Digital marketing campaigns: The effect of channel strategy and advertising phases on mobile and desktop device performance

(1)

Digital marketing campaigns: The effect of channel strategy and advertising

phases on mobile and desktop device performance

Isabelle van Rongen Student number: 10356002 Thesis 21-06-2018 MSc Business Administration: Digital Business University of Amsterdam Supervisor: Shan Chen

(2)

Statement of originality

This document is written by Isabelle van Rongen who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the

(3)

Abstract

In recent years, several academic studies have investigated the concept of digital marketing and the rise of mobile marketing. Nevertheless, current research is lacking knowledge of customer behaviour on different devices. In addition, researchers highlight that the performance of a digital marketing campaign is expected to be different across devices. Therefore, the aim of this research is to combine past insights and study device performance in digital marketing campaigns. By evaluating 700 campaigns that have been executed in the Netherlands over the past two years, an extensive statistical analysis is conducted. First, two regression analyses are performed and subsequently, a mediation effect via PROCESS is tested. The variables that are under investigating are the channel strategy, which includes the overall budget of a campaign and the division of this budget over the channels search, social and display. Second, the overall length of a campaign, as well as the advertising phases split up into the awareness, consideration and purchase phase, are being examined. Finally, the device performance is measured through click-through rates and conversion rates of mobile and desktop devices. The conclusions of this research are that devices differ significantly from each other. Overall, increasing the budget or length of a campaign shows a positive effect on device performance. Nevertheless, when examining an increase in a budget per channel, or increasing the length of a specific advertising phase, several negative relationships are found. Research contributions, as well as limitations and recommendations for future studies, are presented in an extensive discussion.

Keywords_{: Digital marketing, device performance, budget division, channel strategy,}

(4)

List of Tables

Table 1: Means, Standard Deviations, Correlations, Cronbach’s alpha Table 2: Click-Through Rates across devices

Table 3: Conversion Rates across devices

Table 4: Regression model of Mobile Click-Through Rate and Budget per channel Table 5: Regression model of Mobile Conversion Rate and Budget per channel Table 6: Beta values of Mobile Performance per channel

Table 7: Regression model of Desktop Click-Through Rate and Budget per channel Table 8: Regression model of Desktop Conversion Rate and Budget per channel Table 9: Beta values of Desktop Performance per channel

Table 10: Regression model of Mobile Click-Through Rate and Length of Phases Table 11: Regression model of Mobile Conversion Rate and Length of Phases Table 12: Beta values of Mobile Performance per advertising phase

Table 13: Regression model of Desktop Click-Through Rate and Length of Phases Table 14: Regression model of Desktop Conversion Rate and Length of Phases Table 15: Beta values of Desktop Performance per advertising phase

Table 16: Mediation analysis of Mobile CTR, Channel Strategy and Length of Phases Table 16a: Direct, total, and indirect effect (Mobile CTR)

Table 17: Mediation analysis of Mobile CR, Channel Strategy and Length of Phases Table 17a: Direct, total, and indirect effect (Mobile CR)

Table 18: Mediation analysis of Desktop CTR, Channel Strategy and Length of Phases Table 18a: Direct, total, and indirect effect (Desktop CTR)

(8)

Table 19: Mediation analysis of Desktop CR, Channel Strategy and Length of Phases Table 19a: Direct, total, and indirect effect (Desktop CR)

Table 20: Summary of findings

Appendix

Table 4a: Regression model _{of Mobile Click-Through Rate and Channel Strategy} Table 5a:_{Regression model of Mobile Conversion Rate and Channel Strategy} Table 7a: _{Regression model of Desktop Click-Through Rate and Channel Strategy} Table 8a:_{Regression model of Desktop Conversion Rate and Channel Strategy} Table 10a: Regression model of Mobile Click-Through Rate and Total Length Table 11a: Regression model of Mobile Conversion Rate and Total Length Table 13a: Regression model of Desktop Click-Through Rate and Total Length Table 14a: Regression model of Desktop Conversion Rate and Total Length Table 21: Normality tests - Skewness and Kurtosis statistics

Table 22: Tests for independence of variables and multicollinearity Table 23: tests for independence - Durbin-Watson

Table 24: Histograms for normality Table 25: P-P plots

Table 26: Tests for linearity - Scatter plot

(9)

List of Figures

Figure 1: Conceptual framework

Figure 2: Conceptual framework mediation analysis Figure 3: Mobile ads influence conversions on desktop Figure 4: Web Click-Through Rate Comparison Figure 4a: Web Conversion Rate Comparison Figure 5: Customer journey

(10)

1. Introduction

There is no doubt that mobile represents the greatest transformation of marketing in this generation (WARC, 2015). Since 2014, the number of global users on mobile has outgrown desktop users (ComScore, 2017). Next, total minutes spent online are dominated by mobile, highlighting the stake of the device in digital marketing in general (ComScore, 2017). Although mobile has a large share in terms of minutes and users, conversion rates are lower on mobile than on a desktop devices (Adobe, 2016; Criteo, 2015). Nevertheless, mobile shows a strong contribution to a digital marketing campaign’s performance, driving results across the entire path to purchase (WARC, 2015). Although marketing departments are pressured to invest in channels with the highest ROI (Return on Investment), there is still little understanding of the ROI of mobile marketing (WARC, 2015).

In the past decade, research has focused on digital marketing in general whilst keeping an eye on the emerging topic of mobile marketing (Berman, 2016). More researchers highlight that customers on mobile devices are behaving differently from desktop users (Berman, 2016; Criteo, 2015). Nevertheless, no research has been done so far on how the different devices should be used in digital marketing. Next, marketers are putting emphasis on measuring marketing by the large amount of data that can be gathered nowadays (Wang, Malthouse & Krishnamurthi, 2015). Though, when marketers are analysing the performance of campaigns, the link between a campaign’s objectives and device types is not included yet when setting a campaign’s strategy (Grewal, Bart, Spann, Zubcsek, 2016).

First of all, this research will highlight the structural differences between mobile and desktop devices in terms of performance. Second, this study will connect the goals of a digital marketing

(11)

campaign and its channel strategy to the different device types. When a campaign is build up, marketers decide how to use digital channels including search, social and display channels (Li and Kannan, 2014). By dividing budget of the campaign over the three channels, the strategy is set. Next, the campaigns from the dataset are classified into three different phases, namely an awareness phase, a consideration phase, and a final purchase phase. Each advertising phase as a particular goal such as spreading awareness and thus increasing a campaign’s reach, or generating conversions in the final purchase stage (Grewal et al., 2016). Hence, the length of each phase is expected to influence a campaign’s performance. Research has been done in the past on both channel strategy and advertising phases. Therefore, this study will build on past research and corroborate the importance of the two concepts. They will be examined in relation to the performance of the campaigns on different devices, which is measured by click-through rates and conversion rates (Kireyev, Pauwels & Gupta, 2014). Thus, this study elaborates on previous research that has explored the topic of digital marketing already (Grewal et al., 2016; Hofacker, de Ruyter, Lurie, Manchanda & Donaldson, 2016; Kannan & Li, 2017).

