MSc Marketing Intelligence and Marketing Management
Omni-channel retailing: Analysing the dynamic and the interaction effects of both traditional and online advertising.
Theme: Accountability of marketing activities
By
Rixt Gerlofsma s2364565
January 15, 2018 Master thesis First supervisor:
Prof. Dr. J.E. Wieringa Second supervisor:
Dr. K. Dehmamy
Rixt Gerlofsma Marwixstraat 14 9726 CD Groningen +31623704801
rixtgerlofsma@hotmail.com University of Groningen
Faculty of Economics & Business Duisenberg building
Nettelbosje 2 Groningen
ABSTRACT
There is nothing more difficult, controversial and complex in marketing, then measuring the influence of advertising on sales (Zhou, Zhou and Ouyang, 2003). The continuing rise of new communication channels changed retail business models, resulting in multi-channel retailing and following in omni-channel retailing. Omni-channel retailing takes all marketing initiatives into account by which customers are influenced and use in their search and buying process (Verhoef, Kannan and Inman, 2015). Whereas customers want a perfect integration of the physical and the digital, it is crucial for retailers to create an integrated omni-channel strategy. Rigby (2017) even poses that a successful omni-channel strategy guarantees a retailer’s survival. Nevertheless, omni-channel retailing challenged retailers to optimally allocate their media budget across channels. Moreover, the combination of multiple offline (traditional) and online media advertising channels, hinders retailers to thoroughly understand to which extent a particular channel contributes to sales.
With the rise of big data, retailers can make an auspicious step in understanding the effect of both traditional and online advertising (Grewal, Roggeveen and Nordfält, 2017).
Using data from a large non-food retailer, this paper analyses these omni-channel retailing challenges. More specifically, this paper examines the dynamic and interaction effects of both traditional (print, radio, TV) and online (display) advertising channels and how these channels influence offline sales. We find strong evidence for both strong short-term and long-term effects of traditional and online advertising channels. In the short run, print advertising and TV advertising negatively affect sales, whereas radio advertising and display advertising positively affect sales. In the long run, only traditional advertising channels significantly affect sales. While print advertising and TV advertising only display a negative relationship, radio advertising shows both a positive and negative relationship regarding sales. Examining the timing of a specific marketing channel and their profit effects enabled the interesting finding that the wear-in and wear-out effects differ among media advertising channels. That is, the various media advertising channels show significant effects at differing times. Next, we propose interesting insights concerning interaction effects. Our results indicate that there are both within-synergy effects when combining traditional advertising forms and cross-synergy effects when combining traditional advertising forms with online advertising forms.
Keywords: Omni-channel retailing, advertising channels, traditional advertising, online advertising, dynamic effects, interaction effects
PREFACE
Employing all knowledge, methods and analytical skills learned during my studies in previous years, together with many hours of labor, I hereby proudly present my Master Thesis. This thesis finalizes my Master of Science study in both Marketing Management and Marketing Intelligence at the University of Groningen.
First, I would like to thank my thesis supervisor Prof. Dr. J. E. Wieringa for his valuable comments and contributions to this thesis. Especially, I am very grateful to the many things I learned in the field of data analysis, which I would certainly continue to employ in my future career. My gratitude goes also out to the other staff members of the Marketing faculty for all their professional and passionate insights and for most teaching skills.
Next, I am thankful to my fellow students in my thesis group. Together, it was a very pleasant ride to the so-called finish line of finalizing our masters. While, everyone chose a completely different topic to investigate, still I have learned valuable insights when discussing about things together.
Finally, I am very grateful to all external sources of support, especially those of friends and family. They have always encouraged and motivated me during my entire studies. Also, they had the confidence in me that I can reach any goal I want. Even though they might not grasp all the contents of this thesis, I know they will be proud of me. To all, thank you for your support, love, patience, and confidence in me.
Rixt Gerlofsma Groningen, January 15, 2018
TABLE OF CONTENT
ABSTRACT ... 2
PREFACE ... 3
CHAPTER 1 – INTRODUCTION ... 6
1.1 INTRODUCTION ... 6
1.2 RESEARCH PURPOSE ... 7
1.3 THEORETICAL AND SOCIAL RELEVANCE ... 9
1.4 STRUCTURE OF PAPER ... 10
CHAPTER 2 - LITERATURE REVIEW ... 11
2.1 CONSUMER PURCHASE DECISION MAKING PROCESS ... 11
2.2 THE EFFECTIVENESS OF MEDIA ADVERTISING CHANNELS ... 13
2.2.1 Traditional advertising ... 13
2.2.2 Online advertising ... 14
2.2.3 Effect on sales amount ... 16
2.3 THE DYNAMIC EFFECTIVENESS OF MEDIA ADVERTISING CHANNELS ... 17
2.3.1 Short- and long-term effects ... 18
2.3.2 Wear-in and wear-out effects ... 20
2.4 INTERACTIONS AMONG MEDIA ADVERTISING CHANNELS. ... 21
2.4.1 Synergy effects ... 21
2.4.2 Cannibalization effects ... 22
2.5 CONCEPTUAL FRAMEWORK ... 24
CHAPTER 3 – RESEARCH DESIGN ... 26
3.2 RESEARCH METHOD ... 26
3.1.1 Tobit-2 Model ... 26
3.1.2 Dynamic advertising effects ... 28
3.1.2.1 Short-term and long-term effects ... 29
3.1.2.2 Wear-in and wear-out effects ... 33
3.1.3 Interaction advertising effects ... 33
3.2 DATA COLLECTION ... 34
3.2.1 Dependent variables ... 34
3.2.2 Independent variables ... 37
CHAPTER 4 – RESULTS ... 43
4.1 THE EFFECTIVENESS OF MEDIA ADVERTISING CHANNELS ... 43
4.2 THE DYNAMIC EFFECTIVENESS OF MEDIA ADVERTISING CHANNELS ... 46
4.2.1 Short- and long-term effects ... 46
4.2.2 Wear-in and wear-out effects ... 52
4.3 INTERACTIONS AMONG MEDIA ADVERTISING CHANNELS ... 53
4.3.1 Synergy and cannibalization effects ... 53
4.4 VALIDATION ... 57
4.4.1 Face validity ... 57
4.4.2 Statistical validity ... 57
4.4.3 Modelling issues ... 59
CHAPTER 5 – CONCLUSIONS AND RECOMMENDATIONS ... 60
5.1 DISCUSSION ... 60
5.1.1. The effectiveness of media advertising channels ... 60
5.1.2. The dynamic effectiveness of media advertising channels ... 61
5.1.3. Interactions among media advertising channels ... 63
