radio, television and print advertising. However, most of the researches use only one or two traditional mediums. Research done by Sundar et al., (1998) compared print and online advertisements and found that online advertisements compared to print advertisements show difference in recognition, however not in recall. Nysveen and Breivik (2005) continues on this comparison by adding radio advertising and found that banner advertising and print advertising are more effective than radio advertising. This finding is slightly in contrast with Danaher and Rossiter (2011) who found that traditional mediums (television, radio, newspapers, direct mail) are more trustworthy and reliable than online advertising (email advertising). Further, this study uses almost the same traditional advertising mediums as the study of Danaher and Rossiter (2011), however it takes the lagged effects of radio, television, internet and print advertising into account and it examines whether there are synergy effects between all the advertising mediums. Investigating which advertising channel is more effective or reliable compare to other advertising channels is one way. Another way to investigate this, is by combining advertising channels. Edell and Keller (1989) examine whether there are synergy effects between the two traditional mediums, television and radio, and found that both advertising methods could strengthen each other.
After that, attention has shifted from a interplay between traditional media to interaction with online media, which started with a research done by Chang and Thorsen (2004), who showed that synergy effects exist between television and banner advertising. Joo et al. (2014) also investigated television advertising with regard to online search advertising and found that television advertising increases online search advertising. However, research done by Olbrich and Schultz (2014) concerning the relation between print advertising and online search advertising did not found a significant effect.
4 Table 1: Literature framework on selected studies
Authors, Year Traditional media Demographic factors Online media Dependent variable
Synergy Major findings
Edell and Keller, 1989 Television, radio - - Retrieval, evaluative processing
yes Television and radio advertising could strengthen each other Chang and Thorsen, 2004 Television - Banner advertising Source credibility, Ad credibility, Number of positive thoughts, Brand credibility, Attitude toward the brand, Actual purchase behaviour
Yes Synergies between television and banner advertising leads to significant higher perceives message credibility, higher attention and a greater number of total and positive thoughts compared to repetition. Olbrich and Schultz, 2014 Print - Search engine advertising Impression, conversion, rank, click, contract
No Print advertising had no direct significant effect on search advertisements impressions. Sundar et al., 1998
Print - Online ads Recall,
recognition No Online advertisements show compare to print advertisements difference in recognition, however not in recall Nysveen and Breivik, 2005
Print, radio - Banner
advertising Attitude to advertised product, attitude to the product, decision support No Banner advertising and print advertising are more effective than radio advertising
Voorveld, 2011
Radio - Online Brand recall,
brand recognition, brand attitude, actual purchase behaviour
Yes Combining online and radio
advertising leads to more positive affective and behavioural
5 Table 1 (continued) Authors, Year Traditional media Demographic factors Online media Dependent variable
Synergy Major findings
Danaher and Rossiter, 2011 Television, radio, newspapers, direct mail Age, gender, education, employment, status, presence of children Email Trust, reliability of information No Television, radio, newspapers and direct mail advertising are the most trustworthy and reliable source of advertising Joo et al., 2014 Television - Online search advertising Consumer search behaviour Yes Television advertising increases online search advertising for financial services This study Radio, television, print, lagged effects Age of the housewife, income of the household educational level Internet Actual purchase behaviour
Yes The combination of folder advertising and the lagged effect of television advertising have the strongest synergy, followed by the combination of print and radio advertising.
To analyse the effects of advertising on purchasing, it is of great importance to identify the various customer groups who shop across different channels, so companies can optimally invest their advertising budget. As mentioned before, television is an important medium when it comes to advertising. Research suggest that customers with a low income spend twice as much time viewing television, prefer the medium more than the general population and mentioned television as a more trustworthy media above other media (Greenberg and Dervin, 1970). When investigating whether age has an effect on the relation between recall and persuasion in advertising, Phillips and Stanton (2004) conclude age is a crucial factor. They found that younger customers (between 18 and 35 years) are more likely to recall content in an advertisement, however they are less likely to be persuaded by the information. The other way around applies to mature customers (65 years and older). They are less likely to recall the information in an advertisement, however they are more likely to be persuaded by it.
6
3. Conceptual model
In this section a framework is developed to clarify the interplays between advertising media and actual purchase behaviour. In addition, the demographic factors: age of contact person, income, and educational level are taken into account to see whether they have an effect on the interplays. One could think of the following questions; do customers have contact with traditional advertising when purchasing a product or do they use online advertising? Does income has an effect on the relation between online advertising and actual purchase behaviour ?
Figure 1 represents the framework where on the left hand side the four advertising mediums are displayed, namely internet advertising, radio advertising, print advertising and television advertising. On the right hand side, the actual purchase behaviour is displayed. The middle of this framework consist of the three demographic factors: Age of contact person, income of the household and educational level of households main earner. These demographic factors could have an influence on the relationship between the advertising mediums and actual purchase behaviour.
7
3.1 Own effects of advertising mediums
3.1.1 Internet advertising
Traditional and internet advertising are two mediums that a company could apply to inform or to persuade customers. Internet advertising is growing very fast since its appearance in 1994 and several types of advertisements have been created since that time (Robinson, Wysocka, and Hand, 2007; Burns and Lutz, 2006). It has several advantages for companies above traditional advertising, for example: it is cheaper to advertise on the internet, it is a more specific method to target customers by knowing their shopping behaviour and with online advertising it is possible to get immediate feedback with online surveys and online orders (Mehta and Sivadas, 1995; Markham, Gatlin-Watts, and Bounds, 2001). However, there are some disadvantages concerning internet advertising. Research showed that some internet users might appreciate seeing relevant advertisements, many say they find the relevant advertising creepy and don’t like the idea of being followed on the internet. (McDonald & Cranor, 2010; Turow et al, 2009). This finding is supported by Dewitt-Elmer (1994) who suggest that internet users resent advertising on the internet.