All in all, the topic of mobile versus desktop in digital marketing will be centric for this research. Within this study, several digital marketing campaigns that are targeted on both desktop and mobile users will be evaluated. The real-life campaigns are set up by different companies in the Netherlands. In final, the central research question of this study is: To what extent is device performance of digital marketing campaigns influenced by channel strategy and advertising phases? By conducting a statistical analysis on 700 digital marketing campaigns that have been executed over the past two years, the aim is to answer the central research question and discover

(12)

the relationships among the variables channel strategy, advertising phases and device performance.

The research will be structured as followed: In the next section, previous literature regarding digital marketing and specifically mobile and desktop devices will be discussed. In subsequent sections, the conceptual framework and methodology of this research will be described. Afterwards, the steps of all statistical analysis are explained and the results will be displayed and reviewed. The discussion will summarize all findings and include both theoretical as managerial contributions. Moreover, the limitations of this study, the recommendations for future research will be discussed. In the final chapter, the conclusion of this research will be presented.

(13)

2. Literature review

1. Research on mobile marketing

In the current landscape of digital marketing, mobile is no longer a topic that companies might consider. Unquestionably, mobile marketing plays a significant role in a customer’s user experience. Mobile marketing offers a great opportunity for companies as it allows them to communicate with their customers directly without time or location barriers (Haghirian, Madlberger & Tanuskova, 2005). Therefore, marketers have been highlighting the importance of mobile marketing and will continue to do so in the coming years.

1.1 Past insights

1.1.1 The start of mobile marketing

Over the past years, an increasing number of articles on mobile marketing have been published (Varnali, Toker, 2010). Main areas of focus are user behaviour and attitudes towards mobile marketing (Leppäniemi, Sinisalo & Karjaluoto, 2006; Shankar, Venkatesh, Hofacker & Naik, 2010). In addition, Leppäniemi et al. (2006) found that there is an increasing interest for mobile marketing effectiveness and fewer studies focus on the role of mobile marketing in branding, in the value chain and mobile marketing business models. Next, Leppäniemi and Karjaluoto (2008) highlighted that marketing strategies should include a planning and implementation of mobile strategies for businesses in their overall marketing strategy. According to Leppäniemi et al. (2006), several key areas in mobile marketing calling for further research are: Integrated marketing communications, mobile marketing value system, trust, mobile marketing technology

(14)

and mobile-specific metrics (Leppäniemi et al., 2006, p. 10). This final area of mobile-specific metrics will be discussed in more detail in part 2.3 of the literature review as the topic will be centric for this research.

1.1.2 Continued research on mobile marketing

In 2005, researchers Scharl, Dickinger and Murphy classified short message services (SMS) as the most successful form of mobile communication. Scharl et al. (2005) identified three dependent success measures: Consumer attention, consumer intention and consumer behaviour. Following Scharl et al. (2005), several researchers highlighted the importance of permission-based advertising via mobile phones, which is considered as one of the first techniques of mobile marketing (Barwise & Strong, 2002; Tsang, Ho & Liang, 2004; Varnali et al., 2010). In particular, researchers Tsang et al. (2004) highlighted the success of SMS, whilst shortly after this research was published in 2006, SMS has disappeared almost completely (Shankar & Balasubramanian, 2009). The success measures of mobile marketing identified by Scharl et al. (2005) are nowadays replaced by metrics such as ‘viewability’, ‘active page dwell’, ‘interaction rate’, and ‘total exposure time’ (Google, 2013).

1.1.3 Growth in the market for mobile marketing

Several researchers expressed their belief that the market for mobile would grow significantly. One of their expectations was that mobile ad spending on mobile advertising in the U.S. that was at approximately $644 million in 2007, would rise to over $3.5 billion by 2011 (eMarketer 2007). Next to the growth in the mobile market, mobile ad spend was expected to account for 72% of U.S. digital ad spend by 2019 (eMarketer 2015). In 2014, the number of global users on

(15)

mobile has outgrown the number of global users on desktop devices (ComScore, 2014). All in all, the market for mobile is growing significantly and marketers see a shift in advertising spend by companies towards mobile.

1.2 Current state of knowledge

According to Kaplan (2012, p. 130), the definition of mobile marketing is “Any marketing activity conducted through an ubiquitous network to which customers are constantly connected using a personal mobile device”. Researchers Shankar, Kleijnen, Ramanthan, Rizley, Holland & Morrissey (2016, p. 37-38) provided an overview of research focus on mobile marketing in practice thus far, which include the areas: “the scope of mobile marketing (Shankar et al., 2009), mobile service delivery (Kleijnen, De Ruyter & Wetzels 2007), mobile interface usage and usability (Venkatesh, Thong & Xu 2012), mobile browsing experience (Adipat, Zhang & Zhou 2011), applications to retailing (Shankar et al. 2010), interfaces for mobile devices (Brasel and Gips 2014), mobile app demand (Garg and Telang 2013), mobile shopping carts (Van Ittersum et al. 2013), mobile advertising and promotions (Andrews, Luo, Fang & Ghose, 2015; Bart, Stephen, and Sarvary 2014; Fong, Fang, and Luo 2015), and mobile shopping (Wang et al., 2015)”.

According to a report by ComScore (2017), mobile devices dominate total minutes spent online. Marketers no longer underestimate the area of mobile marketing and customers have accepted mobile marketing in sharp contrast to previous customer behaviour towards mobile marketing (Tsang et al., 2004; Bauer, Reichardt, Barnes & Neumann, 2005; Grewal et al., 2016). All in all, both researchers and marketers no longer neglect the concept of mobile marketing but highlight the important role that mobile is playing.

(16)

1.3 Future of mobile marketing

To identify gaps within research on mobile marketing, it is important to investigate what has been done so far, what is the current understanding of this topic, but also: Where is current research on mobile marketing going (Shankar, 2016)?