5.3 LIMITATIONS AND FUTURE RESEARCH ... 65
BIBLIOGRAPHY ... 66
APPENDICES ... 73
APPENDIX A – DESCRIPTIVE STATISTICS OF NUMBER OF HOUSEHOLDS ... 73
APPENDIX B – DESCRIPTIVE STATISTICS OF SALES ... 73
APPENDIX C – DESCRIPTIVE STATISTICS OF OFFLINE SALES ... 73
APPENDIX D – NUMBER OF SALES OVER TIME ... 74
APPENDIX E – VALE OF SALES IN EUROCENTS OVER TIME ... 75
APPENDIX F – VIF SCORES BASIC LOGIT SALES MODEL ... 76
APPENDIX G – VIF SCORES BASIC SALES LOG-LINEAR MODEL ... 76
APPENDIX H – VIF SCORES DIRECT LAG LOGIT MODEL ... 76
APPENDIX I – VIF SCORES DIRECT LAG LOG-LINEAR MODEL ... 77
APPENDIX J – VIF SCORES GEOMETRIC LAG LOGIT MODEL ... 78
APPENDIX K – VIF SCORES GEOMETRIC LAG LOG-LINEAR MODEL ... 78
APPENDIX L – VIF SCORES LOGIT MODEL WITH INTERACTIONS OVER T ... 78
APPENDIX M– ESTIMATES AND VIF SCORES LOGIT MODELS WITH INTERACTIONS OVER TIME ... 79
APPENDIX N – VIF SCORES LOG-LINEAR MODEL WITH INTERACTIONS OVER T ... 81
APPENDIX O – ESTIMATES AND VIF SCORES LOGIT MODELS WITH INTERACTIONS OVER TIME ... 81
CHAPTER 1 – INTRODUCTION 1.1 Introduction
‘’ Half the money I spend on advertising is wasted; but the trouble is, I don’t know which half’’
The oft quoted line of John Wanamaker captures the desire of retailers to know the effectiveness of marketing initiatives which retailers’ use as a source of attracting customers and persuading them to make a purchase (Bluefly, 2011). Nonetheless, there is nothing more difficult, controversial and complex in marketing, than measuring the influence of advertising on sales (Zhou, Zhou and Ouyang, 2003). Already a promising step has been made with the introducing of the online channel, in which the online channel allows retailers to draw a direct line from advertisement to sale (Hoffman and Novak, 2000). Nonetheless, by the introducing of the online channel retail business models have changed dramatically, resulting in first multi-channel retailing (Verhoef et al., 2015), and second, with the ongoing digitalization resulting in a omni-channel retail model (Rigby, 2017). Omni-channel retailing takes all marketing initiatives into account by which customers are influenced and use in their search and buying process (Verhoef et al., 2015).
Omni-channel retailing changed how customers behave (Sorescu, Frambach, Singh, Rangaswamy and Bridges, 2011). Accordingly, customers use different advertising channels in different stages of their buying process. They want a perfect integration of the physical and the digital, such that they value the advantages of both (Rigby, 2017). However, traditional advertising channels and online advertising channels target customers in the same stages of their buying process. For example, traditional advertising including TV and print and online display advertising are both advertising forms which target consumers early in the buying process (Dinner, van Heerde and Neslin, 2014). Consequently, this can enhance the availability of synergies across the media advertising channels or contrary enhance the availability of cannibalization of media advertising channels (Kumar and Gupta, 2016;
Kumar, Choi and Greene, 2017).
In addition, customers are becoming more active and can be reached anywhere and anytime (Hennig-Thurau et al., 2010). Therefore, many retailers are allocating increasing parts of their advertising budgets to online advertising to target the always on-the-go customer (Lobschat, Osinga and Reinartz, 2017). With the rise of the internet, banner advertising became available. Today, 47% of online advertising is allocated to banner advertising (ZenithOptimedia, 2016). Yet, there is still debate whether banner advertising positively
effects sales and whether online advertising can really generate both online and offline sales.
Also, 70% of consumers are under the impression that retailers are getting the digital experience wrong (Rigby, 2017).
Recently, Rigby (2017) poses that the future of retailing is a fusion of digital and physical experiences in which a successful omni-channel strategy guarantees a retailer’s survival. With the above discussion in mind this raises the question if indeed combining traditional and online advertising is most effective or that in some cases a retailer can better choose for either traditional advertising or online advertising (Hoffman and Novak, 2000).
Yet, retailers rely on trial-and-error or gut feelings when making budget allocation decisions for their marketing initiatives (Jordan, Mahdian, Vassilvitskii and Vee, 2010).
However, with the rise of big data, retailers can make an auspicious step in understanding the effectiveness of both traditional and online advertising. Now, retailers can draw effective insights from big data which are promising by making budget allocation decisions (Grewal et al., 2017).
1.2 Research purpose
Data from a large non-food retailer will be used to analyse the omni-channel retailing challenges. The retailer has a 360° approach for their communication and marketing initiatives, including TV, radio, print, digital and social media. The behaviour of customers is changing and evolving, which they need to follow. Therefore, they believe that physical and online stores will intertwine completely with each other and that they need to find the right balance between the various advertising forms both offline and online. From a retailer’s perspective knowing the return of advertising channels is necessary to develop an efficient marketing plan. However, translating the rich data obtained from big data, into measures of advertising response has proven difficult (Zantedeschi, Feit and Bradlow, 2016). Thus, retailers need more insights in which marketing initiates when and how to use.
Within the omni-channel retailing literature there is a strong emphasize on the effectiveness of a specific channel or multiple channels on performance metrics of retailers.
Specifically, the integration of channels and their effectiveness is particularly relevant (Verhoef et al., 2015). However, despite the emphasize on integrated marketing advertising in the literature, retailers seldom coordinate their various marketing campaigns (Joo, Wilbur, Cowgill and Zhu, 2014). Thus, in this paper the dynamic and interaction effects of both traditional and online advertising on sales will be investigated. The focus is will be on offline
sales, because the retailer in question predominantly sells their products offline. These insights have been brought together, resulting in the following research question:
‘’What are the dynamic and interaction effects of traditional and online advertising and how do these channels influence offline sales?’’