3.1.2 Traditional advertising
Traditional advertising consist of radio, print and television advertising. Martín-Santana, Reinares-Lara & Muela-Molina (2015) did research on radio advertising and found that music is one of the most important factors in radio advertising, especially the combination of music and words. This combination has greater relevance on the radio than in audio-visual media, since messages rely solely on sound (Martín-Santana, Reinares-Lara & Muela-Molina, 2015). However, radio advertisements are externally controlled, this means that receivers have little opportunity to select or focus on relevant or interesting parts of these advertisements and to control the length of the message. This is because they cannot look at it as long as they want, like print advertising in a newspaper (Petty and Priester, 1994).
Sundar et al., (1998) did research on print advertising and online advertisements and conclude that online advertisements show compare to print advertisements a negatively difference in recognition, however not in recall. In addition, print advertising has the advantage that it is re-readable so people have more control, they are able to select or focus on the relevant parts of the print advertisement. So what could be seen as an disadvantage for radio advertising could be an advantage for print advertising.
The last advertising medium that will be used in this a study is television advertising. Austin and Husted, (1998) did research on the relationship between radio advertising and television advertising in the public mental health education. They found that television advertising is more expensive compared to radio advertising, radio advertisements also have the benefit that people could listen to it in the car while driving to their work or home.
8 that banner advertising and print advertising have greater advertising effects than comparable single advertising mediums. Following this research by Wakolbinger et al., (2009), it is expected that print advertising comes in second place of most effective advertising medium with regard to actual purchase behaviour.
Further, this study expect that customers are slightly more influenced by television advertising compared to radio advertising since the combination of seeing the product and the related price could be of more importance than only hearing information about the product and the price on the radio. Hence, this leads to the following hypotheses.
H1a: Internet advertising leads to more actual purchase behaviour than print, radio and television advertising, whereas print advertising leads to more actual purchase behaviour than radio and television advertising, and television advertising leads to more actual purchase behaviour than radio advertising
3.2 Synergy effects between advertising mediums
3.2.1 Synergy effects between online and traditional mediums
Radio and online:
It could be that offline and online advertising mediums invigorate each other, which means that the combined effect of the advertising mediums is stronger than the sum of all the individual parts (Belch & Belch, 2003). When investigating consumers responses to internet advertising it is important to take the influence of media multitasking into account. Media multitasking is especially common when people are using computers (Roberts & Foehr, 2008). For example; when people use computers, they could also listen to the radio (Pilotta et al., 2004). This is supported by Carrier et al. (2009) who showed that 91% of the web surfers listen to the radio when surfing online. This study expect that when people are working on their computer, they will be influenced by radio advertisements.
Television and online:
Another synergy effect that is expected in this study is between television advertising and online advertising. When people watch commercials on the television, they might get interested in the product or services and go online to search for more information. The other way around is also possible, when people are getting interested by a banner advertising and later on see the same sort of advertisement on the television. This expectation is supported by Chang and Thorson, 2004 who found that the combined effect of television and web synergies results in a higher attention rate, increased message credibility, and greater amount of positive thoughts. Havlena, Cardarelli, and De Montigny (2007) also support this finding of media synergies.
Print and online:
9
3.2.2 Synergy effects between traditional mediums
Synergies between the traditional advertising methods could also be possible. First of all, the cross-effects between radio advertising and print advertising. Jagpal (1981) was the first in presenting synergies in multimedia advertising by studying radio advertising and print advertising for a commercial bank. However, the model created by Jagpal ignored the carryover effect in dynamic markets. This study expects synergies between the two methods because one could listen to the radio and hearing an interesting advertising and thereby consult a print advertising for more information.
Second, television advertising could also be used in a combined effect with print advertising. Naik and Raman (2003), did research on this relationship and found empirical evidence on the existence of this combined effect in consumer markets. They also showed that the combined effect of multimedia advertising activities such as, television, radio, print and internet can be much stronger than the sum of all the individual media activities. This study expects a synergy effect between print and television advertising, because one could see a great offer of a product or service on a television and want to know more about this so could search for more information in a print advertising. Finally, synergy effects could exist between television advertising and radio advertising. It is possible that one could see an advertisement on the radio and later on recall this by hearing the same advertising on the radio. In this way the two advertising mediums invigorate each other, which is in line with Edell and Keller (1989), how found that consumers recall television commercials when listening to the radio advertisements. The expectation of the synergy effect of radio and television is also supported by study called ‘’Image Transfer Study’’. This field study sponsored by UK Radio network companies proved that radio advertisements reinforce messages created by television advertisements by showing that 73% of the participants remembered visuals of television advertisements when hearing radio advertisements. In addition, 57% relived the TV commercials while listening to the radio advertisement.