Mobile marketing has grown substantially in scope from its first definition in the early 2000s, but also in its development from permission-based advertising such as SMS to customers receiving push-up notifications on their mobile devices when entering a physical store (Aalto, Göthlin, Korhonen & Ojala, 2004; Unni & Harmon, 2007; Shankar et al., 2009). Next, mobile marketing has had a significant impact on the current business environment (Shankar, 2016). The important features of mobile devices such as the level of individualisation and personalisation, the ability to support location-based applications, but also their dynamic options for retargeting and delivering relevant content or promotions to customers, displays how mobile strategies interact with or complement overall marketing strategies (Grewal et al., 2016; Shankar, 2016). Mobile marketing is positioned by marketers as a high potential area to reach more customers better and more efficient, with higher levels of engagement and an increase in overall marketing effectiveness (Grewal et al., 2016; Hofacker et al., 2016; Shankar et al., 2016; Berman, 2017). Overall, Shankar (2016) provides key insights in his research on the potential future for mobile marketing. First, both researchers Shankar (2016) and Hofacker et al. (2016) highlight the rise of gamification together with customers and product factors as the key determinants of marketing outcomes such as engagement, purchase and retention. Second, mobile marketing is expected to influence the shopper behaviour of customers increasingly (Kourouthanassis & Giaglis, 2012; Shankar, 2016; Shankar et al., 2016; Xu & Whinston, 2014). Finally, behavioural outcomes

(17)

depend on a complex amount of factors and several researchers highlight the importance of measuring these outcomes together effectively (Shankar, 2016; Grewal et al., 2016).

The effectiveness of marketing promotions and digital marketing campaigns on mobile devices, in general, are influenced by many variables (Shankar et al., 2016). In addition to this, Grewal et al. (2016) highlight that a research gap exists on what factors determine the effectiveness of mobile advertising. Therefore, this study will focus on the effectiveness of digital marketing campaigns whilst comparing device behaviour and investigate the concept of mobile marketing in more detail.

2. Digital marketing campaigns 2.1 Digital marketing framework

Researchers Kannan and Li (2017, p. 41) define digital marketing in “the broadest sense” and have developed a framework in which they identify the touchpoints that play a role throughout the marketing process as well as the process of marketing strategies where digital technologies play an important role. Next, they identify unresolved questions in each area of the framework as guidance for future research. Kannan et al. (2017) argue that within digital marketing, mobile devices are becoming more important in determining a customer’s path to purchase and more research is focusing on all aspects of mobile devices as a channel. Not only is research necessary to understand the contribution of mobile marketing to final marketing outcomes, but also designing media and building mobile marketing strategies to optimize their effectiveness will be an important topic for future research and point of focus for marketers in the coming years (Kannan et al., 2017).

(18)

2.2 Channels in digital marketing

In an earlier research by Li et al. (2014), they identify two types of channels, namely customer-initiated and firm-initiated channels. These can be further divided into search (organic and paid), direct and referral for the customer-initiated channel, whilst firm-initiated channels are divided into display and e-mail (Li et al., 2014). In addition, Google (2018) adds to this framework the channel ‘affiliate’ and makes a distinction between referral and social channels. The channel division of Google (2018) is used by most marketers nowadays as over half of all websites are using the program Google Analytics (Forbes, 2017). For this study, however, the channels search, social, and display will be included as they are initiated by the firm and can be controlled by managers when deciding on the channel strategy. In other words, managers set the budget per channel, which leads to a certain performance of a campaign. This research will review the strategy of the budget division per channel and intends to find what channels are working best on certain devices.

2.3. Digital marketing metrics

When evaluating digital marketing campaigns, marketers are looking at a set of metrics to define the performance (Kireyev et al., 2014). Metrics are necessary for developing a digital marketing campaign as the metrics are linked to the campaign’s goals and besides, the metrics are used for evaluation of the campaign’s performance (Kitchen, Kim, Schultz, 2008). When developing campaigns, the objectives of a campaign are translated into digital goals (Forbes, 2014). KPIs (Key Performance Indicators) are these quantifiable goals to track and measure a campaigns’ performance (Forbes, 2017a). Not all metrics will be in the scope of this research and therefore,

(19)

the main metrics in digital marketing which are important for this study will be summarized below.

2.3.1 Impressions

Each time an advertisement is viewed by a customer, this is called an impression. According to Google (2018), an impression is counted each time your advertisement is shown on a search result page or website.

2.3.2 Clicks

When a customer clicks on a link/advertisement, this is measured and reported as a customer who did not just view an advertisement, but a customer who acted on the advertisement by clicking on the ad (Barger and Labrecque, 2013). This action by a customer is called a ‘click’. 2.3.3 Click Through Rate

Click Through Rates (CTR) measure how many people who viewed your advertisement, eventually clicked on your ad (Kireyev et al., 2014). The CTR can be calculated by dividing the number of clicks by total amount of impressions. A high CTR might indicate that a customer is interested in a company’s advertisement (Google, 2018). Nevertheless, CTR is sometimes criticized for having an attribution problem, which will be discussed in part 2.4 of this literature review (Kireyev et al., 2014). CTR is measured in this study and evaluated as one of the two outcome metrics that define a digital marketing campaign’s performance.

(20)

2.3.4 Pricing models

Several pricing models exist: CPI, CPM, CPT, CPC, CPL and CPA (Berman, 2017; Kireyev et al., 2014). An overview of what each pricing model stands for:

CPI: Cost Per Impression - pay for every impression. CPM: Cost Per Mille - pay for every 1,000 views. CPT: Cost Per Time - pay per time unit, e.g. day. CPC: Cost Per Click - pay if a customer clicks the ad

CPA: Cost Per Acquisition - pay if one clicks the ad and then continues to a conversion Pricing models are a strategy that managers can set for their digital marketing campaigns. All pricing models are performance-based metrics. When a manager would choose the CPC model, costs will only be charged when a customer clicks on an advertisement. In contrast, if a CPM model is introduced, a manager would pay for a 1,000 impressions viewed to a customer. Pricing models are often connected to the goal of a campaign. For example, when the aim is to spread awareness, managers want to exploit as many advertisements as possible and therefore they would adopt a CPI or CPM strategy. On the other hand, when the goal would be to purchase an item or any other form of conversion, the pricing model CPA would be best suitable.

2.3.5 Engagement

Although the correct definition of engagement is highly debatable according to Barger, Peltier and Schultz (2016), most often engagement is referred to as a customer “taking some action beyond viewing or reading” (Paine 2011, p. 60). In a Social Media context, for example, this could be ‘liking’ or ‘commenting’ to a post. According to Barger et al. (2013, p.26) engagement

(21)

stands for “the number of comments on, replies to, likes of, and shares of a given post”. In a research done by Huizenga (2009), higher engagement showed a significant positive effect on a user’s experience.