This research question can be split up into several more specific questions. First, to develop an efficient marketing plan, retailers need to know the full return from all media advertising channels. Thus, what is the effectiveness of media advertising channels?
Second, he relationship between advertising and sales is not straightforward. For instance, consumers see an ad that may encourage them to make a purchase, yet they may wait a few weeks before actually buying the product. This implies that there is a non- immediate effect of advertising, also called long-term effects (Kerho, 2010). Many studies stress the importance of only short-term effects on offline sales (Breuer and Brettel, 2012), nonetheless long-term effects are as well or even more important. Therefore, both short-term and long-term effects of the media advertising channels and their effect on offline sales will be analysed. More specifically, this paper examines when an effect of a media advertising channel ‘hits in’ and how long this effect last. Nonetheless, there are various ways to incorporate non-immediate effects of advertising. This paper will examine how to incorporate these effects within a model. In addition, our analysis of indirect and direct marketing effects on sales and sales amount provides a rationale for the wear-in and wear-out effects of various media advertising channels. This is a key managerial issue which has received little attention in previous literature (Wiesel, Pauwels and Arts, 2011).
Lastly, traditional advertising including TV and print and online display advertising are both advertising forms which target consumers early in the decision-making process (Dinner et al., 2014). Therefore, it can be expected that there are interactions among these media advertising channels. To be more specific, are the media advertising channels in synergy or do they cannibalize each other? These questions motivate and guide the research done in this paper.
1.3 Theoretical and social relevance
Omni-channel retailing has a growing importance in practice. Rigby (2017) even poses that retailers need an effective omni-channel strategy to survive. Therefore, it is critical for retailers to reassess their competitive strategies. To create competitive advantages retailers, need to invest in a omni-channel retailing strategy, in which the physical and online experience are seamlessly intertwined with each other (Verhoef et al., 2015). In a omni- channel retailing environment, customers go through a series of touch points across various advertising forms both offline and online. Thus far, retailers do not fully understand consumers’ preference for each advertising channel in their buying making process, resulting in poor omni-channel strategies (Frambach, Roest and Krishnan, 2017). Nonetheless, with the availability of big data the effectiveness of advertising forms can be measured, which in turn can enhance the performance and efficiency of advertising channels (Kumar & Gupta, 2016).
Thus, it is of top priority for both academic and retailers to assess the effectiveness of both traditional advertising forms and online advertising with the help of big data (Rutz and Bucklin, 2012).
Moreover, literature and the recent Marketing Science Institute state that to adopt an integrated marketing advertising strategy, the dynamic and interaction effects of advertising channels are of high importance (Verhoef et al.,2015; Marketing Science Institute, 2016).
That is, by aiming to create a seamlessly intertwined experience of the physical and online experience for the always on the go consumer one should not only look at the direct effectiveness of advertising channels but as well at how advertising influence each other and change over time. Despite the high relevance of these topics, the role of these effects in planning an omni-channel advertising strategy is not well understood by retailers (Mantrala, 2002; Naik and Raman, 2003). In practice, the multiple touches a customer makes before making a purchase are rarely taken into account when measuring campaign effectiveness across advertising channels (Li and Kannan, 2014). In addition, existing literature lacks a study that simultaneously considers the dynamic and interaction effects of offline and online advertising forms. (Neslin and Shankar, 2009; Kannan, Reinartz and Verhoef, 2016).
The key contribution of this paper is that not only the effectiveness of both offline and online marketing initiatives is investigated but more specifically dynamic and interaction effects. Also, by the use of recent technologies such as big data, one is able to track consumer journeys on an individual level. Hence, this data will give deeper insights into the individual purchase decision processes. This paper is together with Danaher and Dagger (2013) and Anderl, Schumann and Kunz (2015) one of the first which focusses on omni-channel
marketing measuring individual-level exposures to multiple offline channels. Nonetheless, Anderl et al. (2015) closely study how channel usage and preference along the customer journey assist consumers in their purchase decision processes whereas this paper focusses on the effectiveness of advertising channels on consumers purchase decision process. Danaher and Dagger (2013) demonstrate an advertising response model in which only short-term marketing effects are taken in to consideration whereas this paper both short- and long-term effects are included. Also, Danaher and Dagger (2013) do not analyse the interaction effects across channels. Therefore, it is assumed that this paper is unique in its kind and contributes to both the knowledge of academia and retailers.
1.4 Structure of paper
The remainder of this paper is organized as follows. Chapter 2 reviews the literature omni- channel retailing followed by chapter 3 in which the research design and the data used will be discussed. Next, chapter 3 also elaborates on the model and methodology employed in this study. The results are discussed in chapter 4. Finally, some concluding remarks, recommendations, limitations and suggestions for future research are provided in chapter 5.
CHAPTER 2 - LITERATURE REVIEW
Research investigating the effectiveness and interactions of different media advertising channels influencing offline sales can be categorized in four streams of literature: the consumer purchase decision making process, the effectiveness of media advertising channels, the dynamic effectiveness of media advertising channels and the interactions among media advertising channels.
2.1 Consumer purchase decision making process
One of the oldest fields in academic literature is the purchase decision making process (Perloff, 1993). This process which consumers go through when making a purchase, constitutes three stages: pre-purchase, purchase, and post-purchase. The first stage, pre- purchase, encompasses all aspects of consumer interactions with the brand, product, and environment before a purchase is made. The second stage, purchase, covers all consumer interactions with the brand, product and its environment during the purchase itself. The third stage, post-purchase reflects customer interactions with the brand, product, and environment following the actual purchase (Lemon and Verhoef, 2016).
This buying process has been the same for ages, however omni-channel retailing changed the customer. Now, customers are better informed and undertake the buying process at their convenience and their terms (Cook, 2014). Hence, consumers shift between offline and online channels when they move through the buying process stages (Ahuja, Gupta and Raman, 2003). This customer purchase journey is the process customers go through, across all stages and marketing channels, and is also called the ‘customer experience’ (Lemon and Verhoef, 2016). While the buying process still encompasses three stages: pre-purchase, purchase, and post-purchase, customer experience examines a more holistic customer journey.
This process (figure 1) can function as a guide to empirically examine customer’s experience over time during the customer journey, as well as to empirically modelling the effects of different marketing channels on the customer’s experience (Lemon and Verhoef, 2016).