Taken all the synergy effects into account, this study expects the greatest synergies by combining internet advertising with one of the traditional adverting mediums. The reason is, this study wants to investigate which (combined) advertising medium should be applied to increase the actual purchases. So from a conative perspective, what is the best (combined) media advertising method to influence customers to make a purchase? One way of increasing overall purchases is to attempt to have less delay between the part that a customer gets influenced by an advertisement and the purchase that will be made. Internet advertising could be recommend for this, because the internet capacitate integration of advertising and purchases. When looking at television advertising and radio advertising, both methods could strengthen each other, however to make a purchase, the customer still has to visit a store or website to order the product. In summary, this study expects the strongest synergy effects between internet advertising and traditional advertising and will expect less synergies among the traditional advertising methods with regard to actual purchase behaviour. Hence, this results in the following hypotheses:
10
3.3 Demographic factor age
The demographic age could have an influence on the relation between selecting the right advertising medium and actual purchase behaviour. According to Phillips and Stanton (2004) is age an important factor when it comes to recall and persuasion of content in advertisements. More precisely, research by Denney (1982) found that around the age of 40 detrimental effects begin to start. This means that older adults have a generalized decline in the rate of cognitive processing and the capability of processing information (Phillips & Sternthal, 1977). However, the knowledge base of mature customers remain intact, so the difficulties consist of acquiring and using new information under an abundant amount of information (John and Cole, 1986). Compared to younger adults, older adults make more use of schema-based concepts (Yoon, 1997) and heuristics. In more detail, both concepts are slightly different. Schemas are methods to process information more easily based on categorization and generalization (Yoon, Cole, and Lee, 2009). For example, when there is an information overload in an online advertising, older adults intend to focus more on a keyword, instead of focussing and reading the whole content in the advertisement. Heuristics can be described as simple decision rules, which involve not much processing effort (Chaiken, 1980). Banners could be used for this sort of advertising, because banners mostly consist of pictures and phrases that denote the quality or features of the product (Chaiken, 1980; Petty, Cacioppo, and Schumann, 1983). Since mature customers are expected to show a greater attention to online advertisements than younger adults, one could expect that older customers would be more influenced by online advertising, when purchasing a product. Moreover, using radio and television advertising receivers have little opportunity to select and focus on the relevant content than with internet or print advertising. This is because print and internet advertising are re-readable in contrast to the first two advertising mediums. Taken age into account, one could suggest that older customers have more difficulties with radio and television advertising than with internet and print advertising. Hence, this results in the following hypotheses:
H2a: Age enhances the effect of internet advertising and print advertising on actual purchase behaviour H2b: Age weakens the effect of radio advertising and television advertising on actual purchase
11
3.4 Demographic factor income
Another demographic factor that might affect the selection between traditional advertising and internet advertising could be income. Shavitt, Lowrey and Haefner (1998), showed that customers with a lower income are less offended by advertising. This is in contrast to wealthier customers, who are more likely to dislike advertising (Alwitt & Prabhakar, 1992). However, not much academic literature is available about the effect of income on the choice for using internet or traditional advertising. Investigating from an indirect way, research found that time spend is an important source when people go shopping (Bhatnagar et al., 2000). According to Stigler (1961) value customers with a higher income their time more than customers with a low income, because of its opportunity cost. Normally, less time is taken when shopping online than shopping in brick-and-mortar store, because of the time consuming activities when shopping in a retail store (driving to the store, search for parking space, etc.) (Bellman et al., 1991; Rohm and Swaminathan, 2004). This could suggest that customers with a higher income prefer online shopping above going to the store to purchase products and in this manner deal with more internet advertising compared to the traditional advertising methods. This leads to the following hypotheses:
H3a: Income enhances the effect of internet advertising on actual purchase behaviour
H3b: Income weakens the effect of print, radio and television advertising on actual purchase behaviour
3.5 Demographic factor educational level
The last demographic that might have an effect on the relation between selecting the right advertising medium is educational level. Educated customers are less likely to enjoy advertising and are less likely to rely on advertising in making purchasing decisions compared to less educated customers (Shavitt, Lowrey and Haefner, 1998). Not much academic literature is available about the relation between educational level and the way the customer is influenced by online or traditional advertising methods. However, looking at the internet usage of customers is an indirect way to analyse the relationship. According to Teo (2001) educational level is negatively related to internet usage in terms of messaging, browsing, downloading and purchasing. This is in contrast with research done by Igbaria (1993) who states that higher educational level is positive related to usage of computers. However, Igbaria investigates educational level with regard to the usage of computers in terms of usefulness and anxiety, whereas Teo examines educational level regarding purchasing. So looking at the four advertising mediums, this study expect that customers with a high educational level should have less contact with internet advertising and more contact with print, radio and television advertising concerning actual purchase behaviour . Hence, this results in the following hypotheses:
H4a: Educational level enhances the effect of print, radio and television advertising on actual purchase behaviour.
12
4. Data description
To test the conceptual model
,
a panel dataset provided by GFK will be used. GFK is a world-wide company who has relevant market and consumer information in which it enables their clients to make smarter decisions. The company uses innovative technologies and data sciences to turn big data into smart data so their clients can improve their business (GFK, 2016).For this case, the study will use a panel data set of the durable company X, which is a well-suited company to use because it takes advantage of different advertising channels. For example, it uses tablets in store so customers can see the whole catalogue online in store. In this manner, X is able to provide its customers with a greater amount of products. Customers are also able to decide whether they would the product to be delivered at home or purchase the product in store. Furthermore, X uses several advertising mediums like television, radio, print, folder and internet advertising to inform and to persuade its customers to purchase a product of X. So one could conclude that this company is appropriate when investigating which advertising channel should be most effective if the company uses an omni-channel strategy. The panel dataset consists of weekly household level data of 11672 households covering 31 weeks between week 48, 2010 and week 26, 2011 and contains several demographic factors and advertising mediums of which per week the number of contacts households had with the certain advertising medium are recorded.
4.1 Advertising mediums
The variables that resemble the advertising channels are radio advertising, television advertising, print advertising, folder advertising, internet advertising and the lagged effects of television, radio, print and internet. These lagged effects mean that one could see or hear an advertisements and respond to it a week later. However, the lagged effect of folder advertising is not taken into account since the folder is distributed weekly.
In order to give a well-grounded answer to the research question, some observations of the independent and dependent variables will not be taken into account and some observations will be replaced with more credible values.
First of all, the weeks between the three weeks in 2010 (week 48 until 51) will not be taken into account because no purchases are made by any household during this period. Further, the weeks of 21-2011 until 26-2011 of the variable ‘’print advertising’’ will not be taken into account since there were no print advertisements during that period and finally the weeks between week 20 until 26 in 2011 will not be included for advertising mediums television and folder since no commercial was send out by television and no folder was distributed this period.
Second, the variable folder advertising will not be taken together with print advertising, because the correlation between the two methods is very low (0,12; see table 2). However, the medium folder advertising is an important medium which should be taken into account since it comes in fourth place when looking at the sum of the contacts customers had with folder advertising (see table 2).
third, the variable that displays internet advertising has one value (2253) in the dataset, which is really high compared to the other values. For this reason the value of 2253 is replaced by 491, which value comes in at second place.