Engagement is often discussed when setting targeting strategies. In this research, targeting methods are being investigated in digital marketing campaigns. Therefore, it is important to consider whether highly engaged customers are being targeted or a broad targeting strategy is applied.

2.3.6 Conversion Rate

Whether your goal is to gather valuable information about your website visitors and potential customers or convert sites visits into sales, monitoring your Conversion Rate could potentially define digital marketing success (Google, 2018). A Conversion rate (CR) is the percentage of visitors to a website that complete the desired goal (a conversion) out of the total number of visitors (Berman, 2017, Grewal et al., 2016, Xu et al., 2014). Conversion Rate is the second outcome metric that is being investigated in this study.

2.3.7 Bounce rate

Some visitors immediately leave or “bounce” should they find the content of a website irrelevant to their needs. Often, landing page experiences are below average for customers which is sometimes shown by a high bounce rate. A bounce rate shows the percentage of visitors to a particular website who navigate away from the site after viewing only one page (Google, 2018). Bounce rates can be calculated by the total number of visits viewing one page only divided by the total entries to the page.

(22)

2.3.8 Return On Investment

Identifying what area in a digital marketing campaign is driving sales and revenue and what area could potentially increase its sales, is of high importance for managers (Kireyev et al., 2016). Return On Investment (ROI) is calculated as the revenue gained from a digital marketing campaign minus the cost of the campaign, divided by the cost of the campaign (Barger et al., 2013). ROI is used by managers to measure success over time and is often considered when making future business decisions (Rust, Lemon, Zeithaml, 2004).

2.4 Attribution (problem)

A fully integrated digital marketing campaign includes a balance across all channels that are used in the campaign (Kannan et al., 2017). Once a campaign has ended, marketers start analysing campaigns by using some (or all) of the metrics stated above (Berman, 2017, Grewal et al., 2016). As marketers want to quantify the value of each channel (e.g. search, social, display), several issues arise when trying to analyse all channels (Barger et al., 2013). Marketers are faced with a question of attribution; that is, “deciding which channel should get credit for the conversion” (Barger et al., 2013, p.17). Several years ago, marketers started analysing all channels using the last-click metric which depicts that the final channel before a customer continues to a conversion is the most responsible in the entire process (Li et al., 2014). But Li et al. (2014) were not the only researchers to also claim that the last-click metric is not an appropriate metric for understanding the real impact of firm-initiated channels as well as customer-initiated channels on conversions. In reality, Berman (2017) highlights that customers are likely to encounter multiple touchpoints across channels on their user journey and all of the

(23)

touchpoints have a certain contribution to the final conversion. Nevertheless, calculating this attribution per touchpoint/channel still remains to be an issue for marketers and still has not been solved yet (Barger et al., 2013; Berman, 2017; Grewal et al., 2016; Kannan et al., 2017; Shankar et al., 2016).

3. Mobile vs. desktop

At first, when digital marketing was becoming increasingly important for marketers, no distinction was made between different devices (Book & Wallach, 2015; Ong, Järvelin, Sanderson & Scholer, 2017; Güler, Kılıç & Çavuş, 2014 ). Nevertheless, several researchers including Ong et al. (2017) and Güler et al. (2014) found that a substantial difference between mobile and desktop devices exists. In addition, Baker, Fang and Luo (2014) highlight in their research when mobile ads are most effective, introducing time as a determining factor for the effectiveness of a mobile advertisement. This provides companies the option to engage in mobile targeting strategies by the hour to increase its effectiveness (Baker et al., 2014). Overall, many differences between desktop and mobile exist, which will be identified in this section.

3.1 Cross-device effects

Ghose, Han & Park (2013) find that 59% of customers sometimes visit a website on mobile and then follow up on desktop. In contrast, 34% of customers visit a site on a desktop device and then continue on a mobile device. These ‘cross-device effects’ show that mobile as a channel is a starting point of a customer journey for a lot of customers. Nevertheless, recent research has identified that more customer journeys are mobile-only or end on mobile when they have started on a desktop device (Metrixlab, 2017).

(24)

According to Google (2013), 48% of customers start their path to purchase on search engines on mobile devices. Next to that, if a conversion has been made, chances are that mobile has played a substantial role in this as 93% of people who used mobile to research go on to make a purchase (conversion) (Google, 2013). Most customers convert primarily in store (82% purchases in store) whilst 45% of customers purchases online (on desktop devices). In contrast, 17% of customers purchase directly on their mobile device (Google, 2013). Criteo (2015) argues in his research that understanding cross-device behaviour will be the biggest challenge as the majority of customers visit their websites via multiple devices, but perhaps this could also be an opportunity for managers.

3.2. Conversions on mobile-only, desktop-only or combined campaigns

According to the traditional digital marketing analysis, impressions of mobile ads should lead to clicks, which subsequently lead to conversions (Ghose et al., 2013). Nevertheless, Ghose et al. (2013) found that mobile ads not only influence conversions on mobile, but also conversions on desktop. See figure 3 in Appendix for the model by Ghose et al. (2013). Also, in their research they find that Click-through rates are 34% higher when advertising is done on both mobile and desktop (in comparison to desktop only). In addition, Conversion Rates are 36% higher when both desktop and mobile advertising are used (again in comparison to desktop only). See figure 4 and 4a in Appendix for full results of this study. In conclusion, total clicks and total sales are higher when mobile and desktop are used together instead of mobile- or desktop-only campaigns (Ghose et al., 2013).

Similar to addressing the attribution problem in section 2.4, a problem arises for marketers when analysing the influence of mobile when a conversion is done on a desktop, and controversially,

(25)

the influence of desktop when a conversion is done on mobile (Berman, 2017; Grewal et al., 2016). This research will focus specifically on the influence of different devices on a campaign’s performance, measured in CTR and CR.

3.3 Difference between mobile and desktop

When comparing mobile and desktop from a customer point-of-view, significantly different behaviour can be distinguished (Berman, 2016). A comparison of two samples from the UK and the US in a report of Google Consumer Barometer (2015) is used in an attempt to generalize these contributions to Western cultures and highlight the key differences between the devices (Google, 2015).

3.3.1 Objectives for using mobile vs desktop

When looking at the purchase process, customers use their mobile devices for the following activities (Google, 2015): Searching for early inspiration and making initial discoveries online (45% of total sample of nearly 2000 respondents), comparing choices online, seeking advice online (reviews), and preparing online for immediate offline purchase (search for locations etc). All show events before the actual conversion takes place. What customers do at least weekly on their mobile devices: Use search engines, visit social networks, watch videos online, and look for product information (Google, 2015). Only 20% of respondents mentioned purchasing products or services on their mobile devices. Again, this highlights that the objectives for mobile are mostly for awareness and consideration stages of a customer journey (Grewal et al., 2016).