Figure 1: Process model for customer journey and experience
Lemon and Verhoef (2016) employed the phases in the customer journey as an iterative and dynamic process (figure 1). The pre-purchase stage can be characterized as behaviour such as need recognition, consideration, and search. The purchase stage is characterized by behaviour such as choice, ordering, and payment. The post-purchase stage includes behaviour such as consumptions, usage, engagement, and service requests. Within the customer journey, customers might interact with touch points in each stage of the experience. Lemon and Verhoef (2016) identified four categories of customer experience touch points: brand-owned, partner-owned, customer-owned, and social/external.
Within this paper, the focus is on brand-owned touch points. More specifically, we will examine the role of various media advertising channels on the customer experience.
Retailers employing an omni-channel strategy need to pay attention to the various drivers of consumers’ actual channel selections in the different stages of their purchase decision making process (Kollman, Kuckertz and Kayser, 2012). Omni-channel retailing can be defined as
‘’the synergetic management of all the available advertising channels and customer touch points, in such a way that the customer experience across all channels and the effectiveness over channels is optimized’’ (Verhoef et al., 2015). Thus, it is important that retailers try to reach consumers through both offline and online channels and understand which advertising
2.2 The effectiveness of media advertising channels
Within omni-channel retailing retailers are required to balance their advertising across both offline and online advertising channels. Though, studies which investigate online marketing initiatives typically omit offline marketing initiatives and inversely. (Li & Kannan, 2014;
Blake, Nosko and Tadelis, 2015; De Haan, Wiesel and Pauwels, 2016). Offline advertising channels can be referred to as traditional media such as television, print and radio (Dinner et al., 2014). In contrast, online techniques used in order to drive customers to make a purchase are search and display advertising. These, different media advertising channels generate different marketing results and will have different effects on offline sales (Chen and Hsieh, 2012).
2.2.1 Traditional advertising
Traditional advertising can be defined as offline advertising including forms as TV, print and radio. It is hard to overstate the importance of traditional advertising. The average Dutch consumer listens 190 minutes to the radio (GFK, 2017). Moreover, Grimm (2014) poses that print advertising is one of the best performing media channels, with an ROI of 130. That is, print advertising is selective in audience, but fits the content in magazines magnificent. Also, when retailers act as a market leader, they have to choose large-scale media exposure, such as TV advertising (Grimm, 2014). Thus, traditional advertising remains a trusted source of advertising (Danaher and Rossiter, 2011; Joo et al., 2014; Grimm, 2014)
There is extant literature available which document the effectiveness of offline advertising (Tellis, 2009; Sethuraman, Tellis, and Briesch, 2011; Danaher and Dagger, 2013;
Dinner et al, 2014). Advertising effectiveness can be defined as ‘the consumer response to a retailers advertising in the form of making a purchase’ (Tellis, 2009). The effectiveness of advertising is often measured in terms of advertising elasticity, or the percentage increase in sales for a 1% increase in advertising (Sethuraman et al., 2011).
Sethuraman et al. (2011) made based on prior research, generalizations about traditional advertising effectiveness. They meta-analysed 751 brand-level short-term advertising elasticities and 402 long-term elasticities. The average short-term elasticity is 0.12 and the average long-term elasticity 0.24. There is a decline over time in both short- and long- term advertising elasticity. Researchers point to advertising clutter and competitive advertising to explain this phenome. Whereas traditional advertising elasticity is on average equal to 0.12 both radio elasticity and print advertising elasticity are lower than television advertising elasticity. One reason for this effect is that television advertising, with its ability to
arouse emotions, may be more effective than print and radio advertising, which primarily rely on information appeals (Sethuraman et al., 2011).
Danaher and Dagger (2013) studied the relative effectiveness of advertising channels asking consumers about their viewing, listening, reading, browsing, and searching behaviour.
They were in a position to create single-source data by developing a database of individual- level exposure to ten advertising, coupled with the purchase activity of these same people.
Results indicate that traditional advertising still remains effective. Especially, catalogues, television and radio significantly influence purchase incidence. More specifically, the advertising elasticities for these mediums are respectively 0.1 for catalogues, 0.08 for television, and 0.02 for radio. Together, these ideas suggest that there is a positive within- effect of traditional advertising on offline sales, resulting in the following hypothesis:
Hypothesis 1: Traditional advertising positively affect offline sales.
2.2.2 Online advertising
There is a remarkable growth of customers who use the Internet in their purchase decision making process (Joo et al., 2014). Online advertising can be divided into search advertising and display advertising. Search advertising refers to advertisements on a search engine (e.g., Google, Bing) which are displayed when consumers are searching for keywords in a search engine (Anderl et al., 2015). Within search advertising one can distinguish between paid search (SEA; Search Engine Advertising) and organic search. Paid search is sold through bid auctions, in which retailers pay for each consumer who actually clicks on the advertisements (Abou Nabout, Lilienthal, and Skiera, 2014). In contrast, organic search is free of charge, though retailers invest time to optimize their advertisement positions (Dou, Lim, Su, Zhou, and Cui, 2010). Online display advertising is a type of advertising with the use of banner ads (Dinner et al., 2014). These banner ads present visual and textual information about the brand (Naik and Peters, 2009).
Li and Kannan (2014) propose a conceptual framework which estimates the effects of both search and display advertising by allocating credit for conversions using individual-level data on customers who made a purchase at the online channel (conversions). Their results demonstrate that both search and display advertising positively affect conversions.
Nonetheless, although display advertising overall increases the conversion probability, it is important that they are not used indiscriminately to target all consumers. That is, a segment of consumers shows negative behaviour towards display ads, mostly because display ads are
Kannan, 2014). For the paid search channel, Li and Kannan (2014) state that the generalisability of the incremental impact depends on the brand strength of the retailer.
Specifically, the stronger the brand, the lower the incremental effect of paid search on conversion. Suggesting, that when a brand is strong, much of the impact could be recaptured through the organic search channel (Li and Kannan, 2014). While the outcome variable in the above discussion is online sales, this paper focusses on offline sales. Nonetheless, it still grasps an idea of how consumers are affected through online advertising channels which in turn influence their purchase behaviour.