13 obtained. The dataset shows that the most contacts have been with television advertising (720557) followed by print advertising (543645) and radio advertising (131748). The two channels with the least contacts are folder of X (102754) and internet advertising (27542). It should be mentioned that the variables that reflect traditional advertising (i.e. radio, television, print, folder) are measured by using the Reach, Recency, Frequency method, which means that a chance is determined regarding what respondents read/listened to/watched, when they did so for the last time and how often they do so normally. So in this manner the variable print advertising can have high values, while folder advertising is always between 1 and 0. This is the case, because of the fact that a folder is distributed weekly, someone has the ability to see the same folder once a week. This is in contrast with print advertising that consist of prints in newspapers or posters which can be seen in more newspapers or one could see several posters during a week.
4.2 Demographics
The dataset also consist of the demographic variables age of the housewife, income of household and educational level of the main earner of the household. When looking at the demographic variable age it is important to mention that this study will use people who are 18 years or older. So the observations where a respondent was below 18 will not be taken into account. Further, the dataset showed that some household wouldn’t give information about their income, which is displayed by the value “99”. Those values are not included in the analysis, because it gives no information about the income of a household.
When going deeper into the data, some insights can be acquired regarding the demographic variables. The data reveals that the average age is almost 50 years. In addition, the youngest respondent is 15 years and the oldest 109. Looking at income, the data shows that most of the respondents (17,2 percent) do not want to mention their income. Furthermore, 2,2 percent of the households have an income which is 700 euro or lower and 6,0 percent of the households have an income which is higher than 4100 euro. Finally, the data displays that most of the respondents have a secondary vocational education (28,1 percent) followed by high professional education (21,3 percent).
4.3
Household level
14 Table 2: Correlations among all advertising mediums
15 Table 3: descriptive statistics among all advertising mediums, demographic variables and the dependent variable
Variables Mean Std. dev. Minimum Maximum Sum
Radio (contacts) .36 1.87 0 28 131,748 Television (contacts) 2.57 4.13 0 58 720,557 Print (contacts) 1.86 3.18 0 18 543,645 Folder (contacts) .28 .41 0 1 79,552 Internet (contacts) .07 1.12 0 164 26,955 Google Masthead (contacts) .01 .15 0 31 2,118 GDN (contacts) .22 3.33 0 491 78,024 Banner (contacts) 0 .27 0 131 723 Age (years) 50 14 15 109 - Income (scale) 1.62 .63 1 3 - Education (scale) 2.08 .80 1 3 - Purchases (nominal) 0 0,07 1 0 1,321
4.4 Dependent variable of interest
Actual purchase behaviour:
Actual purchase behaviour will be measured by the variable called aankoop_JN_52_26, which means that this study will take the purchases of a household between the period of week 52 in 2010 until week 26 in 2011 into account. The variable takes the value one if a customer purchased a product and when the product is not purchased the value of zero is displayed.
4.5 Independent variables of interest
Print advertising & lagged effects:
This channel consist of all prints and folder advertising. Prints are messages that are displayed in newspapers, posters, magazines, etc. and folder advertising consist of the folder that is distributed weekly. As mentioned before, both advertising methods will not be taken together, because of the fact that the correlation between the two is very low (0,12; table 2).
Radio advertising & lagged effect:
16 Television advertising & lagged effect:
Television advertising contains all the commercials that will be send out by the television and lagged effect means that people could respond to an advertising on television a week later. Or in other words, could purchase a product a week later. Television advertising is measured by a chance that households had contact with television advertising, which is also a applied to the lagged effect of television advertising.
Internet advertising & lagged effects:
This channel consist of Google Display Network, Google masthead and banner advertising. First of all, Google Display Network (GDN) consist of contacts with advertisements of X in the network of Google on the internet. Second, Google masthead contains the actual contacts with the advertisements of X, which shows up in the first search options if you are searching for a product X has. Third, banner advertising can be seen as a graphical advertising material on a website. Visitors can click on it and will be redirected to the website of the advertiser in the banner. An average will be made of the those three variables due to the fact that those variables are count variables (actual contacts).
4.6 Demographic variables of interest
The demographic variables will consist of age of the housewife, educational level of the main earner and income of the household.
Income:
The variable income will display the income of the household, which is split up in 21 various ranking orders. The rankings 1 until 9 will display low income, rankings 10 until 19 will display middle income and finally the value of 20 will resemble incomes higher than 4100 euro.
Educational level:
The variable educational level of the main earner of a household will be used in this study which consist of 15 categories. These categories will be split up in low education, middle education and high education.
Age:
17
5. Model section
The model will be analysed by means of binary logistic regression, so the significance of the hypotheses can be tested and one could conclude which variables are most important. The binary logistic regression is used to estimate the probability of a binary dependent variable on a set of independent variables. It measures the relationship between the dependent variable and independent variables by estimating probabilities using a cumulative logistic distribution. Logistic regression is comparable to a probit regression, because it uses the same sort of techniques, which result in a very similar distribution shape except the tails, which are heavier for the logistic distribution (Greene, 1997). In addition, the probit model uses a cumulative normal distribution instead of a cumulative logistic distribution (Blattberg, Kim, & Neslin, 2008).In addition, the reason for choosing a binary logistic regression is because the interpretation of a logistic regression is easier compared to a probit model.
The logistic cumulative density function, which is applied to the binary logistic regression is provided below.
𝐹(β′Xi) = exp(β ′𝑋
i) 1 + exp(β′𝑋i)
Figure 2: Cumulative distribution function for a binary logistic model
The cumulative density function is a function that always satisfies the condition of a percentage, which must be between zero and one. The maximum outcome that the function can have is one and the minimum outcome is zero. That is the reason why this function is used for this study since the dependent variable takes the value of zero if no purchase is made and takes the value of one if a
purchase is made.