When asking respondents what type of device they use to make their purchases, 81% (of over 4500 respondents from the US and the UK) answers that they use a desktop device to make a

(26)

purchase in sharp contrast to a percentage of 8.5% on mobile (Google, 2015). All in all, the conclusion can be drawn that mobile and desktop devices are used for different purposes by customers.

3.3.2 User journey of desktop and mobile

Using the classical customer journey model (see figure 5 in Appendix), mobile is mostly used for the first two stages of the model, namely awareness and consideration (Grewal et al., 2016). Desktop devices may be used for the first two phases as well, but when customers need to make a purchase, they use desktop far more often than mobile (Google, 2018). Mobile can play a significant role in the start of a customer journey, but final conversions happen on a desktop device or in a physical store most of the time. This is in correspondence with past research, which has also found that conversions on mobile are much lower than on desktop (Criteo, 2015). When analysing a mobile conversion funnel, three phases are distinguished. The first stage of the funnel is the number of products viewed, which is higher for desktop than mobile (3.2 products viewed per user in comparison to 3.0 products on mobile). The percentage of products ending up in a customer’s basket and eventually being bought is also higher for desktop than for mobile. See figure 6 in Appendix for a full report of this research by Criteo (2015). All in all, several issues exist regarding mobile marketing in measuring its effectiveness and explaining its role in the full conversion funnel (Grewal et al., 2016).

4. Mobile marketing strategy

Mobile is a major disruption across industries and a business that has been ‘booming’ over the past decade. Therefore, mobile has been identified as one of the fastest-growing advertising

(27)

formats (Bart et al., 2014; Grewal et al., 2016; WARC, 2017). Next, projections suggest that mobile marketing will account for three-quarters of all digital advertising spending by 2019 (eMarketer, 2015). According to Graham (2015) over one-half of all searches on Google are now performed on mobile sites and this amount is only expected to increase even further.

4.1 Benefits of mobile marketing

Mobile marketing can be a very important component of a firm’s overall marketing strategy (Luo, Andrews, Fang & Phang, 2014). ‘Can be’ as, despite the increased use and potential of mobile marketing, there is evidence of marketers struggling to find the right strategy (Berman, 2016). Several benefits of mobile marketing exist which will now be presented. First of all, mobile devices are always on, connected, and carried by customers (Baker et al., 2014, Ström, Vendel & Bredican, 2014 ). This offers the potential to reach the on-the-go customers who are difficult to reach through traditional online and offline methods (Baker et al., 2014; Berman, 2016; Book et al., 2015; Grewal et al., 2016). Second, mobile offers the potential for location-based data allowing marketers to target customers based on their location with f.e. price discounts for nearby shops (Andrews et al., 2015; Berman, 2016; Book et al., 2015; Grewal et al., 2016; Luo et al., 2013). Third, due to the highly personalized features of mobile devices, companies can send relevant personalized messages and offers to customers to get higher response rates and higher levels of engagement (Baek & Morimoto, 2012; Berman, 2016; Book et al., 2015). Also, personalized advertising provides the opportunity to apply one-to-one marketing communication for customers, targeting of “prospective audiences”, and the option to measure responses directly from digital marketing campaigns (Baek et al., 2012, p. 59). Finally, mobile marketing campaigns can reach a certain age or occupational group such as younger

(28)

people or people who are very busy at work as these might be more engaged with their mobile devices (Book et al., 2015; Persaud & Azhar, 2012).

4.2 Issues for mobile marketing

Whilst the market for mobile is growing a lot, Berman (2016) argues that marketers have failed to use mobile marketing effectively (Berman, 2016). According to Berman (2016, p. 437), “developing an effective mobile marketing program is much more challenging than developing a traditional program aimed at desktop users”. This can be seen by high bounce rates on mobile devices, low completion rates and poor average sales in comparison to campaigns on a desktop device (Berman, 2016). Also, Google (2015) highlights that mobile devices are used for different purposes than desktop, which explains the lower conversion rates on mobile devices.

A popular metric to assess the effectiveness of mobile marketing is measuring the Click-Through Rate (Berman, 2016). Nevertheless, WARC (2017) finds the biggest barrier to the growth of mobile is measurement and metrics. WARC (2017) finds that mobile marketing effectiveness is measured through engagement, not through ROI and work needs to be done to track behavioural metrics, audience delivery metrics, but most importantly: business metrics. Mobile has established itself in a primary position worldwide, where half of all digital minutes are spent on mobile (ComScore, 2017). Thus, mobile is “quite important” but additional evidence about its effectiveness is needed in order to create more urgency (WARC, 2017). In this research, no distinction will be made between the behaviour of customers on mobile Apps versus mobile web behaviour as the behaviour of customers on mobile will be measured collaboratively.

(29)

5. Conclusions

It is a fact that digital marketing has been evolving significantly throughout time. Research thus far has also shown that money spent on digital marketing campaigns will only increase over the next decade. Hence, the world of digital media that offers options offline media cannot, opens a new world for marketers and a growing need for insights on how to efficiently set up digital marketing campaigns. There is a clear need present to identify the different objectives for customers to use their mobile versus desktop device and the difference in user experience on each device. This research, therefore, aims to close the gap of efficiently using several devices in digital marketing campaigns by focusing on channel strategy and advertising phases. Thus, the central research question for this study will be: To what extent is device performance of digital marketing campaigns influenced by channel strategy and advertising phases? In the next chapter, the methodology of this research will be explained in more detail.

(30)

3. Conceptual framework

This section will describe the conceptual framework that is investigated in this research. First, the research design of this study will be presented. Subsequently, the research method and the variables that are included in the analysis will be defined. Also, the three analyses that will be tested via SPSS and the hypotheses will be displayed.

1. Research design

To explore the difference between mobile and desktop performance in digital marketing campaigns, this study contains a large amount of data. The research will be a quantitative analysis of a number of digital marketing campaigns of several companies in the Netherlands. The analysis of the campaigns will be done by examining the relationship between the goals of a campaign, the channel strategy set per campaign, and the device performance on both mobile and desktop. When analysing the campaigns, the aim is to identify to what extent device performance is influenced by the variables ‘channel strategy’ and ‘advertising goals’. Therefore, a field study is designed to test the hypothesis in an explanatory manner. This explanatory research attempts to establish causal relationships between the variables by performing a quantitative data analysis by means of two regression analyses and a mediation analysis via PROCESS (Hayes, 2017). 2. Research method and variables

The research is an online experiment in a natural setting. Campaigns were assigned to the conditions of a 3 (channel strategy: Search, Social and Display) x 3 (advertising goals:

(31)

Awareness, Consideration and Purchase) design. The variables that are included in the analyses are now presented. Overall, fourteen variables are being tested.