The effect of online advertising on offline sales is still sparse, however due to its growing importance more and more researchers have begun to investigate cross-channel effects of online advertising on offline sales (Chan, Wu, and Xie, 2011; Danaher and Dagger, 2013; Pauwels, Leeflang, Teerling and Huizingh, 2011; Van Nierop et al., 2011). More specifically, Verhoef, Neslin and Vroomen (2007) studied the phenomenon ‘research shopping’ in which they state that it is not uncommon for customers to use online advertising channels to get informed and the offline stores as the purchase channels. Likewise, Dinner et al. (2014) pose that customers prefer to purchase in particular channels, meaning that consumers use the Internet as a search channel and the offline store as purchase channel. As a result, it is very likely to expect cross effects of online advertising on offline sales.
Nonetheless, much of the early work on cross-channel effects used brand awareness, brand recall, or purchase intentions as outcome variable, rather than the outcome variable sales (Cho, Lee, and Tharp, 2001; Dahlen, 2001; Dreze and Hussher, 2003; Manchanda, Dubé, Goh and Chintagunta, 2006).
Though, Abraham (2008) investigated how online display advertising influenced both online and offline sales. Customers who saw a banner ad made compelling more purchases than customers who did not saw a banner ad. Moreover, online display advertising effects were significantly larger on offline sales compared to online sales. Furthermore, Lewis and Reiley (2014) conducted a large-scale field experiment investigating the causal effects of online advertising on sales. They fundamentally found that online advertising is profitable for the retailer, imposing that the incremental revenues exceeded seven times the cost of the advertisements. Especially, the effect on offline sales was large; 93% of the online ads’ total effect was on offline sales. Additionally, research of Lobschat et al. (2017) provides evidence that retailers who predominantly sell through the offline channel can benefit from online advertising. Moreover, they present that online display advertising with the use of banner ads, positively changes consumers’ preferences by reminding them of the retailers’ brand. In turn,
consumers who saw banner ads made more purchases compared to consumers who did not saw banner ads. This effect was even stronger for the offline purchase funnel than for the online purchase funnel.
Elaborating on the above discussion it is expected that there is a positive cross-effect of online advertising on offline sales, resulting in the following hypothesis:
Hypothesis 2: Online advertising positively affect offline sales.
2.2.3 Effect on sales amount
Traditional and online advertising, not only affect purchase incidences but as well purchase amount (Lewis and Reiley, 2014). In light of omni-channel retailing literature which combines different advertising channels and their effects on purchase incidence and purchase amount is scarce. Nonetheless, recently Chang and Zhang (2016) modelled the relevant customer behaviours: channel choice, purchase incidence, and purchase amount, using data from an omni-channel clothing retailer. They identified two dynamic relationship states among consumers which differ in retention and profitability: an active state and an inactive state. Consumers in an active state exhibit a stronger relationship with the retailer through higher purchase amount and more frequent purchases (Chang and Zhang, 2016).
Prior research suggests that traditional advertising is experientially more immersive, which in turn is helpful in building strong relationships with consumers and enhance consumers retailer experience (Verhoef et al., 2007). Accordingly, traditional advertising has an educational purpose migrating consumers from an inactive to an active state (Chang and Zhang, 2016). In contrast, online advertising has a retention purpose keeping currently active consumers active (Chang and Zhang, 2016). Moreover, online advertising channels provide low transactional cost and convenience for those consumers who are already familiar with the retailer experience (Lal and Sarvary, 1999). Chang and Zhang (2016) conclude that for consumers which are in the inactive state, traditional advertising is far more impeccable than online advertising in not only migrating consumers from the inactive state to the active state, but as well in increasing purchase incidence and more importantly purchase amount.
Nevertheless, omni-channel consumers which are for most consumers in active states, tend to appreciate advertising initiatives of various channels and hence are more profitable (Chang and Zhang, 2016). The differential use of channels among shoppers of omni-channel retailers is both associated with channel characteristics and basket composition (Kalyanam, Lenk and Rhee, 2017). Kalyanam et al. (2017) examine the impact of basket composition on
the choice among various channels. Their estimates show that basket composition indeed impacts channel choice and suggest that one could target households based on basket size thresholds. Basket size threshold can be defined as ‘the basket size at which the total cost of using one channel is equivalent to the other channel, and the household is indifferent between channels’ (Kalyanam et al., 2017). Outcomes revealed that a retailer should target households with a small basket size by means of online advertising. Specifically, 43.2% of the households in consideration find the internet more attractive that the offline channel when basket size is one. However, even a greater percentage: 44% of the households in consideration exhibit multi-channel shopping behaviour. This implies that a household has been influenced by more than one channel in a given time period and is more attracted to both the offline and online channel. Hence, these households should be targeted with the use of traditional and online advertising initiatives instead of solely using one of these channels (Kalyanam et al., 2017).
Considering the above discussion, it is expected that (1) traditional advertising is more impactful compared to online advertising in increasing sales amount. Nevertheless, keeping the omni-channel customer in mind it is expected that (2) combining traditional advertising and online advertising is more impactful than traditional advertising in increasing sales amount, resulting in the following hypothesis:
Hypothesis 3a: Traditional advertising is more impactful than online advertising in increasing sales amount.
Hypothesis 3b: Combining traditional advertising and online advertising is more impactful than traditional advertising in increasing sales amount.
2.3 The dynamic effectiveness of media advertising channels
As stated in literature, marketing is dynamic in essence (Leeflang, Wieringa, Bijmolt and Pauwels, 2015). As a result, the relationship between advertising and sales is not necessarily straightforward. Therefore, to properly assess the impact of media advertising channels we should account for dynamic effects of these marketing advertising channels, or simply called marketing dynamics. For instance, the effects of traditional and online advertising can be both on the long-term, increase sales even after the campaign is over, or/and on the short-term, increase sales only during the period of the campaign (Lewis and Reiley, 2014). Or, looked at somewhat differently, this implies that there is a non-immediate effect of advertising (Kerho, 2010). In addition, the dynamic effectiveness of media advertising channels refers as well to the duration media advertising channels and their effectiveness. That is, how long does an
advertisement have an impact and refers to the wear-in and wear-out effects of the various media advertising channels.
2.3.1 Short- and long-term effects
Still many retailers believe in the conventional wisdom that the impact of advertising is only on the short-term (Zhou et al., 2003). However, traditional and online advertising have both short- and long-term effects on sales (Lewis and Reiley, 2014; Wildner and Modenbach, 2015). For instance, consumers see an ad that may encourage them to make a purchase, yet they may wait a few weeks before actually buying the product. This suggests that there is indeed a non-immediate effect of advertising (Kerho, 2010). These effects are the dynamic effects of advertising focussing on ‘when’ advertising is effective.
