IN
DEGREE PROJECT INDUSTRIAL ENGINEERING AND MANAGEMENT,
SECOND CYCLE, 30 CREDITS ,
STOCKHOLM SWEDEN 2020
Understanding how automatized
personalization with AI can drive
value in B2B marketing
A case study of a Swedish industrial equipment
manufacturer
ELIZABETH ANZÉN
LUKAS EKBERG
KTH ROYAL INSTITUTE OF TECHNOLOGY
1
Understanding how automatized
personalization with AI can drive value in
B2B marketing
by
Elizabeth Anzén
Lukas Ekberg
Master of Science Thesis TRITA-ITM-EX2020:138 KTH Industrial Engineering and Management
Industrial Management SE-100 44 STOCKHOLM
2
Hur automatiserad personalisering med AI
kan driva värdeskapande i
B2B-marknadsföring
by
Elizabeth Anzén
Lukas Ekberg
Examensarbete TRITA-ITM-EX2020:138 KTH Industriell teknik och management
Industriell ekonomi och organisation SE-100 44 STOCKHOLM
3
Master of Science Thesis TRITA-ITM-EX2020:138
Understanding how automatized
personalization with AI can drive value in B2B
marketing
Elizabeth Anzén Lukas Ekberg Approved 2020-06-04 Examiner Kristina Nyström Supervisor Mana Farshid Commissioner Atlas Copco Contact person Claire Geffroy AbstractIn the last decade, marketing automation, a tool for automatic personalization, has been gaining significant traction among marketing professionals. In parallel with the growing adoption trend, many marketing automation platform providers have been extending their offers to include AI features. However, there is a lack of research regarding how AI can enhance the process of marketing automation in a way that creates value, which is the studied topic in this thesis.
A qualitative and exploratory case study has been conducted in collaboration with the global B2B company Atlas Copco, a manufacturer of industrial equipment. Digital marketing practitioners were presented with two use cases of AI, segmentation and cross-selling, for personalization and asked about the marketing automation process and the expected impact on value.
The findings reveal what would be required in the marketing automation process for the use cases in terms of data needs, learning about customer insights, marketing output and evaluation. In our findings value creation strongly revolve around the value types: ‘excellence’, ‘efficiency’ and ‘privacy’. To conclude, AI will enable more advanced personalization and value creation can be substantial if customer sacrifices are addressed in an appropriate way. Depending on relevance, tone of voice, time and use of channel, different feelings of value are perceived, which are factors that AI can help to determine.
Key Words
4
Examensarbete TRITA-ITM-EX2020:138
Hur automatiserad personalisering med AI kan
driva värdeskapande i B2B-marknadsföring
Elizabeth Anzén Lukas Ekberg Godkänd 2020-06-04 Examinator Kristina Nyström Handledare Mana Farshid Uppdragsgivare Atlas Copco Kontaktperson Claire Geffroy Sammanfattning
Under det senaste årtiondet har verktyg för automatisk marknadsföring blivit populära bland
marknadsförare. Automatiska marknadsföringsplattformar fungerar som ett verktyg för att automtiskt leverera personaliserade marknadsföring. Många leverantörer av automatiska marknadsföringsplattformar har utökat sina erbjudanden till att innefatta tjänster. Den befintliga forskningen kring hur sådana AI-tjänster ska utnyttjas på ett sätt som skapar värde är begränsad och därav behandlas ämnet i den här uppsatsen.
En explorativ och kvalitativ fallstudie har genomförts i samarbete med Atlas Copco som är ett globalt b2b-företag. Vid varje intervju presenterades antingen merförsäljning eller kundsegmentering sedan ställdes frågor om den automatiska marknadsföringsprocess och värde. Resultaten indikerar vad som skulle krävas för de undersökta användningsfallen i den automatiska marknadsföringsprocessen samt att värdeskapande är starkt kopplat till värdetyperna excellens, effektivitet, privatliv och datasäkerhet.
Slutsatserna indikerar att AI kommer göra den personalisering som uppstår till följd av automatisk marknadsföring mer avancerad. Värdeskapandet från nya AI lösningar kan vara signifikant om implementeringen tar hänsyn och adresserar uppoffringar kunder behöver göra.
Nyckelord
5
Acknowledgements
We would like to thank all the members of the marketing and sales department at Atlas Copco who have been incredibly helpful during our thesis project. Without their contribution, in terms of time, resources and domain knowledge, this thesis would never have been finished.
Moreover, our supervisor Claire Geffroy at Atlas Copco has helped us significantly by providing us with great insights into the procedures in B2B and the latest trends in digital marketing among many other things. Finally, we would like to thank our supervisor Mana Farshid at KTH who has continuously challenged us to improve our thesis throughout the process.
6 Table of Content Acknowledgements ... 5 1. Introduction ... 9 1.1. Background ... 9 1.2. Problem Statement ... 11 1.3. Research Purpose ... 13 1.4. Research Question ... 13 1.5. Delimitations ... 13 1.6. Sustainability Aspects ... 13 1.7. Outline of Thesis ... 14 2. Literature Review... 15 2.1. Personalization ... 15 2.1.1. What is Personalization? ... 15 2.1.2. Levels of Personalization ... 17 2.2. Value ... 18 2.2.1. What is Value? ... 18
2.2.2. Value in Business Marketing as a Developing Concept ... 19
2.2.3. Value Typologies and Dimensions ... 22
2.2.4. Value in Personalization ... 27
2.3. Marketing Automation... 29
2.3.1. Marketing Automation and Personalization Processes ... 29
2.3.2. Conceptualization of the Marketing Automation Process ... 34
2.4. Artificial Intelligence ... 37
2.4.1. Big data and AI ... 37
2.4.2. How AI can Enhance Personalization ... 39
2.5. Frame of Reference... 42 3. Methodology ... 43 3.1. Research Purpose ... 43 3.2. Research Approach ... 43 3.3. Research Design ... 43 3.4. Research Strategy ... 44 3.5. Case Company ... 44 3.6. Data Collection ... 44 3.7. Research Instrument ... 45 3.8. Sample Selection ... 46
7 3.9. Data Analysis ... 47 3.10. Research Quality ... 48 3.11. Ethical Considerations ... 49 4. Empirical Analysis ... 50 4.1. Empirical Context ... 50
4.2. Findings & Analysis ... 52
4.2.1 Marketing Automation Process ... 52
4.2.2. Value Types ... 58
5. Conclusion ... 68
5.1. How Can AI be Used in the Marketing Automation Process to Create Value in a B2B Context?.. 68
5.2. Limitations and Suggestions for Future Research ... 69
References ... 71
Appendix ... 79
A. Value measurements ... 79
8
List of Figures
Figure 1. Sample model for perceived value as a second-order formative construct (Lin et al., 2005: 325). Get and Give components can be any number and include dimensions as price,
privacy and security. ... 24
Figure 2. A conceptual framework of personalization (Vesanen, 2007: 4147) ... 28
Figure 3. The personalization process by Vesanen and Raulas (2006) ... 31
Figure 4. Marketing automation framework adopted from (Heimbach et al., 2015: 131) ... 32
Figure 5. The marketing and sales funnel related to content marketing, marketing automation, and CRM adopted from (Järvinen and Taiminen, 2016: 170) ... 33
Figure 6. Process of automized personalization. ... 36
Figure 7. Frame of reference used in this study ... 42
List of Tables Table 1. Selection of definitions of personalization mentioned in marketing literature, (Montgomery and Smith, 2009:131; Vesanen, 2007: 412-413). ... 16
Table 2. The value dimensions in PERVAL (Sweeney and Soutar, 2001: 211). ... 23
Table 3. Total value proposition with value dimensions (Lapierre, 2000: 125) ... 23
Table 4. Revised value typology adapted from Holbrook (1996: 139-140) ... 26
Table 5. Comparison of personalization processes, adapted from Vesanen and Raulas (2006) .. 30
Table 6. AI framework (Davenport et al., 2020) ... 39
Table 7. Applications of AI in personalization. ... 41
Table 8. List of contextual interviews... 46
Table 9. List of title of interviewees in main research stage. ... 47
Table 10. Themes identified relating to the digital marketing process ... 52
9
1. Introduction
This chapter sets the scene by introducing the emergence of AI, marketing challenges in B2B and how personalization can help to facilitate these challenges to create value. The motivation for the research is laid out and thesis’s purpose is described along with the research question. Finally, delimitations, sustainability aspects and the structure of the thesis is presented.