2.1 Channel Strategy

The first independent variable in this research is the_{Channel Strategy}

. This includes how much

of the campaign’s budget will be used for the channels Search, Social and Display. The variable is expressed in percentages where all three channels add up to 1 (or 100%). As the variable channel strategy includes the total budget spent on a campaign, the variable is also split up into the budget division per channel. Thus, besides the variable _{Channel Strategy}

, the variables

Budget Search, Budget Social

and _{Budget Display} are being included in the regression analysis.

2.2 Length of Phases

The advertising goals as they have been identified by Grewal et al. (2016) will be followed in this research. The six ‘Ad goals’ by Grewal et al. (2016, p. 4) include awareness, engagement, purchase intent, conversion, repurchase and advocacy. The stages of Grewal’s framework are summarized into three phases in this research, namely the phases: Awareness, consideration and purchase. This is due to the fact that the data is collected via a media firm who builds up the campaigns in this three-phase structure.

Thus, the second independent variable in this research is the _{Length of Phases}

. This variable

includes the total length of the digital marketing campaign, expressed in days. Subsequently, the variable is split up into the length of each advertising phase, meaning how many days the Awareness, Consideration, and Purchase phase have taken. The variables _{Length Awareness,} Length Consideration and _{Length Purchase are investigated in the regression analysis. The}

(32)

variable is expressed in days but will be transformed in percentages where the three phases together add up to 1 (or 100%). In addition, the variable _{Length of Phases will be tested as an} independent as well as a mediator variable in this research. This will be explained in more detail in section 3 on the Conceptual Framework.

2.3 Device Performance

Third, _{Device Performance on mobile and desktop will be the dependent variables in this} research. Mobile and desktop performance, measured by CTRs and CRs (see part 2.3 in the literature review), will show the effect of the length of phases and the channel strategy on mobile and desktop devices. All in all, the four dependent variables are _{Mobile CTR, Mobile CR,} Desktop CTR and Desktop CR

. The variables are expressed in percentages as CTRs and CRs are

usually also expressed in this way. 2.4 Control variables

Finally, several control variables exist: First of all, the _{Type of Industry will be included in this} research. Several companies are included in the sample set from two types of industries. As this could potentially influence the dependent variables, this variable will be controlled in the research. Therefore, the distinction in the sample set is made between companies in Commercial and Governmental industries.

Second, the _{Targeting Method will be controlled in this research as it could also influence the} dependent variables. For example, a campaign with narrow targeting and therefore a smaller targeting audience could have an effect on the dependent variables. By targeting the right customer in a campaign, hence customers with higher levels of engagement, the performance

(33)

metrics are expected to be higher in comparison to broadly targeted campaigns. The control variables are nominal and will be coded as a dummy variable (Lomax, 2007).

3. Conceptual framework

To analyse the relationship between the variables _{Length of Phases}

of a campaign, _Channel

Strategy

, and_{Device Performance}, three analyses are distinguished before the actual analysis can

take place. In the regression, interactions between independent and dependent variables will be analysed to answer the central research question. The conceptual framework is divided into three analyses. The first analysis measures the direct effect between independent variable Channel Strategy and dependent variables Mobile Performance and Desktop Performance. The second analysis measures the direct effect between independent variable Length of Phases and dependent variables Mobile Performance and Desktop Performance. Finally, a mediation analysis is being executed which measures the indirect relationship of Length of Phases on Channel Strategy and Device Performance. In all three analysed, the control variables are added. Figure 1 below includes the framework that will be investigated.

(34)

Figure 1: Conceptual framework

Based on the central research question, the following research hypotheses are proposed for further analysis:

H1: Channel Strategy has a positive effect on Device Performance

H1a: Channel Strategy has a positive effect on Mobile Performance H1b: Channel Strategy has a positive effect on Desktop Performance H2: Length of Phases has a positive effect on Device Performance

H2a: Length of Phases has a positive effect on Mobile Performance H2b: Length of Phases has a positive effect on Desktop Performance

(35)

H3: The relationship between Channel Strategy and Device Performance is partly or completely mediated by Length of Phases

H3a: The relationship between Channel Strategy and Mobile Performance is partly or completely mediated by Length of Phases

H3b: The relationship between Channel Strategy and Desktop Performance is partly or completely mediated by Length of Phases

The next chapter will discuss the method of this research in more detail and show how the hypotheses will answer the central research question. Also, the sample and data collection will be described.

(36)

4. Methodology

This chapter presents the research methodology employed in the analysis of data on digital marketing campaigns, with the purpose of testing the proposed research hypotheses. This section addresses, first of all, what type of companies are in the sample set and what type of data is collected. Second, information will be given on how the data of these companies will be analysed and how this should answer the central research question.

1. Population and sample

The population of interest includes all companies in the Netherlands, who are active in online marketing. The sample will consist of a mix of Dutch companies. Overall, the sample set includes 700 campaigns which will be analysed in this research. All companies in the sample set will be coded due to privacy restrictions. Due to data restrictions, only companies in the Netherlands can be analysed as no data on other countries was available. Nevertheless, in all countries worldwide these type of companies will exist and, therefore, this research aims to generalize the findings to the wider population of companies in developed countries who are active in digital marketing. Also, the prognosis is that there will be no significant differences between countries in the Netherlands and other developed countries but further research could investigate this. Whilst potentially the type of industry of a certain digital marketing campaign could influence the model, it is expected that this will not have any influence on the research. Nevertheless, the type of industry, as well as the targeting method of the campaigns, are included in the research as control variables. Also, both B2B and B2C companies are included in the

(37)

sample set and no distinction will be made between these types of companies (Grewal et al., 2016).

2. Data collection

The data is collected via a media firm in the Netherlands which gained access to all digital marketing campaigns that have been executed in the past two years. The campaigns that have been selected contain all three phases of the advertising goals (Awareness, Consideration and Purchase) and are active on all channels (Search, Social and Display). Every campaign contains all three advertising goals which are launched respectively so a campaign will start with spreading awareness, then continue to make customer consider, and finally stimulate purchases. Next, campaign budget is divided over the three channels and together the budgets add up to 100%. In addition, all campaigns in the sample have been active on both desktop and mobile devices. Potentially, there are also desktop-only or mobile-only campaigns, but these will not be included in this research as it is the aim of this research to compare the performance of the two devices. The campaigns that have been selected contain a budget of over 7,500.00 Euros as this assures that the campaign has a significant impact and therefore relevant data can be drawn. Also, this budget restriction is necessary to assure a certain amount of budget is available per channel so significant impact can be measured when evaluating the device performance.