18 The following variables will be included in the models:
Table 3: variables used in the model
Variable Label Measurement
Pht Purchase of household h at X in
week t
1=purchase, 0=no purchase
TVAhᵼ Number of television advertising
contacts of household h in week t
A chance that households had contact with television advertising
TVAλh,t-1 Carry-over-effect of the number of
television advertising contacts of households h in week t-1
A chance that households had contact with the lagged effect of television advertising
RAht Number of radio advertising contacts
of household h in week t
A chance that households had contact with radio advertising
RAλh,t-1 Carry-over-effect of the number of
radio advertising contacts of households h in week t-1
A chance that households had contact with the lagged effect of radio advertising
PAht Number of print advertising contacts
of household h in week t
A chance that households had contact with print advertising
PAλh,t-1 Carry-over-effect of number of print
advertising contacts of household h in week t-1
A chance that households had contact with the lagged effect of print advertising
FAht Number of folder advertising contacts
of household h in week t which is between 0 and 1
A chance that households had contact with folder
advertising
IAht Number of internet advertising
contacts of household h in week t
Average of households actual contacts with Google Display Network, Google masthead and banner advertising IAλh,t-1 Carry-over-effect of number of
internet advertising contacts of household h in week t-1
Average of households actual contacts with Google Display Network, Google masthead and banner advertising
Ah Age (in years) housewife of household
h
19 Table 3: (continued)
Variable Label Measurement
NIh Net income of household h Low income means the
households have an income between 0 and 2100 euro. Middle income means the households have an income of 2100 euro up to 4100 euro and high income means that a households have an income more than 4100 euro. Dummy NIh middle Dummy for middle income of
household h
Middle income of a household is 2100 up to 4100 euro
Dummy NIh high Dummy for high income of household
h
High income of a household is more than 4100 euro
ELh Educational level main earner of
household h
Low – middle - high educational level
Dummy ELh middle Dummy for middle educational level
of the main earner of household h
Middle educational level consist of general secondary school, gymnasium,
secondary vocational
education and apprenticeship training.
Dummy ELh high Dummy for high educational level of
the main earner of household h
20
6. Results and discussion
In this section, five models will be estimated of which the first is a model without interaction. It test the effects of print, folder, radio, television, internet advertising and the lagged effects on actual purchase behaviour. The second model will examine whether there exist synergies between the different advertising methods this study uses. After that, three models will be tested of which all three consist of interaction effects. The first one will take the moderator effect of age of housewife into account, the second one will use the net income of a household as a moderator variable and the last one will take the moderator variable educational level of the main earner into account.
6.1 Direct effect
The aim of this study is to examine which advertising medium is most effective for an omni-channel company. This is investigated by means of a logistic regression model. First, the direct effect of the advertising methods to actual purchase behaviour is examined, whereas the direct effect consist of print advertising, folder advertising, radio advertising, internet advertising, television advertising and the lagged effects of print, radio, television, internet to actual purchase behaviour . In addition, the variables age, income and educational level serve as control variables.
The output showed that the model as a whole is highly significant (p=0.000) and looking at the output of the parameters, one could conclude that the lagged effect of folder advertising is positive significant (p=0.000) at a 5 percent level, which means that an additional contact with the lagged effect of radio advertising result in an increase in odds of purchasing a product of X. This is also true for the lagged effect of radio advertising of X (p=0.017). As indicated that the two variables are significant, it is necessary to actually investigate which variables (advertising mediums) are most important to generate purchases. As mentioned before is the folder of X highly significant followed by the lagged effect of radio advertising. When looking at the odds ratios the following insights could be acquired. First of all, for an additional contact with a folder of X, the odds of purchasing a product at X is higher by 99.57 percent. Secondly, for an additional contact with the lagged effect of radio advertising, the odds of purchasing a product of X is higher by 6.38 percent.
21
6.2 Synergy effects
To investigate the synergy effects, a binary logistic regression is applied of which the output showed that model as a whole is highly significant (p=0.000). When looking at the output of the parameters, one could conclude that interaction between print and the lagged effect of print, the interaction between folder and the lagged effect of television and the combined effect of print and radio advertising are significant (resp. p=0.051; p=0.076 and p=0.090) at a 10 percent level. The odds ratios for the latter two are positive, which indicate that an additional contact with a folder of X and noticing the commercial a week before on the television leads to a 3.95 percent increase in odds of purchasing a product of X. For the interaction between print and radio advertising is this 2.21 percent. In contrast to these two synergy effects is the combination between print and the lagged effect of print negative. This means that an additional contact with print advertising and noticing a print a week earlier leads to a decrease in odds of purchasing a product of X by -0.98 percent. An explanation for this could be that customers get an aversion due to too many prints. However, not much academic literature is available concerning this relationship.
Further, the results do not fully support the hypothesis (1b), which state that ‘’Synergy effects exists between radio, television, print, internet advertising with regard to actual purchase behaviour and the synergies between internet advertising and traditional advertising are greater than the synergies between the traditional methods.’’ The results do not indicate that there are synergy effects between all the advertising methods we used. The synergy effects that exist are the two mentioned above, namely the interaction between folder and the lagged effect of television and the interaction between print and radio advertising.