1.1. Background
The rapid development of digital solutions, fueled by progress in digitization, information and communications technology (ICT) and artificial intelligence (AI), has sparked a belief that we are now entering a new epoch, referred to as the fourth industrial revolution. This revolution is believed to create a shift in decision-making, from human to machine (Chatterjee et al., 2019; Syam and Sharma, 2018). Traditionally, information technology has helped with the processing of data to enhance and support human decision-making. Now there are algorithms that process data, learn from data and use data to make well-informed decisions. A process that renders jobs both faster and easier, by being able to use more data than humans could ever dream of analyzing by themselves. The brain behind this process is AI; the concept where machines have the ability to mimic intelligent human behavior, including learning and problem solving (Syam and Sharma, 2018).
With AI, firms can automate some routine functions in the sales process, but the more interesting aspect is its ability to augment sales by using personalization, customization and enhanced service, while simultaneously increasing effectiveness (Moncrief, 2017; Paschen et al., 2019). This includes using AI as an advanced analytics tool to engage in activities such as creating tailor-made offers to specific customers, conversing online with virtual agents and proactively suggesting maintenance (Balducci and Marinova, 2018; Miklosik et al., 2019).
The potential of AI opens many doors for marketing practitioners, who must be agile in an ever-changing sales environment that is getting increasingly more complex. During the last decades, the way customers conduct their purchasing has changed (Steward et al., 2019). With the explosion of information, buyers are now able to do independent research and set their own purchasing criteria. Before even interacting with a sales representative, B2B customers are normally 50-60% down the purchasing process (Adamson et al., 2012; Gartner, 2018a). In addition, online channels are being increasingly more used throughout the purchasing process,
10 where it has been found that 83% of buyers are accessing digital channels for more information even during later stages of the purchasing process (Gartner, 2018b). Business customers are content-driven, technically savvy and comfortable with engaging via digital channels (Vieira et al., 2019). This shift in B2B buyer behavior highlights the importance of marketers to adapt their practices towards digital solutions.
Traditional marketing with the goal to build brand awareness and to generate qualified leads that hopefully result in transactions is no longer enough. Today, marketing must support the entire customer journey, which is the entire process the customer through and touchpoints they interact with (Steward et al., 2019). A challenge that has arisen due to this is how to align multiple touchpoints with marketing actions to provide relevant information. Another challenge is
understanding both the customers, i.e. those making the buying decision, as well as the users, i.e. those who ultimately will use the service or product (Paschen et al., 2019). In complex B2B contexts using the correct marketing action is even more difficult. These contexts are characterized by (Schmitz et al., 2014):
• Technical complexity;
• Few, infrequent transactions with large economic value; • Buying-center involvement with many stakeholders; • Heterogeneous customer requirements;
• Long-term decision processes; • Highly individualized solutions.
At the same time customers are more demanding and more value conscious than ever before (Sweeney and Soutar, 2001), leading to the creation of value being key in marketing (Anderson et al., 1992; Woodruff, 1997). As a result, B2B sales in complex settings need to have a high level of personalization and customization to provide the best possible offering, which in turn will increase profits for the firm and value for the customer (Montgomery and Smith, 2009). To facilitate these challenges, the emergence of marketing automation tools has become advent. At its core, marketing automation is automatic personalization and customization of the marketing mix (Heimbach et al., 2015). Marketing automation is seeing substantial growth in the private sector. Forrester estimates that marketers will spend 25 billion dollars on marketing automation
11 by 2023 (Adams, 2018). In addition, the number of available marketing automation platforms has grown drastically from 10 to 292 between 2011-2018, with tech giants like Adobe entering the scene (Murphy, 2018).
This interest is due to many recognized benefits of marketing automation, such as increased ability to generate a higher quantity of more qualified leads (Järvinen and Taiminen, 2016; Sandell, 2016; Todor, 2017), multichannel view of prospect behavior (Todor, 2017), improved lead conversion (Heimbach et al., 2015; Todor, 2017) and return on marketing investments (Montgomery and Smith, 2009; Świeczak, 2013; Todor, 2017). The lead qualification process is claimed to be both improved as well as accelerated by delivering personalized content to
potential buyers (Järvinen and Taiminen, 2016).
1.2. Problem Statement
Although marketing automation has been present in the literature for almost two decades (Heimbach et al., 2015), the research field is still in its infancy (Murphy, 2018). While the
impact on research has been modest, the adoption of the technology in the business community is surging (Murphy, 2018). The primary functionality of marketing automation platforms is to deliver content automatically to users according to a specific set of rules (Järvinen and Taiminen, 2016). This functionality requires the user of the platform, the marketer, to define rules based on existing customer insights. The ability to generate such insights is one of the most prominent challenges for many marketing executives (Leeflang et al., 2014). Forrester predicts that AI technology will emerge as a prominent feature in marketing automation platforms to ease this challenge (Hussain, 2019). How companies should utilize these new features, and which benefits and challenges that arise with its applications, is an area which is not present in the marketing automation literature.
Prior related work by Järvinen and Taiminen (2016) demonstrated how marketing automation and content marketing can be integrated in the selling process in a way that creates business benefits. However, their work did not touch upon these new emerging features discussed above. Additionally, their study focused solely on business benefits which does not capture the broader concept of value creation. The need for such research is stressed further by Russo-Spena et al. (2019), who highlights a general lack of academic examination of the impact of AI on complex business system interactions and how the integration of resources can bring value.
12 In addition to the gap in marketing automation literature, researchers have acknowledged the need for studies regarding AI and personalization (Kannan and Li, 2017; Syam and Sharma, 2018; Wedel and Kannan, 2016). Since marketing automation is a tool for automatized personalization, it is highly connected to the research field of personalization.