Academic literature poses that there are strong dynamic effects of both traditional and online advertising (Lecinski, 2011; Kireyev, Pauwels, and Gupa, 2015). Nevertheless, these dynamic effects depend on the consumer purchase decision making process. That is, as consumers tend to consider products over a long-time horizon these products benefit from strong dynamic effects and as consumers tend to consider products only over a short-time horizon these products may not benefit from strong dynamic effects as much (Kireyev et al., 2015). A recent study conducted by Google poses that products which are considered over a long-time horizon are for most financial investments such as home or car purchases.
Consumers are most influenced by the advertising efforts 1-2 months prior to conversion (Lecinski, 2011). Next, the study finds that consumers are most influenced by advertising efforts 1-4 weeks prior to conversion as they purchase for technology and electronic goods and advertising efforts which target consumers on the same day as conversion are most effective for consumer-packaged goods. Likewise, literature on advertising and consumer behaviour suggests that consumers purchase decision making process is different for high- and low-involvement products (O’Cass, 2000; Shimp, 2000; Zhou et al., 2003). Particularly, consumers are generally more highly involved when purchasing nondurables compared to durables, because they are more concerned about reducing risk. As a consequence, advertising tends to create a long memory effects in the minds of consumers (Zhou et al., 2003). This suggests that advertising campaigns can have long-lasting impacts on sales.
Literature imposing the dynamic effects of the various advertising channels is scarce.
Nonetheless, a generalization study of Tellis (2009) poses that the carryover elasticity of advertising is twice as large as that of the current effect of advertising. More specifically, Joseph (2006) poses that television advertising affects sales several periods after the original
exposure, which is referred to as the decay effect. The decay effect implies that for each new exposure to advertising the awareness increases to a new level. The awareness will be higher if there have been recent exposures and lower if there have not been recent exposures (Joseph, 2006). Similarly, Wildner and Modenbach (2015) studied the effectiveness of TV advertising in the age of social media. They particularly considered both the short- and long-term effects of TV advertising to answer the question if TV advertising has become obsolete in the age of social media. According to their results, TV advertising achieved a positive ROI for all examined brands. The average short-term ROI was 1.15 whereas the average long-term ROI was 2.65. Long-term effects are therefore essential for obtaining a fair evaluation of advertising effects (Wildner and Modenbach, 2015).
In line with the foregoing, concerning online advertising Lewis and Reiley (2014) found that online display advertising, constituting to enhance the image of the retailer, led to persistent positive effects on sales for a number of weeks after the banner ads stopped showing. Also, while consumer behaviour tends to be exploratory at initial stages, consumers who have been exposed to display ads may click on these ads to get more information and may eventually convert in later more goal directed search stages (Novak and Hoffman, 2003;
Kireyev et al., 2015). Moreover, Kireyev et al. (2015) posed that display ads had a significant effect on the increase of the retailers’ revenue. Nonetheless, the majority of this increase took effect only after two weeks. This entails that returns of online advertising may be underestimated when only looking at the short-term sales effects.
Concluding, by only examining short-term effects (Danaher and Dagger, 2013) both the effectiveness of traditional advertising and online advertising can be undervalued. Thus, it is expected that (1) media advertising channels exhibit long-term effects which impact offline sales and that (2) both traditional and online advertising positively affect offline sales in the long run, following in hypotheses 4a and 4b.
Hypothesis 4a: Media advertising channels exhibit long-term effects which impact offline sales.
Hypothesis 4b: Traditional and online advertising both have positive long-term effects on offline sales.
2.3.2 Wear-in and wear-out effects
When focussing on the duration of the effectiveness of media advertising channels one need to examine two aspects: wear-in and wear-out (Tellis, 2009). Wear-in occurs when the effect of an advertisement has a significant positive effect on consumers, resulting in making a purchase (Pechman and Stewart, 1988). Contrary, wear-out transpires when the effect of an advertisement continues to decrease with repetition of the advertisement within a campaign, resulting in that media advertising no longer significantly affects sales (Pechman and Stewart, 1988). Tellis (2009) performed an empirical generalization study leading to the following important findings regarding wear-in and wear-out effects: (1) wear-in effects happen quite rapidly, (2) wear-in effects occur more slowly when exposures are spread apart and when consumers are not actively processing the advertisements, (3) wear-in effects are stronger for advertisements that have higher persuasion scores, (4) wear-out effects occur more slowly when advertisements are less effective and exposures are spread apart, (5) wear-out effects occur more slowly when consumers’ are light viewers of TV, (6) wear-out effects arise faster when there was a break in the campaign.
In addition, in the consideration of wear-in and wear-out effects one should also acknowledge the concept marketing persistence. Marketing persistence can be defined as ‘the extent to which a relatively short-term change in the marketing mix, such as advertising initiatives, leads to a long-lasting effect on sales (Dekimpe and Hanssens, 1995). Academic literature shows divergent views regarding marketing persistence. For instance, Jones (1995) states that more and longer uses of advertising are better than less and shorter uses of it. This implies that retailers are better off using advertising campaigns which last a long-time compared to advertising campaigns which only last for a short-time.
Likewise, Kerho (2010) showed that marketing persistence differs among media advertising channels. For instance, consumers have deep personal connections to magazines, which often leads to high levels of engagement with magazine ads. In turn, highly involved consumers tend to have good memory of relevant advertising messages. While a positive brand image is built in consumers’ minds, the advertising message has a strong marketing persistence (Zhou et al., 2003). In contrast, banner ads are often lower-funnel focussed which implies that the goal is to persuade consumers to make a purchase (Kerho, 2010). Little involvement is required leading to a less strong brand image in consumers’ mind. Hence, the advertising message needs to be repeated more often to create a strong marketing persistence (Zhou et al., 2003). These insights result in the following hypothesis:
Hypothesis 5: Wear-in and wear-out effects differ among advertising channels
2.4 Interactions among media advertising channels.
Interactions among media advertising channels is an important theme within marketing (Naik and Raman, 2003). Numerous studies have found that a particular advertising channel may influence other advertising channels (Naik and Raman, 2003; Assael, 2011; Vakratsas and Ma, 2005; Joo et al., 2014). Now, consumers live in a world of simultaneous media usage (Schultz, 2002). They read the newspaper while they listen to the radio. They watch TV while they browse on the Internet. Consequently, this can enhance the availability of synergies across the media advertising channels or contrary enhance the availability of cannibalization of media advertising channels (Kumar and Gupta, 2016; Kumar et al., 2017).