The second part of the hypothesis, which state that “the synergies between internet advertising and traditional advertising are greater than the synergies between the traditional methods” is also not supported since the results show that the synergy effects between traditional advertising, consisting of the interaction between folder and the lagged effect of television and the interaction between print and radio, are the only positive significant interactions with regard to actual purchase behaviour. This could be explained by the ability that television and radio advertising have by making use of sound for radio advertising and a combination of sound and visuals by television advertising which could have a stronger effect of influencing a customer to purchase a product. According to Martín-Santana, Reinares-Lara & Muela-Molina (2015), who state that messages rely solely on sound, could be an explanation why radio and television advertising have a positive effect in combination with a folder of X compared to internet advertising. Internet advertising consist, as mentioned before, in this study of three types. Google Display Network has the highest contacts (78024) with customers among the other two types and means that a advertising shows up at different sites in the network of Google. When assuming that these advertisements are without sound most of the time, one could follow the results of the research by Martín-Santana, Reinares-Lara & Muela-Molina, who showed that sound is an important factor to deliver a message. This finding could be an explanation why the combination of folder advertising with the lagged effect of television advertising and the combination of print advertising with radio advertising have a stronger influence to actual purchase behaviour compared to an interaction with internet advertising.
22 Table 4: Outcomes binary logistic regression for direct model and synergy model
Variable Odds Ratio Std. Error p-value
Radio .9928 .0266104 0.787 TV .9843 .0246042 0.527 Folder 1.9957 .20988 0.000 Print 1.0034 .0241625 0.889 Internet 1.0210 .0318116 0.504 Lag Radio 1.0638 .0275658 0.017 Lag TV 1.0108 .0123228 0.375 Lag Print .9954 .0199051 0.819 Lag Internet .5149 .365805 0.350 Print*Folder 1.0467 .0433662 0.271 Print*TV 1.0002 .0029751 0.931 Print*Radio 1.0221 .0131621 0.090 Print*Internet .9926 .0108585 0.495 Radio*Internet .9737 .0208017 0.213 Radio*TV .9919 .0067517 0.220 Radio*Folder 1.0103 .0474749 0.826 TV*Internet .9917 .0104985 0.443 TV*Folder 1.0403 .0251748 0.102 Internet*Folder 1.0349 .061782 0.566 Lag Print*Lag TV 1.0035 .0021754 0.103
Lag Print*Lag Radio 1.0198 .0154121 0.194
Lag Print*Lag Internet 1.0058 .0069059 0.397
Lag Radio*Lag TV .9992 .0055223 0.885
Lag Radio*Lag Internet .9872 .0260204 0.627
Lag Internet*Lag TV .9981 .0015278 0.709 TV*Lag TV .9980895 .0020105 0.342 TV*Lag Radio 1.005994 .006326 0.342 TV*Lag Print 1.004419 .0028343 0.118 TV*Lag Internet .788199 .2207591 0.395 Radio*Lag TV .9994 .0051977 0.908 Radio*Lag Radio .9991493 .0019352 0.660 Radio*Lag Print .9808453 .0164926 0.250 Radio*Lag Internet .980861 .0216363 0.381 Folder*Lag TV 1.039516 .0227068 0.076 Folder*Lag Radio .9691826 .0451199 0.501 Folder*Lag Print .9499295 .0340767 0.152 Folder*Lag Internet .9478014 .0983125 0.605 Print*Lag TV .9979224 .0047328 0.661 Print*Lag Radio .9825389 .0110083 0.116 Print*Lag Print .9901667 .0050196 0.051 Print*Lag Internet .9888163 .0316142 0.725 Internet*Lag TV 1.003248 .0131313 0.804 Internet*Lag Radio 1.02338 .0153061 0.122 Internet*Lag Print 1.00993 .0065588 0.128 Internet*Lag Internet .9952876 .0046041 0.307 Age .700235 .0675973 0.000
Dummy income middle 1.175643 .1070722 0.076
Dummy income high 1.445105 .2340452 0.023
Dummy education middle .9895268 .1113331 0.925
Dummy education high 1.097591 .1269792 0.421
23
6.3 Demographic factor age
To investigate whether age has an influence between the relation of all advertising methods to actual purchase behaviour, a binary logistic regression is created. The model as a whole is highly significant (p=0.000), however the output of the parameters shows that only the combination of age with the lagged effect of radio advertising is significant (p=0.095) at 10 percent level. This means that customers with an age of 40 and older who have an additional contact with the lagged effect of radio leads to an increase in odds of purchasing a product of X by 6.28 percent compared to customers with an age of 18 up to 39.
The hypothesis (2a) which state that, “Age enhances the effect of internet advertising and print advertising on actual purchase behaviour”, is not supported by the data. A possible explanation why age does not enhance the effect of print advertising on actual purchase behaviour could be that older customers use less internet compared to young customers and in this manner, deal with less internet advertising. An explanation why age does not enhance the effect of print advertising on actual purchase behaviour could be that print advertising is independent of age. This means that both, young customers and older customers, use print advertising to purchase products.
The next hypothesis (2b) which state that, ‘’Age weakens the effect of radio advertising and television advertising on actual purchase behaviour’’, is in contrast with the results which shows that age enhances the effect of lagged radio advertising on actual purchase behaviour. This finding could be explained by the concept of impulse buying. Research by Ekeng et al., 2012 showed that older customers are less impulsive than young customers in purchasing products. In this manner, it could make sense that older customers, comparing to young customers, hear the advertisement on the radio in week t-1 and need more time to consider before actual purchasing a product in week t. The next part of the hypotheses which state that, “age weakens the effect of television advertising on actual purchase behaviour ” is also in contrast with the data. This could be explained by that young and older customers are influenced by television advertising when purchasing products, so it is independent of age.
Table 5: Outcomes binary logistic regression for model with demographic factor age
Variable Odds Ratio Std. Error p-value
24
6.4 Demographic factor income
To examine whether income has an influence on the relation between all advertising channels and actual purchase behaviour, a logistic model is created. Following the output one could see that the model is highly significant. Looking at the output of the parameters, one can conclude that print advertising combined with an average (middle) income is significant (p=0,016) at a 5 percent level. This means that customers with an income between 2100 - 4100 euro who have an additional contact with print advertising leads to an increase in odds of purchasing a product of X by 9.54 percent compared to customers with an income between 0 - 2100 euro. Further, the output shows that print advertising combined with a high income is also significant (p=0.036). This means that customers with an income of more than 4100 euro, who have an additional contact with print advertising leads to an increase in odds of purchasing a product of X by 18.04 percent compared to customers with an income between 0 – 2100 euro.