Syam and Sharma (2018: 141) underline the research question “how can machine learning and AI enhance smart and continuous customer targeting in real time?”. A similar topic is
emphasized by Kannan and Li (2017: 40) who point to the need of research into “methodologies that provide real-time, accurate targeting across platforms as well as the development of
intermediaries who can help in personalization”. Furthermore, Wedel and Kannan (2016) brings up research gaps regarding the role of AI, cognitive systems and automated attention analysis systems in delivering personalized customer experience.
The need for further research within personalization is particularly prominent in the context of B2B since most of the current personalization research is centered around the B2C context (Strycharz et al., 2019). Generally, marketing research is centered around B2C, despite the fact that B2B and B2C account for approximately the same economic value of transactions (Lilien, 2016). Lilien (2016) points out three reasons for this being the case. Firstly, the B2B problem domain is heterogenous and complex, as discussed in the previous section. Secondly, there is a lack of easy data availability in B2B. Data in B2B are less voluminous than in B2C, due to comparably fewer and larger transactions. In addition, there are difficulties in collecting data, where multiple organizations normally must cooperate and align. Lastly, there is a lack of domain knowledge in B2B. All researchers act as consumers in their daily lives, therefore it is easier to relate to the field of B2C. Hence, there is a lot of potential for additional research focusing on the B2B domain.
From a broader perspective, how sales and marketing practices will be impacted by AI is an area where several scholars stress a need for further research (Flaherty et al., 2018; Moncrief, 2017; Paschen et al., 2019; Russo-Spena et al., 2019; Singh et al., 2019; Syam & Sharma, 2018). Understanding and implementing AI in a successful way will be paramount for businesses to remain competitive in the future. AI enables a deeper level of personalization, which in turn can create additional value for the customer.
13
1.3. Research Purpose
The purpose of this thesis is to investigate how a company in a complex B2B context can benefit from using AI in personalization to create value. Moreover, what kind of value it is perceived to bring will be examined. Marketing automation has been found to be a prominent tool to
automatize personalization processes but has yet to reap the full benefits of AI, which will be explored further in this thesis.
1.4. Research Question
To achieve the sought-after purpose, the following research question has been defined:
RQ: How can AI be used in the marketing automation process to create value in B2B contexts?
1.5. Delimitations
The study focuses on one large organization within industrial equipment, therefore, the findings cannot be considered as general. As the studied phenomenon is not applied to a full extent in the studied organizations, the results will be based on perceptions of individuals and hypothetical situations. Real applications might differ. In addition, the amount of applications of AI in personalization are endless and we will focus on the most prominent ones in regards of the studied organization.
The dynamics and processes for creation of value between supplier and customer is out of scope. Instead the study focuses on the component of value, if it is created and if so, what type of value. Even though the research question addresses AI, the scope is limited to marketing research and the technical details will not be central in the study.
1.6. Sustainability Aspects
This study also considers its impact on sustainability. Sustainability is often defined by three interconnected pillars: environmental (ecological), economic and social (Elkington, 1999). The environmental aspect is considered as the management of the physical environment in a way that supports life on the planet within ecological limits and protection of natural resources. Social sustainability entails the impact on people and society, often related to well-being, justice and equality. Lastly, economic sustainability refers to practices that support long-term economic growth and the ability to create value.
14 As value is a central building block in our study, the economic aspect will be assessed. In
addition, we’ll consider sustainability by considering social and ecological benefits & costs as value types. More in general, AI has the potential of freeing up time from repetitive work, leaving certain jobs redundant, therefore impacting the social aspect. Personalization on an individual level could lead to recommendations on how to use products in a way that saves resources and extends the lifetime of products, therefore impacting environmental aspects, which is related to UN sustainability development goal 13: climate action. However, personalization could also be used for buying recommendations, leading to more transactions and thus
potentially increasing total consumption. This relates to UN sustainability development goal 12: responsible consumption and production.
1.7. Outline of Thesis
The first chapter introduces the studied topic and presents the research question. Chapter 2 discusses previous related work and lays out theory that will be used throughout the study. The concepts of value, marketing automation, personalization and AI are discussed in more detail. The frame of reference in this chapter presents the theory that is used a basis for the study. Chapter 3 describes the methodology, along with the research approach, data collection and how data will be analyzed. Also, ethical considerations are discussed. In chapter 4 the findings from the study are laid out and analyzed with respect to theory. Lastly, chapter 5 presents the
15
2. Literature Review
This chapter discusses previous related research and serves as a foundation for the theories, models and concepts used in this study. The chapter is divided into four parts: personalization, value, the marketing automation process and AI. The chapter starts off with a presentation of the concept of personalization and the current views present in the literature. Thereafter, the value research field is presented, culminating in foundational characteristics of value and a value typology. Marketing automation and personalization process are scrutinized, and the concept of AI is introduced as well as how AI can impact the field of personalization. Finally, the
conceptualization of the most relevant theories which serves as theoretical foundation for the study are presented in the frame of reference.
2.1. Personalization
2.1.1. What is Personalization?
The concept of personalization is broad and occurs in several research fields besides marketing, including human-computer interaction, machine learning and data mining among others (Zanker et al., 2019). In the context of marketing, personalization generally refers to a customer-oriented marketing strategy that aims to deliver the right message, to the right person at the right time (Aguirre et al., 2015; Dangi and Malik, 2017).
Even though the interest in personalization among researchers has been cultivated by increasing internet usage (Montgomery and Smith, 2009) and rising e-commerce (Dangi and Malik, 2017), the use of personalization preceded the internet (Montgomery and Smith, 2009). Vesanen (2007) asserts that it is likely that personalization is as old as trade itself. Aguirre et al., (2015) describe the existence of both online and offline personalization. Examples of offline personalization include face-to-face service encounters where the employee adapts their behavior to the
characteristics of the customer, e.g. greets them by name (Aguirre et al., 2015). The internet has advanced the possibilities and usage of personalization tremendously. An example of online personalization is individualized search results. Google personalize by refining a user’s search result based on past behavior (Montgomery and Smith, 2009).
There are several definitions of personalization in the marketing literature (Strycharz et al., 2019). Vesanen (2007) provides a summary of available definitions and discusses the many faces
16 of personalization. Some definitions are centered around the context, e.g. (Allen et al., 2001) whom describes personalization as individualization of web experiences. Montgomery and Smith (2009: 130) propose the following definition where technology is required as an enabler for personalization and thus context centered: “the adaptation of products and services by the
producer for the consumer using information that has been inferred from the consumer's behavior or transactions”. Other early definitions are more general (e.g. Imhoff, 2001.; Wind and
Rangaswamy, 2001) and some are more focused on value creation (Peppers and Rogers, 1999). An overview of definitions is summarized in Table 1 adopted from (Vesanen, 2007) and
(Montgomery and Smith, 2009).
Table 1. Selection of definitions of personalization mentioned in marketing literature, (Montgomery and Smith, 2009:131; Vesanen, 2007: 412-413).