2.4.1 Synergy effects
There are numerous studies which investigate the joint effects of media advertising channels called synergy effects (Naik and Raman, 2003; Vakratsas and Ma, 2005; Joo et al., 2014).
Synergy effects can be defined as ‘’when the added value of one advertising channel resulting from a combined use with another advertising channel, exceeds the sum of the individual effects of both advertising channels’’ (Naik and Raman, 2003). A distinction can be made between within-media synergies and cross-media synergies, whereas within-media synergies refers to synergies within-offline channels or within-online channels, cross-media synergies refers to synergies across online- and offline-channels (Naik and Peters, 2009).
Naik and Raman (2003) created an integrated marketing communications framework.
This framework demands that each advertising channel enhances contributions of other advertising channels. Specifically, they showed that within-synergy effects are present when combining both TV and print advertising in consumer markets. Regarding online within- synergy effects, Abraham (2008) found evidence for synergies between online search ads and online display ads. Results showed that when retailers use both search and displays ads in a campaign, both online and offline sales increase more compared two separate campaigns for each advertising channel.
Combining both traditional advertising (TV, print and radio) and online advertising (banners and search) there is evidence that there are cross-media synergies (Naik and Peters, 2009). Naik and Peters (2009) proposed a hierarchical synergy model which includes both within- and cross-media synergies. This model generates higher-order media interactions to investigate more complex natures of media synergy effects for a compact car brand. They
have found that indeed cross-media advertising synergies exist. In addition, both Chang and Thorson (2017) and Lim, Ri, Egan and Biocca (2015) present that there are synergistic effects among TV and internet advertising. They pose that both advertising channels provide unique features which are absent in the counterpart. When used in conjunction, TV and online advertising, can attract more attention than they could by themselves, resulting in a positive effect on sales. Likewise, Havlena et al. (2007) pose that there are synergies between print and online advertising. Retailers can internalize these synergies to use both print and online versions of ads to target their customers and persuade them to make a purchase.
Even though studies suggest that there are synergistic effects among media advertising channels, there is still no sufficient information available which can be used by retailers to enhance their media advertising strategies (Woo, Ahn, Lee and Koo, 2015).Nonetheless, still valuable propositions can be drawn of the reviewed literature, (1) combining both TV and print advertising affects offline sales more positively than only using one of them, (2) combining traditional advertising with online advertising will affect offline sales more positively than solely using traditional advertising or online advertising, following in the subsequent hypotheses:
H6a: There are within-synergy effects when combining traditional advertising forms or online advertising forms, which positively affect offline sales.
H7a: There are synergy effects when combining traditional and online advertising, which positively affect offline sales.
2.4.2 Cannibalization effects
Cannibalization can be defined as ‘’when multiple advertising channels are being used and two or more are viewed as substitutes, resulting in that a specific advertising channel is responsible for a decline of another advertising channel’’ (Sridhar and Sriram, 2015).
According to Dijkstra, Buijtels and Raaij (2005) employing multiple media advertising channels can be effective, nonetheless when using too many media advertising channels the effect on offline sales can be negative. Moreover, retailers can perceive redundancies when reaching customers via multiple media advertising channels (Sridhar and Sriram, 2015).
When this occurs the media advertising channels used are probably viewed as substitutes; a particular advertising channel is accountable for the decline of another advertising channel effectiveness.
Various studies have shown that cannibalization exists among offline or among online advertising channels. Kolsarici and Vakratsas (2011) argued that consumers can become saturated when they convey similar information by different media advertising channels.
Specifically, they found that cannibalization effects exist when using both TV and print advertising. Online advertising channels show cannibalization effects as well. Kireyev et al.
(2015) pose that search advertising due to its strong dynamic effects, can lead to a decreased effectiveness of display advertising. Specifically, they propose that ad budgets should be allocated by increasing search ad budget by 36% from its current level, and decreasing display ad budget by 31%.
Across channels, customers who desire service, rather than risk aversion, prefer offline channels more than online channels. When the desire of service becomes even more important, online advertising channels face cannibalization (Kollman et al., 2012). In addition, according to Sundar, Hesser, Kalyanaraman, and Brown (1998) print advertising is superior to online advertising when it comes to memory measures. This implies that customers are more likely to recall a print ad compared to for instance a banner ad. In turn, this imposes that customers who saw a print ad are more likely to make a purchase compared to customers who saw a banner ad.
In contrast, when online channels are added to a retailers’ distribution system, traditional advertising channels can also face cannibalization (Montoya-Weiss, Voss and Grewal, 2003; Steinfield, 2004). Alba and Hutchinson (1987) posed already in 1987 that online media advertising channels surpass traditional advertising whenever online media advertising provides more appealing characteristics. Furthermore, cannibalization effects are most likely to occur when online advertising channels closely mimic the offline setting according to Deleersnyder, Geyskens, Gielens and Dekimpe (2002). Likewise, the growth in online media advertising accounts partly for the decline in traditional marketing advertising posed in the article of Sridhar and Sriram (2015). Specifically, Sridhar and Sriram (2015) found that 7-17% of the loss in print advertising can be traced back to cannibalization effects, resulting from online newspaper advertising growth. This implies that consumers view print advertising and online newspaper advertising as substitutes. Another substantial fraction of the decline in print advertising is related to the emerging of other online advertising channels, such as search and display advertising. This implies that cannibalization effects occur with the rise of the Internet (Sridhar and Sriram, 2016).
In summary, investigating whether there are cannibalization effects when using multiple media advertising channels is very valuable and is of high relevance when retailers are making marketing decisions. Drawing on previous literature, we expect that (1) the use of multiple traditional advertising forms or multiple online advertising forms decreases the effectiveness of one another and (2) the use of online advertising cannibalizes the use of traditional advertising or vice versa. This results in the following hypotheses:
H6b: There are cannibalization effects when applying multiple traditional advertising forms or online advertising forms, which negatively affect the effectiveness of traditional advertising.
H7b: There are cannibalization effects when applying both traditional and online advertising, which negatively affect the effectiveness of traditional advertising.