This outcome is in contrast with the hypothesis (3a) which state that income should have a positive effect on the relation between internet advertising and actual purchase behaviour . As mentioned earlier, there is not much academic literature available about the effect choosing the right advertising channel depending on income of the customers. However, one explanation could be that more people can afford a computer or laptop nowadays. So even people with a lower income are making use of computers and internet (Markham, Gatlin-Watts & Bounds, 2001). This could be an explanation why income has no effect on the relation between internet advertising and actual purchase behaviour.
25 Table 6: Outcomes binary logistic regression for model with demographic factor income
Variable Odds Ratio Std. Error p-value
Radio .9887058 .0201544 0.577 TV 1.010293 .0102381 0.312 Folder 1.956 .1954292 0.000 Print .971937 .0177416 0.119 Internet 1.007151 .0230162 0.755 Lag radio 1.058395 .0219114 0.006 Lag TV 1.018597 .0106331 0.078 Lag Print .9841398 .0160125 0.326 Lag Internet .901801 .060987 0.126
Dummy middle income 1.222969 .1155794 0.033
Dummy high income 1.46932 .2554434 0.027
Dummy middle income*Radio 1.011685 .0428765 0.784 Dummy middle income*TV .9812289 .0203605 0.361
Dummy middle income* Print 1.095437 .0415973 0.016 Dummy middle income*Internet .9743875 .0427319 0.554 Dummy middle income*Folder .8330788 .1743946 0.383 Dummy middle income*Lag Radio .993102 .0438285 0.875
Dummy middle income* Lag TV 1.01476 .0221288 0.502 Dummy middle income*Lag Print .9620778 .031081 0.231 Dummy middle income*Lag Internet .8767097 .1016192 0.256 Dummy high income*Radio .9977693 .0738138 0.976
Dummy high income*TV .9397697 .0414127 0.159
26
6.5 Demographic factor educational level
To investigate whether educational level has an influence between the relation of all advertising channels to purchase intention, a logistic model is created. The model as a whole is highly significant (0,000), however the output of the parameters shows that the demographic variable educational level is in no case significant. This is in contrast with the hypotheses (4a & 4b), which state that ‘’Educational level enhances the effect of print, radio and television on actual purchase behaviour’’ and ‘’Educational level weakens the effect of internet advertising on actual purchase behaviour’’. So following the results, it might be the case that educational level is unimportant when it comes to selecting the best advertising medium to purchase a product.
Table 7: Outcomes binary logistic regression for model with demographic factor educational level
Variable Odds Ratio Std. Error p-value
Radio .9963597 .0177696 0.838 TV 1.01373 .0093869 0.141 Folder 1.967432 .1771564 0.000 Print .9850954 .0165236 0.371 Internet 1.015509 .00771 0.043 Lag radio 1.040854 .0183416 0.023 Lag TV 1.01635 .0096011 0.086 Lag Print .9758438 .0150125 0.112 Lag Internet .9360635 .0472783 0.191
Dummy middle education 1.01695 .1066532 0.873
Dummy high education 1.247654 .1271403 0.030
Dummy middle education*Radio 1.048601 .0432453 0.250
Dummy middle education*TV 1.021039 .0242842 0.381
Dummy middle education* Print .9722538 .0401311 0.495
Dummy middle education*Internet .9957714 .018319 0.818
Dummy middle education*Folder 1.300111 .2964486 0.250
Dummy middle education*Lag Radio .9399336 .0413894 0.159
Dummy middle education* Lag TV 1.021588 .0232556 0.348
Dummy middle education*Lag Print 1.063397 .0390033 0.194
Dummy middle education*Lag Internet
1.165759 .1837574 0.331
Dummy high education*Radio 1.041033 .0468827 0.372
Dummy high education*TV 1.029677 .0248249 0.225
dummy high education*Print 1.010519 .0421428 0.802
dummy high education*Folder 1.004033 .0157379 0.797
Dummy high education*Internet 1.108945 .2554261 0.653
Dummy high education*Lag Radio .9341827 .0414477 0.125
Dummy high education*Lag TV 1.022779 .022214 0.300
Dummy high education*Lag Print 1.006855 .0421697 0.870
Dummy high education*Lag Internet 1.102598 .187518 0.566
27
7. Model fit
To aim of this study is to examine which advertising medium should be chosen by an omni-channel company based on the data set given by GFK. So this study is more about descriptive than predictive. Nevertheless, it is worthwhile to investigate model fit. Table 9 shows the goodness of fit for all models estimated.
Table 9: Model fit
Looking at the table, one could see that synergy model fits best based on the log likelihood. Furthermore, the model has the lowest log likelihood, the highest Nagelkerke R² and McFadden R² (resp. 0.015; 0.015). However, based on the BIC the logit model without synergies fits best because it has the lowest BIC (9717). It should be mentioned that the two models are different because the direct model does not account for synergies, whereas the synergy model does. For this reason they are difficult to compare.
To finalize this section, a probit model was created for all the five models, however the outcomes regarding model fit, significances of the parameters and the directions are slightly the same.
Logit
(direct
model)
Logit
(synergy
model)
Logit
(age model)
Logit
(income
model)
Logit
(education
model)
Log likelihood -4767 -4752 -5845 -4835 -5797 Likelihood ratio chi-square (df) 112,858(14). p=0.000 142.808(50). p=0.000 120.161(19). p=0,000 107.227(29) p=0.000 118.065(30). p=0.000Cox & Snell R² 0.001 0.001 0.001 0.001 0.001
Nagelkerke R² 0.012 0.015 0.011 0.011 0.010
McFadden R² 0.012 0.015 0.007 0.011 0.010
AIC 9564 9606 11609 9622 11656
28
8. Conclusion
In this study, we attempt to find an advertising medium that is most effective for an omni-channel company by investigating which advertising channels customers use to purchase a product. The advertising channels consist of radio advertising, television advertising, internet advertising and print advertising, whereas print advertising consist of prints and folder advertising.