Source Definition
(Peppers and Rogers, 1999: 146)
“Customizing some feature of a product or service so that the
customer enjoys more convenience, lower cost, or some other benefit” (Allen et al., 2001:
32-33)
“Company-driven individualization of customer web experience”
(Imhoff, 2001: 467) “Personalization is the ability of a company to recognize and treat its customers as individuals through personal messaging, targeted banner ads, special offers on bills, or other personal transactions”
(Wind and
Rangaswamy, 2001: 15)
“Personalization can be initiated by the customer (e.g. customizing the look and contents of a web page) or by the firm (e.g. individualized offering, greeting customer by name etc.)”
(Montgomery and Smith, 2009: 131)
“The adaptation of products and services by the producer for the consumer using information that has been inferred from the consumer's behavior or transactions”
Another aspect of the literature is the relationship between personalization and customization. There is no consensus on the relationship between the two terms (Vesanen, 2007). Peppers and Rogers (1999) does not acknowledge the need to distinguish between the terms and Allen et al (2001) highlight the complexity in separating the two terms. Other scholars view customization
17 as a form of personalization (Imhoff et al., 2001) or even as an advanced form of personalization (Wind and Rangaswamy, 2001). In more recent literature (Aguirre et al., 2015; Arora et al., 2008), the difference is centered around who initiates the adaptions of the marketing mix (Cöner, 2003). If the adaptions are requested proactively by the customer, it is customization, while if the adaptions are initiated by the company, it is personalization (Aguirre et al., 2015; Arora et al., 2008). This view creates a similar relationship between the terms personalization and
customization and push and pull personalization (Wedel and Kannan, 2016).
In this thesis we build on the definition presented by Montgomery and Smith (2009), due to its broad nature which enables a variety of data-driven applications to be within the scope of personalization. The amendment we make is that not only products and services can be
personalized, but also other objects such as activities, experiences, technologies, etc. Therefore, we consider personalization as “the adaptation of objects by the producer for the customer using information that has been inferred from the customer's behavior or transactions”. Compared to personalization, customization is viewed as adaptions of the marketing mix requested or initiated by the customer.
2.1.2. Levels of Personalization
Personalization and customization are the two main concepts in one-to-one marketing. One-to-one marketing is a concept that suggests that at least One-to-one part of a company’s marketing mix should be adapted to the individual customer. It is considered as an extreme form of
segmentation where the segmentation size is one (Arora et al., 2008). Personalization can be performed with different levels of granularity and Wedel and Kannan (2016) identified three levels of granularity on which personalization can be performed:
1. Mass personalization is when all customers receive the same offering or adaptions are made based on their average taste;
2. Segment level personalization is when all customers in the same segment receive the same adaptions to the marketing mix;
3. Individual level personalization is when each customer receives adaptions that are based on them only.
Similarly, Huang and Rust (2017) distinguishes static and dynamic personalization. Dynamic personalization adapts to a specific customer’s preferences based on the customer’s active input
18 and by observing the customer’s behavior over the time, rather than relying on cross-sectional customer data from similar customers as in the case of static personalization. This is comparable with the segment and individual levels described above. Static personalization can be achieved by analytics and big data, while dynamic personalization hinges on AI and other cutting edge technologies (Huang and Rust, 2017).
2.2. Value
2.2.1. What is Value?
To create an understanding of the concept of value, one must delve into the development of different value perspectives. The study of value dates to the time of Aristotle (Eggert et al., 2018; Gordon, 1964; Grönroos, 2011). In his value theory he raised the famous value paradox,
distinguishing between two ways a product can be used (Gordon, 1964). A shoe for example can either be used for wearing or used for exchange. Drawing on this distinction, two complementary perspectives on customer value were introduced, use (or use value) and
value-in-exchange (or value-in-exchange value), which have been highly adopted by economics (Smith, 1791). In addition, Gordon (1964) points out that Aristotle treats use value as the subjectively perceived benefit, where demand is a function of use value and exchange value something that is derived from use value.
Early, popular work on value that builds onto the perspectives is that by Lawrence D. Miles (Lindgreen and Wynstra, 2005; Wilson and Jantrania, 1994). Miles (1961: 3) explains the elusive nature of value: “value means a great many things to great many people because the term value is used in a variety of ways. It is often confused with cost and with price. In most cases, value to the producer means something different from value to the user. Furthermore, the same item may have differing value to the customer depending upon the time place, and the use”. In addition to use value and exchange value, Miles identifies two additional types of value.
1. Esteem value is the intrinsic attractiveness, features or properties, which causes a desire to own it.
2. Cost value is the cost of producing a product, meaning the sum of labor, material, and various other costs.
19 Wilson and Jantrania (1994) highlights the importance of setting the discussion of value in relation to the social, economic, political and religious environment surrounding the affected individual(s). They conclude that value is a problematic concept to define, but that it cannot be ignored. Similarly, Holbrook (1994) emphasizes the difference between ‘value’ and ‘values’, where the first is defined as preferential judgements (i.e. benefits vs sacrifices) and the second the criteria that determine those preferential judgements (i.e. enduring beliefs). To avoid
unnecessary complex discussion, the focus in this thesis will be on the general realm of customer value, rather than of philosophical and ethical values.
2.2.2. Value in Business Marketing as a Developing Concept
Marketing, as an offspring of economics, began with an exchange centered view on value (Eggert et al., 2018). With this view, the supplier manufactures and distributes goods and services that are embedded with value. The value is created and determined by the supplier, it can be exchanged and it is the marketing’s job to understand and communicate the value to the customer (Anderson and Narus, 1998). Zeithaml (1988) builds on this logic, emphasizing value as subjectively perceived and as an overall assessment of the utility of a product or service. The perceived value is described as a trade-off between what the customer receives and what it gives in exchange (Woodruff, 1997; Zeithaml, 1988). With the competition taken into account
Anderson et al. (1992) defined value in business markets as the perceived worth in monetary units as the set of economic, technical, service and social benefits received by a customer in exchange for a price for the product offering, taking into consideration competitors offerings and prices. Some scholars consider the value and price as independent elements, the price is what a customer pays for a market offering (Anderson et al., 2000). The difference between the price and the value is the customer’s incentive to purchase a market offering. These findings are reflected in an extensive review by (Ulaga and Eggert, 2005) who reviewed a variety of definitions of value and identified four recurring characteristics:
1. Customer value is a subjective concept.
2. Customer value is conceptualized as a trade-off between benefits and sacrifices. 3. Benefits and sacrifices can be multifaceted.
20 Then a shift in the marketing literature occurred, emphasizing the long-term value of
relationships. As marketing features a continuation of transactions (Dwyer et al., 1987), more value can be accrued through relationships exchanges rather than from transactional exchanges. In a relational context, customer value is not embedded in the transactional exchange of a product or service for money (Grönroos, 1997). Instead perceived customer value is created and delivered over time, as the relationship develop over time. Ravald and Grönroos (1996) explains the value of relationships as the total episode value, or the sum of all the interactions that create value in a relationship. However, due to difficulties in estimating the future value of relationships the identification of high value relationships is problematic (Wilson and Jantrania, 1994).