To make sure a short length of a campaign does not have a significant impact on the performance, total campaign length must be at least two weeks. All budget will be used within the time frame of the campaign and how the campaign has performed will be measured accordingly. The variation in campaign length and budget allows for the collection of several data points per campaign and increases the external validity of this research. For each campaign

(38)

in the sample set, information on the number of impressions, clicks, conversions, costs, length of time, type of industry and targeting method will be collected. Campaigns are monitored on a daily basis and will be analysed on their full performance after the entire campaign is finished. 3. Data analysis

This next section represents the different steps taken in order to analyse the data. The analytical procedure, as well as the results of the analyses, are described in detail.

For this research, SPSS version 24 has been used to perform the statistical analysis with the additional feature of PROCESS (Hayes, 2017). Before the regression and mediation analyses were executed, certain adjustments have been done. First of all, the dataset was checked for outliers. These outliers with extremely high or extremely low values have not been excluded from the sample as the outliers include campaigns with extremely high or very low performance and are, therefore, relevant to include in the sample. Nevertheless, the outliers have been identified and were taken into consideration when executing the analyses.

Second, the regression analyses can only be done when certain assumptions have been met, namely the assumptions of linearity, collinearity, independence, homoscedasticity and normality (Osborne & Waters, 2002). The normal distribution of the residuals has been tested by the Kolmogorov-Smirnov test and the Shapiro-Wilk test. Besides that, Q-Q plots and P-P plots have been analysed for the dependent variables. Finally, histograms and values of Skewness and Kurtosis have been taken into consideration. All in all, the normality of distribution could be confirmed. Next to that, when following the Central Limit Theorem, this sample set consists of enough observations so the sample distribution should be normal when the sample size is big (Field, 2013). Nevertheless, when checking for the normality of the residuals, the results showed

(39)

that transformations of the dependent variables using natural logarithms (log10) increased the normality of the distribution even further. Relevant statistical tables to assure the assumptions are met are included in the Appendix in tables 21 to 27.

Table 1 in the Results section of this research provides the descriptive statistics of the variables in this research, consisting of means, standard deviations, correlations and Cronbach’s alpha. This final statistic assures the internal validity of this research by testing the reliability of the analysis (Field, 2013). In final, regression analyses were executed to test hypothesis 1 and 2. Next, a mediation analysis was executed to test hypothesis 3 using model 4 of PROCESS. To be more specific, regression analyses have been used to examine the effect between independent variables _{Channel strategy}

and _{Length of Phases and dependent variables Mobile Performance}

and _{Desktop Performance}

. Next, the mediation effect of _{Length of Phases is tested by means of a}

PROCESS analysis. 4. Predictions

In this research, a model is created to measure the overall performance of digital marketing campaigns. By analysing the strategy set per channel, as well as the length of each advertising phase, the performance on mobile and desktop is tested. The aim of the research is to identify linear combinations of the independent variables to predict in the best way the value of the dependent variables.

Two factors that are expected to have an influence on the outcome metrics (CTR and CR) are being analysed. The strategy set per channel is expected to have a big influence on the overall performance of a campaign (hypothesis 1). Therefore, the regression analysis examines whether an increase in the budget of a certain channel has a significant impact on a campaign’s

(40)

performance. Next, the length of phases is expected to influence this relationship between the channel strategy and the performance, which is being measured by the mediation analysis (hypothesis 3). Also, it is examined whether there is a linear relationship between the length of phases and performance by means of a regression analysis (hypothesis 2). This allows the study to test the three advertising phases in detail and their effect on the device performance. All in all, the aim of this research is to forecast the performance of digital marketing campaigns by setting the best channel strategy and deciding on the best timings per advertising phase.

As digital marketing has been booming and will continue to do so, the need has arisen for more insights on how digital marketing is used in practice. The focus of this paper, therefore, is on how online marketers can design their strategy around digital marketing campaigns in such a way that they include the advertising phases of each campaign, the budget division per channel, and find a suitable strategy in terms of targeting devices.

(41)

5. Results

This next section presents the analytical procedure, as well as the results of the analyses. First of all, the descriptive statistics of the data are presented. Second, a raw data analysis of the average device performance scores is discussed. Subsequently, two statistical analyses are presented. 1. Descriptive statistics

The total sample consists of 700 observations. The dataset consists of no missing values and next to that, no coding has been necessary for the dependent and independent variables. Dummy variables have been created for the control variables to enable them to be added in the regression. Table 1 given below describes Pearson’s correlation coefficient as well as standard deviations and means of all variables. A bivariate correlation analysis was performed to assess the predictive accuracy of the conceptual model and its variables. From the table below, several assumptions can already be made regarding relationships between certain variables. As mentioned before, the dependent variables have been transformed using natural logarithms (log10). This is due to the fact that the variables must meet the assumption for the statistical analysis (Field, 2013). Through statistical tests and visual inspections of several graphs, the transformation of the variables confirmed that assumptions for the statistical analysis were met. Next to the descriptive statistics of the variables, skewness, kurtosis and normality tests of all items were executed which are presented in the Appendix in Table 21 to 27 (Tabachnick & Fidell, 2001). The tables include VIF values, correlation table for multicollinearity, Durbin-Watson tests for independence and several graphs such as p-plots, histograms for normality, scatter plots for linearity and homoscedasticity.

(42)

Table 1: Means, Standard Deviations, Correlations, Cronbach’s alpha Variables M SD 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1. Channel strategy 4.85 0.34 (0.77) 2. Budget Search 4.15 0.43 0.78**_(0.78) 3. Budget Social 4.36 0.36 0.94** 0.85** (0.77) 4. Budget Display 4.43 0.41 0.83** 0.69** 0.62** (0.80) 5. Length of Phases 1.69 0.34 0.43* 0.51 0.60 0.42 (0.77) 6. Lentgh Awareness 1.37 0.40 0.10 0.04 0.06 0.07 0.87** (0.80) 7. Length Consideration 0.99 0.38 0.08 0.08 0.08 0.07 0.94** 0.69** (0.67) 8. Length Purchase 0.95 0.54 0.11 0.15 0.14 0.03 0.61**_0.45**_0.72**_(0.80) 9. Mobile CTR 1.01 0.28 0.41**_0.27**_-0.38**_-0.28**_0.20**_-0.18*_0.15*_-0.15*_(0.85) 10. Mobile CR 1.40 0.19 0.15**_-0.19**_0.16**_-0.24**_0.31**_0.23*_-0.26*_0.27*_{0.02 (0.84)} 11. Desktop CTR 0.96 0.27 0.57**_-0.49**_0.51**_-0.48**_0.22**_-0.19*_0.10*_-0.17*_{0.02 0.04 (0.85)} 12. Desktop CR 1.43 0.19 0.31**_-0.21**_0.18**_-0.26**_0.12**_-0.15*_0.13*_-0.12*_{0.03 0.02 0.01 (0.84)} 13. Targeting Method 0.44 0.50 -0.05 -0.02 -0.03 -0.06 -0.16 -0.07 0.03 0.05 -0.04*_-0.03*_{0.02 0.02 (0.88)} 14. Type of Industry 0.52 0.50 0.10 0.08 0.10 0.08 0.05 0.04 -0.04 -0.04 0.01 0.01 -0.03 -0.03 -0.03 (0.87)