First, we find that folder advertising is by far the most beneficial advertising medium when taking all the adverting mediums into account. Folder advertising is followed by the lagged effect of radio advertising, which indicate that radio advertising is an important medium.
Second, the study examines whether there are synergies between the several advertising methods. The analysis shows us that the interaction between print advertising and radio advertising is significant, which is also true for the interaction between folder advertising and the lagged effect of television advertising and the combination of print advertising and the lagged effect of print advertising. However, the latter is negatively significant which means that is has no synergies. Furthermore, the combination of folder and lagged television performs slightly better than the combine effect of print and radio advertising.
Third, the study investigates whether the demographic variables: age of the housewife, income of the household and educational level of the main earner could have an effect on the relation between the advertising medium and the actual purchase behaviour. The output of the binary logistic regression shows that age enhances the effect of lagged radio advertising on actual purchase behaviour. Which means that customers with an age of 40 and older who have an additional contact with the lagged effect of radio could lead to more purchases compared to customers between 18 and 40. The following demographic factor ‘’income’’ is positive significant in combination with print advertising. Which means that customers with a middle income (2100 - 4100 euro) and high income (more than 4100 euro) who have an additional contact with print advertising could lead to more purchases compared to customers with a low income (between 0 -2100 euro). Finally, there are no significant effects for the demographic factor ‘’educational level’’.
Model Main findings
Direct model - Folder advertising most effective, followed by
lagged effect of radio advertising.
Model with synergy effects - Combination of folder and the lagged effect of
television advertising performs slightly better than a combination of print and radio advertising
Model with moderator age - Customers with an age of 40 and older who have an
additional contact with the lagged effect of radio advertising could lead to more purchase compare to between the age of 18 and 40.
Model with moderator income - Customers with a middle and high income who have
an additional contact with print advertising could lead to more purchases compared to people with a low income.
Model with moderator educational level - No significant results
29
8.1 Managerial Implications
This study has several managerial implications. First, the results show that there exist synergy effects between traditional mediums. In particular, managers of companies using an omni-channel strategy should focus on a combination of folder and the lagged effect of television advertising and a combination of print and radio advertising, whereas the first combination performs slightly better. A suggestion for the first combination could be that managers highlight the products which sells best in a TV commercial and give more information about these products in a folder. The second combination could exist of a print which shows a discount for a certain product and highlight this discount by a radio ad. Second, this study shows when only focussing on one channel, that managers should take into account that folder advertising is by far the most effective advertising medium to generate sales, followed by the lagged effect of radio advertising. Third, when including demographic factors, managers should know that age and income have an significant positive effect on the relation between the advertising mediums and the actual purchase behaviour. In particular, customers with an age of 40 or older are more likely to use the lagged effect of radio advertising when purchasing a product. Managers could take advantage of this finding by sending out radio advertisements on radio stations for older customers. Next, the results indicate that customers with a middle and high income are more likely to use print advertising when purchasing a product. Managers could benefit from this finding by providing posters to places where people with a middle and high income are located. Finally, no significant effects were found regarding the combination of traditional advertising and internet advertising.
8.2 Limitations
30
References
Alwitt, L. F., & Prabhaker, P. R. (1992). Functional and belief dimensions of attitudes to television advertising: Implications for copytesting. Journal of advertising research, 32(5), 30-42.
Austin, L. S., & Husted, K. (1998). Cost-effectiveness of television, radio, and print media programs for public mental health education. Psychiatric Services.
Belch, G. E., & Belch, M. A. (2003). Advertising and promotion: An integrated marketing communications
perspective. The McGraw− Hill.
Bellman, S., Lohse, G. L., & Johnson, E. J. (1999). Predictors of online buying behaviour. Communications of the
ACM, 42(12), 32-38.
Bhatnagar, A., Misra, S., & Rao, H. R. (2000). On risk, convenience, and Internet shopping behaviour. Communications of the ACM, 43(11), 98-105.
Blattberg, R. C., Kim, B. D., & Neslin, S. A. (2008). Why Database Marketing? (pp. 13-46). Springer New York.
Burns, K. S., & Lutz, R. J. (2006). The function of format: Consumer responses to six on-line advertising formats. Journal of Advertising, 35, 53. Spring
Carrier, L. M., Cheever, N. A., Rosen, L. D., Benitez, S., & Chang, J. (2009). Multitasking across generations: Multitasking choices and difficulty ratings in three generations of Americans. Computers in Human
Behaviour, 25(2), 483-489.
Chaiken, S. (1980). Heuristic versus systematic information processing and the use of source versus message cues in persuasion. Journal of personality and social psychology, 39(5), 752.
Chang, Y., & Thorson, E. (2004). Television and web advertising synergies. Journal of Advertising, 33(2), 75-84. Danaher, P. J., & Rossiter, J. R. (2011). Comparing perceptions of marketing communication channels.
European Journal of Marketing, 45(1/2), 6-42.
Davis, F. D., & Venkatesh, V. (2004). Toward preprototype user acceptance testing of new information systems: implications for software project management. Engineering Management, IEEE Transactions
on, 51(1), 31-46.
Denney, N. W. (1982). Aging and cognitive changes. Handbook of developmental psychology, 807, 827. De Vaus, D. (2013). Surveys in Social Research, 6th edn. Allen & Unwin, Crows Nest.
Dewitt-Elmer, P. (1994). The Battle for the Soul of the Internet. Time, 25, 50-6.
Dinner, I. M., Van Heerde, H. J., & Neslin, S. A. (2014). Driving online and offline sales:
The cross-channel effects of traditional, online display, and paid search advertising. Journal of
Marketing Research, 51(5), 527-545.