The importance of understanding business relationships in terms of value and the creation of value has since been highly discussed (Ulaga and Eggert, 2006). Researchers started
emphasizing the process of co-creation of value between the supplier and customer. In this point of view, relationship value is a “measure of joint outputs, underpinned by co-operation, where the nature of the interaction between supplier and customer is critical in the creation of joint value” (Lindgreen et al., 2012: 209).
With a growing trend towards servitization of business markets, the traditional view of business marketing was challenged. The economic exchange of service provision rather than goods led to the introduction of service dominant (S-D) logic in marketing (Vargo and Lusch, 2004). This logic emancipates the marketing discipline from its economic heritage and focuses on the subjective value in use and integration of resources rather than resource exchange. In this sense the customer integrates resources from the provider such as products and services with other skills and resources to (co-)create value. With the S-D logic, value is perceived and determined by the customer, co-created through resource integration and it is the marketing’s job to identify and facilitate opportunities for co-creation in the use context of the customer (Vargo and Lusch, 2004). Payne et al. (2008) builds on the S-D logic and investigates how value co-creation with the customer occurs. They developed a conceptual framework consisting of three main
21 1. Customer value-creating processes is the processes, practices and resources that the
customer uses to manage its business and relationships with suppliers.
2. Supplier value-creating processes is the processes, practices and resources that the supplier uses to manage its business and relationships with customers.
3. Encounter processes is the processes and practices that facilitate exchange and interaction in a customer-supplier relationship, which needs to be managed to successfully realize co-creation opportunities.
Each component consists of their own set of procedures, tasks, mechanisms, activities and interactions which ultimately supports co-creation of value. In this sense the supplier creates value propositions, where the customer determines value after consumption and the dialogue is seen as interactive process of learning together (Ballantyne and Varey, 2006).
With respect to the S-D logic and previous value literature, seven foundational characteristics for customer value have been identified (adapted from Leroi-Werelds, 2019):
1. Customer value implies an interaction between a customer and an object (e.g. a product, service, store, technology, activity, etc.).
2. Customer value involves a trade-off between the benefits and sacrifices of an object; 3. Customer value is not inherent in an object, but in the customer’s experiences derived
from the object.
4. Customer value is subjective and personal as value perception are based on personal characteristics.
5. Customer value is context-specific, considering situation, time frame, circumstances and location.
6. Customer value is multidimensional and consists of multiple value types.
7. Customer value is co-created by the customer by means of resource integration between supplier and customer.
These seven foundational characteristics will be used throughout the thesis as a common denominator for value.
22 2.2.3. Value Typologies and Dimensions
Many attempts to conceptualize value have been made (Gallarza et al., 2017; Leroi-Werelds, 2019; Lin et al., 2005). One of the earlier defined typologies, that has gained a lot of traction (Gallarza et al., 2017), is the one by Holbrook (1994). His typology comprises three dimensions that emerge in the customer experience: extrinsic vs intrinsic value, self-oriented vs other-oriented value and active vs reactive value. In turn, by combining these dimensions eight types of value arises: efficiency, excellence, status, esteem, play, aesthetics, ethics, and spirituality (Holbrook, 1996). His perspective is in line with the S-D logic, where value is experienced by the customer (Holbrook, 1996) and this is reinforced in his later work where he describes that value only resides in a consumption experience (Holbrook, 2006). The relevance of certain types of values in Holbrook’s typology has been discussed, e.g. ethics (Smith, 1996), other has
combined spirituality and ethics value into ‘altruistic value’ and split ‘excellence’ into ‘service’ excellence’ and ‘product excellence’ (Willems et al., 2016). Also, it has been highlighted that customers increasingly value sustainability (Sudbury-Riley and Kohlbacher, 2016). Likewise, value creation involves a process that increases the customer’s well-being (Grönroos and Voima, 2013), making ecological and societal aspects even more relevant when assessing value. The ethics value dimension by Holbrook (1996) has been argued to be related to “respect and care for the environment from the organization” and “collaboration with social causes” (Gallarza et al., 2017: 741), therefore these can be seen as the value types ‘ecological’ and ‘societal’.
Sheth et al. (1991) proposed a value construct where customer value is a function of five independent value dimensions: social, emotional, functional, epistemic and conditional value. These dimensions affect the perceived utility of a choice and whether a customer should make a purchase decision. Sweeney uses the first three dimensions (see Table 2) to propose PERVAL, a scale to measure perceived value of consumer durable goods (Sweeney and Soutar, 2001). In comparison, he sees the dimensions as interrelated rather than independent and breaks down functional value into the subcomponents price and quality. These dimensions are seen in a trade-off perspective, where price is the only sacrifice component and the remaining as benefit
23 Table 2. The value dimensions in PERVAL (Sweeney and Soutar, 2001: 211).
Dimension Description
Emotional value the utility derived from the feelings or
affective states that a product generates Social value (enhancement of social
self-concept)
the utility derived from the product’s ability to enhance social self-concept
Functional value (price/value for money)
the utility derived from the product due to the reduction of its perceived short term and longer term costs
Functional value (performance/ quality)
the utility derived from the perceived quality and expected performance of the product
Lapierre concurs with seeing value in terms of benefit vs sacrifice components. He conducted interviews with suppliers and customers in an industrial B2B context and empirically identified several value dimensions related to product, service and relationship (see Table 3).
Table 3. Total value proposition with value dimensions (Lapierre, 2000: 125)
Product Service Relationship Benefit Alternative solutions
Product quality Product customization Responsiveness Flexibility Reliability Technical competence Image Trust Solidarity
Sacrifice Price Price Time/effort/energy Conflict
Seeing value as a first-order multidimensional (or unidimensional) construct as in the case of Pierre and Sweeney has been argued to be inadequate (Lin et al., 2005). Lin et al. (2005) instead offers a formative second-order multidimensional construct to assess value perception (see Figure 1). They conceptualize perceived value with different give-get (benefit vs sacrifice) components and consequence dimensions of perceived value (satisfaction and behavioral intentions), which are manifested by insdicators.
24
Figure 1. Sample model for perceived value as a second-order formative construct (Lin et al., 2005: 325). Get and Give components can be any number and include dimensions as price, privacy and security.
Their model is tested on the web site eTail’s service value survey data, where they use the following dimensions to conceptualize value: monetary sacrifice, web site design,
fulfillment/reliability, security/privacy and customer service.
Macdonald et al. (2011) uses a more experiential perspective and conducts a case study in B2B. Based on the case study, they suggest a conceptual framework for assessing value-in-use, with the resulting types of value-in-use: continuity of operation, retention of knowledge, retention of competencies, security and time
Other researchers focus solely on the concept of relationship value. Wilson and Jantrania (1994) examines the creation of value in industrial buyer-supplier relationships by comparing several disciplines and proposes three independent dimensions of relationship value: economic, strategic and behavioral dimensions. In contrast, Ruiz-Martínez et al. (2019) measure relationship value in manufacture-supplier relationships along three axes: core axis, ICT axis and access axis. The core axis represents the essential benefits and sacrifices in a relationship, the ICT axis reflect dimensions related to the use of technologies and the access axis the represent importance of social interaction (Ruiz-Martinez et al., 2018). The investigated dimensions include product quality, low quality cost, ICT costs, customization, order delivery, personnel training, seller
25 support, ICT benefits, electronic notification, switching cost and social interaction. In their study the most notable dimension in contributing to relationship value was seller support.