** Correlation is significant at the 0.01 level (2-tailed) * Correlation is significant at the 0.05 level (2-tailed) List wise n = 700. Cronbach’s alpha on diagonal

(43)

The SPSS output provided a table of correlation coefficients for all of the combinations of variables. First of all, independent variable Channel Strategy shows a significant correlation with Length of Phases, as well as with all dependent variables which provides initial support for H1. The table shows that the relationship between Channel strategy and Mobile performance is characterized by a positive relationship of r = 0.41 for Mobile CTR and r = 0.15 for Mobile CR. The correlation between Channel Strategy and Desktop performance gives a positive relationship of r = 0.57 for Desktop CTR and r = 0.31 for Desktop CR. Thus, initial support for hypothesis 1 is observed. In addition, the correlation between Channel Strategy and Length of Phases is defined by a relatively large positive effect, r = 0.43, p < .05. Therefore, initial support for hypothesis 3 is provided.

The variable Length of Phases shows a significant relationship with all dependent variables presenting initial support for H2. Length of Phases shows a positive relationship of r = 0.09, p < .20 and r = 0.31, p < .01 for Mobile CTR and CR respectively. Also, it displays a positive relationship of r = 0.22 and r = 0.12 for Desktop CTR and CR both for significance level p < .01. This will be taken into consideration when analysing the mediation effect in the next part.

The control variables in this research, Targeting Method and Type of Industry both show a significant relationship with some of the variables. First of all, Targeting Method shows a significant negative relationship with Mobile performance as the correlation between the variables is characterized by r = -0.04 for Mobile CTR and r = -0.03 for Mobile CR, p < .05. No other significant relationships are found for this variable. The control variable Type of Industry, on the contrary, shows no significant relationships. Before conclusions can be drawn on this, the

(44)

results of the regression and mediation analyses need to be examined first. The proposed effects are all tested and analysed in more depth in the next chapters.

Reliability checks test the consistency of the data in the sample. The Cronbach’s alpha, which is presented in table 1 on the diagonal, estimates the internal consistency of this study (Field, 2013). All fourteen variables in the table beside one consist of a Cronbach’s alpha > 0.7. The overall Cronbach’s alpha of 0.73 indicates that the sample consists of a high level of internal consistency with only one variable scoring below the recommended benchmark of 0.7 (Field, 2013).

2. Raw data analysis

Before starting with the regression and mediation analyses, some general analyses using Excel have been performed to look at the overall performance metrics and behaviour of customers across devices. The findings of this analysis are important to incorporate into the conclusion of this research and will be considered when answering the central research question. The following two chapters will discuss the overall Click-Through Rate and Conversion Rate across devices. 2.1 Click-Through Rate

When analysing the performance on mobile versus desktop, the data shows that over the entire sample set the CTR is higher on mobile than on a desktop with 1.22% vs. 1.08% respectively. If the CTR is analysed in more detail, the data shows some interesting differences between CTRs of the three channels. Table 2 below shows the average CTR per channel on mobile and on desktop devices.

(45)

Table 2: Click-Through Rates across devices

Mobile Desktop

Search Social Display Search Social Display

11.68% 0.77% 1.09% 13.16% 0.62% 0.88%

The results show that CTR of the Search channel is higher on desktop whilst for Social and Display channels, the CTR is higher on mobile devices. Also, the CTR is significantly higher on Search versus Social and Display. This can be explained when analysing advertisements across all channels. For the Search channel, the advertisements often do not stand out that much. Research has investigated the click-through behaviour of customers on Search and has found that customers only consider the first page of their search results. Besides that, there is a significant drop in the first result on the page and the second, third and fourth result (Chan, Yuan, Koehler & Kumar, 2011; Google, 2015). As advertisements on Search channels are placed on the very top of the search result page, this high CTR can be partly explained. Second, when evaluating Social channels, customers use the channel more often on Mobile devices. It is therefore interesting to see that advertising on Social Media is more effective on mobile devices as well. Finally, when looking at Display channels, the CTR is higher on mobile devices. Past research has shown that a reason for this high CTR could be that mobile devices have smaller screens and therefore, customers click easier on advertisements than on desktop devices (Le & Nguyen, 2014). Although this traffic from an advertisement causes high bounce rates and low Returns On Investment, it is out of scope for this research to investigate this is more detail. Therefore, future research could potentially build on this finding.

(46)

2.2 Conversion Rate

Besides the CTR, this research will also investigate the Conversion Rate of digital marketing campaigns (Table 3). Conversions coming from campaigns mostly follow during the end of a campaign as the final Purchase phase is mainly focused on generating conversions. The three channels of the digital marketing campaigns show different behaviour when analysing conversion rates in comparison to the click-through rates.

Table 3: Conversion rates across devices

Mobile Desktop

Search Social Display Search Social Display

7.74% 2.50% 1.28% 8.54% 2.04% 1.44%

Overall, the campaigns on mobile have a CR of 1.62% versus 1.74% on desktop. Although the difference is relatively small, it has been measured over 700 campaigns and, therefore, this displays a significant difference. Although desktop devices shows a higher Conversion Rate, it is interesting to see the differences between channels. The results show that CR of the Search and Display channel are higher on desktop whilst for the Social channel, the CR is higher on mobile devices. In addition, the Search channel shows the highest Conversion Rates overall.

Finally, when comparing CRs and CTRs, Search and Social channel behave in the same way across devices as Search performs higher on a desktop and Social performs higher on mobile. This can be explained when referring back to the literature regarding customer behaviour across devices. For example, Social channels (or social media) are being used more often on mobile devices and therefore, it is no surprise that CTRs and CRs are higher on mobile devices for the

Digital marketing campaigns: The effect of channel strategy and advertising phases on mobile and desktop device performance