2.5 Conceptual framework
On the basis of findings from previous research, a conceptual framework has been proposed, which is displayed in figure 2. First, it is expected that impressions from both traditional (TV, print and radio) and online (display and search) advertising channels positively affect consumers purchase decision. Second, not only affect traditional and online advertising purchase incidences but as well purchase amount. Hence, it is expected that traditional advertising is more impactful versus online advertising, but combining traditional and online advertising will outperform traditional advertising. Third, by looking only at short-term effects the effectiveness of advertising can be undervalued. Therefore, it is expected that advertising has as well long-term effects. Fourth, omni-channel retailing includes the use of various advertising channels. Thus, it is expected that positive or negative interactions exist across and among various offline and online advertising channels. Finally, the relationship between advertising and sales is not always straightforward. For example, consumers see an ad that may encourage them to make a purchase, yet they may wait a week before actually buying the product. Therefore, it is expected that there are decaying media effects of advertising channels and that these effects differ among advertising channels.
Figure 2: Conceptual framework media advertising channels influencing sales.
CHAPTER 3 – RESEARCH DESIGN
Within the marketing literature questions about decisions of consumers at their individual level are central (Fok, 2017). For this paper, individual household data from a large Dutch non-food retailer is used. The available data tracks, on a weekly household level, the behaviour of consumers regarding the retailer. Analysing this data gives valuable insights in when and if consumers purchase a product, and if consumers purchase a product how much the profit effects are. Variables affecting the purchase decision of consumers are advertising channels.
Consumers shift between offline and online channels when they move through the buying process stages (Ahuja et al., 2003). Therefore, to persuade consumers to make a purchase, it is important for a retailer to understand which advertising channels are most effective in each stage of the buying process (Frambach et al., 2007).
Due to data limitations, this paper will focus only on the advertising channels which influence consumers earlier in the buying making process (Dinner et al., 2014). Channels which target consumers early in their buying making process are traditional advertising (TV, print, radio) and online display advertising. Therefore, online search advertising is excluded in this research design.
In summary, the focus of this paper is the effectiveness of omni-channel retailing, including the dynamic and the interaction effects of the various advertising forms. More specifically, the short- and long-term effects, the wear-in and wear-out effects, and the synergy and cannibalization effects of the media advertising channels will be investigated.
3.2 Research method
Aiming to assess the effectiveness of omni-channel retailing, including the dynamic and interaction effects of the various advertising forms a Tobit-2 model will be used. Subsequent, different variables for the various interaction effects will be added to the model so one is able to analyse these specific effects.
3.1.1 Tobit-2 Model
Tobit models constitute a specific subset of an integrated model, which are used when the dependent response variable is a limited continuous variable (Leeflang et al., 2015). That is, the response is only observed for those customers who purchase or respond. When a customer does not purchase or respond, the variable equals zero. The individual household weekly level data presents a large proportion of zero expenditures. This is very understandable given that
the data covers purchases for only a 31-week period and the economic-lifetime of durable products sold by the retailer in question is longer than the examined period.
In Tobit models, one specifies two equations: one to explain the decision to respond (purchase) and one to explain the demand (amount spent) (Tobin, 1958). These models integrate which factors determine whether a customer responds or not and, if there is a response, which factors determine the size of the response. One could differ Tobit models in two types: type-1 Tobit model and type-2 Tobit model. A Tobit-1 model is a mix between a logit model for the events Y=0 versus Y>0 and a linear regression model, where the explanatory variables and parameters in both models are assumed to be the same.
Nonetheless, this might be a heroic assumption to expect that the same drivers (1) explain whether or not a purchase is made by a customer and (2) how much will be purchased if a purchase does occur. Likewise, Fok (2017) poses that advertising channels have a different impact on the purchase incidence itself compared to the how much one spends per purchase.
Therefore, in this study we will use a Tobit-2 model also known as a hurdle model, which contains two separate parts. One part investigates the binary choice if a customer makes a purchase or not, while the second part describes the outcome conditional on that a purchase occur. That is, the Tobit-2 model accommodates the cases where the decision to make a purchase is driven by other explanatory variables than the decision on how much to spent.
Basic sales models
To analyse the effectiveness of the various advertising channels on offline sales, a logit model with parameters α has been specified to describe Pr [Y > 0]. In equation 1 the specific model is shown. The explanatory variables are television, radio, print, bannerGoogle, mastheadbannerGoogle, banner, and masthead banner. Hence, it has been decided due to very small average number of impressions per week, to create a new variable Display, which combines all four forms of online display advertising. Also, to control for consumer heterogeneity the variables income and age are as well included. Income controls for differences in spendable euros between households whereas Age reflects the average age and the number of people within an age range of a household. In addition, to compare traditional advertising and display advertising another model is estimated in which the variables TV, radio and print are combined in one variable called Traditional. Second, to analyze the profit effects, conditional on Y > 0, a log-linear model has been estimated. This model has been linearized by taking the logarithm. Now, the coefficients can be interpreted as elasticities.
One can find the specific model in equation 2. The similar variables as used in the logit model are included in the log-linear model.
OfflineSales* = α* + 𝛽-TV* + 𝛽0Radio* + 𝛽4Print* + 𝛽8Display* (1.1) + 𝛽<Income* + 𝛽@Age* + ε*
OfflineSales* = α* + 𝛽-Traditional* + 𝛽0Display* + 𝛽4Income* (1.2) + 𝛽8Age* + ε*
SalesAmount∗F = 𝛼∗* + 𝛽-TV*∗ + 𝛽0Radio*∗+ 𝛽4Print*∗
+ 𝛽8Display*∗ + 𝛽<Income*∗ + 𝛽@Age*∗+ ε* (2) Equation (2) is linear in the parameters 𝛼∗, 𝛽-∗, 𝛽0∗, …, 𝛽@∗, where 𝛼∗ = ln 𝛼 and 𝛽*∗= ln 𝛽*, and is referred to a log-log model.
3.1.2 Dynamic advertising effects
Even though, the time span of the observed households is relatively short: 31 weeks, the marketing advertising variables still exhibit a pattern regarding sales. The time-series plot of sales and each advertising form can be found in figure 3.
Figure 3: Average number of impressions advertising channels and total number of sales over time
One can observe that the sales and advertising patterns are close to each other. Nonetheless, the extreme values seem to have as well opposite signs. That is, when sales are relatively low the impressions of traditional and online display advertising are relatively high, and vice versa. A possible explanation would be that both advertising forms target consumers in the