By revising the typology by Holbrook with the trade-off logic, relationships and other prominent value dimensions, the typology in Table 4 has been proposed. These types can be considered as a “menu card” for a study, where some values are more relevant than others depending on the context (foundational characteristic 5).
26 Table 4. Revised value typology adapted from Holbrook (1996: 139-140)
Value types Brief description Papers
Benefits The perceived advantages
Efficiency/ Convenice
The extent to which an object makes life easier for the customer. E.g. increased output of products from a given time
(Holbrook, 1996; Lin et al., 2005)
Status The extent to which an object enhances a positive impression on others
(Gallarza et al., 2017; Holbrook, 1996; Sweeney and Soutar, 2001) Excellence The extent to which an object is of high quality.
Can both be product and service excellence. Includes reliability and responsiveness
(Holbrook, 1996; Lin et al., 2005; Macdonald et al., 2011)
Aesthetics The extent to which an object is appealing. Relates to sensory appreciation
(Gallarza et al., 2017; Holbrook, 1996; Willems et al., 2016) Escapism/
Spirituality
The extent to which an object allows customer to relax and escape daily routine
(Gallarza et al., 2017; Holbrook, 1996)
Self-esteem The extent to which an object positively affects the attitude and satisfaction of oneself
(Holbrook, 1996; Sweeney and Soutar, 2001)
Relational The extent to which an object improves the relationship with the service provider
(Lapierre, 2000) Social The extent to which an object results in better
relationships with other parties
(Ruiz-Martínez et al., 2019; Wilson and Jantrania, 1994) Epistemic The extent to which an object provides novelty,
arouse curiosity or satisfy a desire for knowledge
(Sheth et al., 1991) Ecological
benefits
The extent to which an object positively impacts environmental well-being
(Gallarza et al., 2017; Holbrook, 1996)
Societal benefits
The extent to which an object positively impacts societal well-being
(Gallarza et al., 2017; Holbrook, 1996)
Sacrifices The perceived loss for the sake of other considerations
Price The extent to which an object demands monetary resources
(Lapierre, 2000; Lin et al., 2005; Ruiz-Martinez et al., 2018; Sweeney and Soutar, 2001) Time The extent an object requires time to prepare,
use, understand, etc.
(Lapierre, 2000; Macdonald et al., 2011)
Effort The extent an object requires effort or energy to prepare, use, understand, etc.
(Lapierre, 2000) Privacy The extent an object can result in loss of privacy (Lin et al., 2005) Security The extent an object can result in security issues,
e.g. being more vulnerable for hacking
(Lin et al., 2005; Macdonald et al., 2011)
Ecological costs
The extent to which an object negatively impacts environmental well-being
(Gallarza et al., 2017; Holbrook, 1996)
Societal costs The extent to which an object negatively impacts societal well-being
(Gallarza et al., 2017; Holbrook, 1996)
27 2.2.4. Value in Personalization
Vesanen (2007) conducted a literature study and proposed a conceptual framework (Figure 2) of personalization. He concluded that the main topics in personalization literature were execution of personalized marketing, personalized marketing output, value for customer and value for
marketer. The value conceptualization used is that value created is the margin between benefits and sacrifices.
The benefits of personalization for the customer include better preference match, service,
communication (Cöner, 2003; Murthi and Sarkar, 2003; Vesanen, 2007; Wind and Rangaswamy, 2001) and the experience of one (Vesanen, 2007). An additional benefit recognized in literature (Ansari and Mela, 2003) as well as among marketing professionals (Strycharz et al., 2019) is reduced information overload for the customer. The sacrifices for the customer listed by
Vesanen (2007) are privacy risks, spam risks, spent time, extra fees and waiting time. Successful personalization requires data about past behavior, which can be perceived as intrusive (Miceli et al., 2007; Montgomery and Smith, 2009; Strycharz et al., 2019; Vesanen, 2007). Strycharz et al. (2019) interviewed marketing professionals and found that privacy risks are viewed as a
boundary condition of personalization success. Privacy concerns can negatively influence the effectiveness of personalization. If customers get cues that their data have been collected without their consent, they tend to have a negative reaction (Aguirre et al., 2015). This serves as the foundation for the personalization-privacy paradox, where personalization leads to an increase of relevance at the expense of an increased sense of vulnerability (Aguirre et al., 2015).
Additionally, Vesanen (2007) addresses value created for the marketer through personalization. The benefits recognized for the marketer includes higher prices, better response rates, customer loyalty, customer satisfaction and differentiation from competitors (Ansari and Mela, 2003; Vesanen, 2007). Potential sacrifices for marketer include investments, risk of irritating customers with marketing output and potential brand conflict.
28 Figure 2. A conceptual framework of personalization (Vesanen, 2007: 4147)
Another framework proposed by Miceli et al. (2007) uses value as a dimension of
personalization. They define value of personalization as the site-specific and content-specific attributes and benefits expected by the customer. Site-specific attributes consist of graphics, content layout and site updating, while content-specific attributes consist of presentation of information. This framework covers several of the same topics as the one proposed by Vesanen (2007) including value and customer marketing relationship. Their view on value solely deals with benefits and not sacrifices.
29
2.3. Marketing Automation
2.3.1. Marketing Automation and Personalization Processes
Marketing automation is closely related to personalization. Heimbach et al.(2015: 130) describe marketing automation in the following way: “the core of marketing automation is an automatic customization or personalization of marketing mix activities”. A similar view is stated by
Järvinen and Taiminen (2016: 165) whom state that the objective of marketing automation is “to attract, build and maintain trust with current and prospective customers by automatically
personalizing relevant and useful content to meet their specific needs”. Furthermore, several scholars acknowledge the process nature of personalization (Vesanen, 2007). Consequently, personalization processes are viewed as marketing automation processes if they can be
performed automatically. Thereby, the following literature review will include processes from the personalization literature in addition to processes from the marketing automation literature. Several scholars have attempted to describe the process of delivering a personalized marketing output. Within the field of personalization, processes have been proposed with varying levels of granularity (Adomavicius and Tuzhilin, 2005; Murthi and Sarkar, 2003; Pierrakos et al., 2003; Vesanen and Raulas, 2006). Within marketing automation research, a general framework for marketing automation has been proposed (Heimbach et al. 2015) and as a part of the wider concept of the sales and marketing process (Järvinen and Taiminen, 2016).
Vesanen and Raulas (2006) reviewed the personalization literature and found that there are nine required basic elements to execute a personalized marketing output. The nine basic elements are: customer, dialogue with customer, customer data, analyses of customer data, customer profile, customization, marketing output and delivery of marketing output (Vesanen and Raulas, 2006). They categorized these elements further into operations or results of operations which are referred to as objects. To incorporate the previously discussed differentiation between
personalization and customization used in this thesis, personalization will be can be seen as an additional tenth element.
Table 5, provides a comparison of personalization processes, adapted from Vesanen and Raulas (2006), and a mapping of the basic elements in the relevant steps. In the marketing automation process proposed by Heimbach et al., (2015) step three and four of the process have been merged to increase comparability.
30 Table 5. Comparison of personalization processes, adapted from Vesanen and Raulas (2006)
Heimbach et al. (2015) Vesanen and Raulas (2006) Pierrakos et al. (2003) Murthi and Sarkar (2003) Adomavicius and Tuzhilin (2005) Basic Elements of Personalization Relevant by process step 1. Data inputs
Interaction Data Collection Learning Understanding the customer by using data to build customer profiles Customer Interaction Data Processing 2. Real time decision rules Processing Data preprocessing Matching offerings to customer Deliver personalized offering by matchmaking Processing Customer profile Personalization Delivery Customer 3. Update and optimize rules Personaliz ation Pattern discovery Evaluation Measure personalization impact and adjusting personalization strategy Interactions 4. Delivery by chosen medium and content Delivery Knowledge postprocessing Delivery
5. Personalization Customer profile
Matching
6. Report Customer data
Processing Customer Profile
The processes proposed by (Adomavicius and Tuzhilin, 2005; Heimbach et al., 2015; Murthi and Sarkar, 2003; Pierrakos et al., 2003; Vesanen and Raulas, 2006) have strong similarities in terms of the basic elements that they incorporate. Murthi and Sarkar (2003) and Adomavicius and Tuzhilin (2005) both divided the process into three main steps:
1. Learning in terms of data collection and data inference. 2. Matching marketing action and customer preference. 3. Evaluating the efficiency of the action.
Adomavicius and Tuzhilin (2005) process included additional subcomponents which are detailed in Table 5. The process by Pierrakos et al. (2003) include additional intermediary steps,
preprocessing and postprocessing, that are more relevant from a data mining perspective. Moreover, they suggest that reports should be a part of the process which differentiates their
31 process further from the ones by (Adomavicius and Tuzhilin, 2005; Murthi and Sarkar, 2003) . Vesanen and Raulas (2006) aimed to synthesize the previously discussed processes by
interlinking the basic elements of the personalization process: customer, customer data, customer profile and marketing output, seen in Figure 3.
Figure 3. The personalization process by Vesanen and Raulas (2006)
Heimbach et al., (2015) proposed a general framework (see Figure 4) for marketing automation based on Little's (2001) five levels of system operations. The five levels of system operations are data inputs, real time decision rules, updates of the decision rules, feedback to site management, and strategy choice. The first step, data inputs, aims to ensure data availability. Heimbach et al., (2015) categorize the data inputs into current information, e.g. behavior on the website and stored information, e.g. transactional history. The second step is usage of real time decision rules, which have been predefined based on observed patterns in the data. These rules are then updated in the third step, which can be automated to some extent, e.g. by A/B testing (Heimbach et al., 2015). The fourth step relates to feedback to management, which enables monitoring and optimization of performance. The fifth step is related to the strategic choices regarding content and medium.
32 Figure 4. Marketing automation framework adopted from (Heimbach et al., 2015: 131)
Järvinen and Taiminen (2016) used the classic sales and marketing funnel in Figure 5 to show the role of marketing automation in the sales and marketing process by doing a case study on a large B2B-company. Instead of describing the process of personalization, they describe
personalization as a part of a sales and marketing process. Therefore, their process is excluded from the comparison in Table 5. The marketing and sales funnel consist of five stages: identified contacts (or suspects), marketing leads (or prospects), sales leads, opportunities (or qualified leads) and deals. In the first stage, marketing automation contributes by identifying suspects through their contact info, login, cookies and IP-addresses (Järvinen and Taiminen, 2016). The basic elements of personalization that occur in this step are customer data, analysis of customer data and customer profile. In the second stage, marketing leads receive automatic nurturing programs and lead scoring (Järvinen and Taiminen, 2016) which can be broken down to the basic elements personalization marketing output and delivery of marketing output. Previous findings by Strycharz et al. (2019) indicate that the sales funnel is a common way to evaluate
33 Figure 5. The marketing and sales funnel related to content marketing, marketing automation, and CRM adopted from (Järvinen and Taiminen, 2016: 170)
Järvinen and Taiminen (2016) limit the applications of marketing automation to the first two stages of the sales and marketing funnel, which significantly reduces the number of potential use cases of marketing automation. Marketing automation can be used for promotion of additional products for cross-selling or up-selling, which occur in the later stages of the sales and marketing funnel. Thus, a need for extension of the process description adopted from Järvinen and
Taiminen (2016) is recognized and we argue that marketing automation can be used throughout the sales and marketing funnel.
34 2.3.2. Conceptualization of the Marketing Automation Process
There have been several attempts to describe the process of delivering a personalized marketing mix. Out of the reviewed processes none of the articles explicitly discuss AI despite the fact that big data and algorithmic personalization is believed to be the future for personalized marketing (Strycharz et al., 2019). The basic elements of the personalization process that has the strongest relationship to AI are those that incorporate data and algorithmic aspects. Therefore, the basic elements customer data, analysis of customer data, customer profile and personalization should be included.
Several types of data of have been acknowledge as useful for personalization by scholars (Kumar et al., 2019; Murthi and Sarkar, 2003; Vesanen and Raulas, 2006). Murthi and Sarkar (2003) separates the data that can be collected through monitoring the user, into user-centric data and site-centric data. User-centric data consist of web behavior that is stored in the user’s machine. Site-centric data captures a subset of activities that are present in user-centric data but cannot be used to explain competitive effects (Murthi and Sarkar, 2003). Moreover, the usage of site-centric data is highlighted by Vesanen and Raulas (2006). Other types of data that have been acknowledged as suitable for AI powered personalization is:
- Data on firm-customer transactions (Kumar et al., 2019; Vesanen and Raulas, 2006) - Customers’ consumption pattern of offering (Kumar et al., 2019; Montgomery and Smith,
2009; Murthi and Sarkar, 2003; Vesanen and Raulas, 2006)
- The communication pattern about firm offerings to customers (Kumar et al., 2019) - Clickstreams (Montgomery and Smith, 2009)
Learning about customer preference requires data collection (Montgomery and Smith, 2009; Murthi and Sarkar, 2003). Data collection can be done overtly, where the company informs the customer about the data collection or covertly which is when the company collect customer data without them being aware (Aguirre et al., 2015; Murthi and Sarkar, 2003). Since data collection and processing can have negative effects related to the value of personalization it is important to include it in a process which is done in most personalization processes (Adomavicius and Tuzhilin, 2005; Heimbach et al., 2015; Järvinen and Taiminen, 2016; Pierrakos et al., 2003; Vesanen and Raulas, 2006), but not as explicitly as by Murthi and Sarkar (2003). A potential negative effect originating from personalized marketing is customer privacy concerns.
35 Several of the reviewed processes have included an explicit step that aim to encapsulate creation of customer understanding (Adomavicius and Tuzhilin, 2005; Murthi and Sarkar, 2003;
Pierrakos et al., 2003; Vesanen and Raulas, 2006). Marketing professionals have a recognized difficulty in translating data into customer behavior insights (Leeflang et al., 2014). Kumar et al., (2019) suggested that AI can be used in personalization to automatically predict the type, timing, and purchase of preferred firm offerings which are examples of how AI can be used to learn about the customer.
The execution of personalized marketing output encompasses the interaction between marketer and customer in the shape of channel and content. This is described as a process that interlinks the customer and the marketer (Vesanen, 2007), a view that is shared by several scholars (Murthi and Sarkar, 2003; Strycharz et al., 2019; Vesanen, 2007; Wind and Rangaswamy, 2001). The interactions created by the personalized marketing output creates the relationship between marketer and customer (Vesanen, 2007; Wind and Rangaswamy, 2001).
To achieve successful personalization the marketing professionals is required to understand the customers preferences (Strycharz et al., 2019). Gaining insights about customers and their preference is a challenge for many marketers (Leeflang et al., 2014). Therefore, a process should include an explicit step for evaluation to ensure qualitative preference match. Research regarding the effectiveness of personalized marketing show varied results (Strycharz et al., 2019).
Moreover, the need for evaluation is stressed further by the need to estimate the positive impact on the company’s profits which is required for successful personalization (Kaptein and Parvinen, 2015). They also use the ability to measure effects as a requirement for successful
personalization. The personalization processes proposed by Adomavicius and Tuzhilin (2005) and Murthi and Sarkar (2003) mention evaluation explicitly. In the marketing automation by Heimbach et al. (2015) optimization is discussed where some form of evaluation is assumed to be required.
The importance of evaluation underlines the advantage of relating the marketing automation process to the sales and marketing funnel, as done by Järvinen and Taiminen (2016), compared to viewing it as a standalone process. It enables for a structured view of the relationship between marketing action and outcome. It creates a more prominent focus on how marketing automation relates to marketing outcomes and thereby a link to value creation. Furthermore, Strycharz et al.
36 (2019) conclude that a funnel perspective is often used among marketers to evaluate
effectiveness. This indicate that a process where evaluation is linked to the sales funnel could be simplify application among marketers.
In summary, the reviewed personalization literature supports that a process, that aims to describe automatic personalization, should incorporate four process steps. The first step is data which is achieved through data collection. The main reason for including this as an explicit step is the associated customer sacrifices related to privacy (Aguirre et al., 2015; Vesanen, 2007) and the potential negative consequences on value (Vesanen, 2007). The second step, learning, aims to describe the understanding of the customer that arises as a result of data analysis. The purpose of the understanding is to be able to match marketing content to a customer. Data analysis is a prominent part of several processes existing in literature (Adomavicius and Tuzhilin, 2005; Murthi and Sarkar, 2003; Pierrakos et al., 2003; Vesanen and Raulas, 2006). When the company has gained an understanding of the customer’s preferences and created a customer profile, it should be matched to an appropriate marketing output, thereby personalized. The importance of these steps are highlighted by Kaptein and Parvinen (2015) who view algorithm scalability as a prerequisite to successful personalization. The delivery of the personalized marketing is the third step, here it is assumed that the operations can be automized. As discussed in the earlier section another prerequisite to successful personalization according to Kaptein and Parvinen (2015) is the ability to evaluate it which is the final step. The process summary can be viewed in Figure 6.
37
2.4. Artificial Intelligence
2.4.1. Big data and AI
With so much value hiding in data, it has been referred to as the oil of the digital economy (Yi et al., 2014). Compared to oil, the amount of data is growing at an unprecedented speed, with an estimated generation of 2.5 quintillion bytes per day (Dobre and Xhafa, 2014).
In 1997, Michael Cox and David Ellsworth were among the first to use the term big data. Since then, numerous definitions of big data have appeared in literature. One of the most popular definitions originates from IBM and describes the three V’s of big data (O’Leary, 2013). It has later been expanded to the following five V’s characterizing big data (Ishwarappa and Anuradha, 2015):
• Volume refers to the sheer amounts of data in terms of stored terabytes, tables and files, transactions and records. The volume of big data renders traditional data warehousing methods unusable.
• Velocity aims to describe the increasing speed of data creation. Attributes of velocity include number of batches, data streams or processes and the use of real time data.
• Variety is an indicator of the different types of data that is stored. There is structured data and unstructured data, where unstructured data is significantly more complex to process. Of all generated data, 90% is unstructured.
• Veracity refers to the degree of correct data in a large data set. Low veracity is detrimental to the quality of analysis based on the big data set.
• Value is said to be the most significant aspect of big data. It refers to the potential value the data can turn into.
Big data and AI are closely related. AI uses voluminous data sets to perform advanced pattern recognition and learning tasks (O’Leary, 2013), therefore the presence of big data enables AI (Thrall et al., 2018). Artificial intelligence is manifested as machines that exhibit some aspects of human intelligence (Huang and Rust, 2018). Furthermore, they described four types of artificial intelligence: mechanical, analytical, intuitive and empathic. The categorization can be used to determine the complexity of AI in certain applications, depending on the nature of the task that it shall perform (Huang and Rust, 2018). The four types of artificial intelligence are explained further below.
38 1. Mechanical intelligence can handle simple and standardized tasks were limited training is
required (e.g. call center agents or retail sales agents). It solely relies on observations to act.
2. Analytical intelligence can perform analytical, rule based, systematic and complex tasks. This type of intelligence can make rational decisions and handle work that require technical training and expertise (e.g. financial analysis, data science and accounting). 3. Intuitive intelligence learns and adapts intuitively based on understanding. An agent that
possess this type of intelligence can handle work that require creative problem solving (e.g. being a doctor, management consultant or marketing manager). Tasks that can be managed by an agent that possess this type of intelligence can be complex, chaotic and idiosyncratic.
4. Empathetic intelligence can recognize emotions and make decisions that incorporate emotions. Characteristics include sociability, being emotional, communicative and highly interactive. Agents that possess this intelligence can perform tasks that require
relationship building and communication for example being a physiatrist or politician. In comparison, Davenport et al. (2020) distinguish two levels of intelligence, task automation and context awareness. The former is described as the case of standardized or rule-based AI, where consistency and logic are required. Context awareness is more advanced and requires machines to “learn how to learn” and therefore, to be able to extend beyond a certain domain or context. The aforementioned two are highly related to the concepts of narrow AI (maps onto analytical and mechanical intelligence) and general AI (maps onto intuitive and empathetic intelligence) (Kaplan and Haenlein, 2019). Davenport et al. (2020) includes the intelligence levels in an AI framework along with two other dimensions, task type and whether AI is embedded in a robot. The proposed AI framework is derived from extant marketing literature revolving around AI, with the goal to create an understanding of AI. Task type refers to whether the AI application analyzes numerical or non-numerical data (e.g. voice, text or images). Usage of non-tabular data is more complex and often translated to numerical data, e.g. pixel color in an RGB scale. The last dimension relates to whether AI is virtual or embedded in a robot. The robot form has elements of physical embodiment, which has been found to offer substantial
