CONFIDENTIAL + ANONIMISED

(1)

1

CONFIDENTIAL

+ ANONIMISED

The new role of health insurance companies in a

health-focussed society

Investigating the effectiveness of direct marketing on health

program engagement

CONFIDENTIAL

Laura Nauta

(2)

2

The new role of health insurance companies in a

health-focussed society

Investigating the effectiveness of direct marketing on health

program engagement

Master Thesis Marketing Intelligence and Marketing Management University of Groningen

Faculty of Economics and Business Department Marketing

PO Box YYY, 97YY AV Groningen (NL)

Laura Nauta

Antillenstraat 1-Y5, 971Y JT Groningen (+31)Y Y3 Y1 Y3 YY

L.Y.Nauta@student.rug.nl SYYYYY3Y

1st supervisor:

Prof. Dr. Ir. K. van Ittersum (K.van.Ittersum@rug.nl)

Ynd supervisor:

Prof. Dr. T.H.A. Bijmolt (T.H.A.Bijmolt@rug.nl)

External supervisors (X): X

X

(3)

3

Management summary

Recently, the global healthcare industry is experiencing a tremendous change in the way the industry works as a consequence of the continuously increasing healthcare expenses. The intention of this enormous change is to increase efficiency while also responding to three new trends in the market: a new market system resulting in more freedom of choice for the consumer, increased focus on a healthy lifestyle and/or healthy living, and increased use of internet and mobile applications (apps) for medical or health purposes.

Building upon these three trends, X developed a platform called X (English translation: ‘HealthyTogether’), which stimulates participants to improve their lifestyle. The X platform is one of the first large health programs developed and, consequently, a lot is still unclear about this type of health intervention programs. This paper will investigate the effectiveness of a health program using three research questions: (1) What is the effect of direct marketing on a participant’s health program engagement (HPE)?; (Y) What is the effect of customer characteristics, relational characteristics, and mass media on a participant’s HPE?; and (3) What is the effect of customer characteristics, relational characteristics, and mass media on the direct marketing-HPE relationship?

A literature review was performed to identify current literature on the topic of health interventions, which highlighted that no conclusive findings on health improvement as a result of a health intervention were found. Another conclusion based on the literature review was that the X platform was highly distinctive from current research.

(4)

4 the full model, including all exploratory variables, was estimated. Besides, for two dependent variables interaction effects were included.

Overall, this study showed that HPE is hard to be maintained over time. However, it can be concluded that in general direct marketing and the intensity of the health insurer-insured relationship have a positive effect on HPE. The most important findings regarding direct marketing in more detail are: overall the recruitment, informative, and birthday email are found to be most effective for increasing the number of the HPE components, while the points and activation email have less effect. Besides, the points, informative, and birthday email are found to be most effective for creating an observation of the HPE components (one instead of zero), while the welcoming, recruitment, and activation email are less effective.

Besides, customer characteristics result in different HPE levels. On the other hand, mass media does not impact HPE. Regarding possible interaction effects on the direct marketing-HPE relationship, only significant interactions are found for customer characteristics (vitality) and relational characteristics. In terms of relational characteristics, health program-health participant relationship length is found to have a positive impact, while health insurer-insured relationship length and intensity of the health insurer-insured relationship are found to have a negative effect on the direct marketing-HPE relationship.

Keywords: engagement, healthcare, health insurance, health intervention, health program engagement

Acknowledgements

(5)

5

1 Introduction

Recently, the global healthcare industry is experiencing a tremendous change in the way the industry works as a consequence of the continuously increasing healthcare expenses (Van der Horst, Van Erp, and De Jong YY11; Thomson et al. YY13; van Weel, Schers, and Timmermans YY1Y). For example, in the Netherlands healthcare expenses have increased from Y% of gross domestic product (GDP) in 197Y to more than 13% of GDP in YY1Y, and the prediction is that in YYYY it will be YY% of GDP (Van der Horst, Van Erp, and De Jong YY11; Stichting X Beheer YY1Y). The intent of the enormous change in the healthcare industry is to increase efficiency while also responding to three new trends in the market: (1) a new market system resulting in more freedom of choice for the consumer, (Y) increased focus on a healthy lifestyle or healthy living, and (3) increased use of internet and mobile applications (apps) for medical or health purposes (Dute, Bemelmans, and Breda YY1Y; Maarse, Jeurissen, and Ruwaard YY1Y; Stephens, Allen, and Dennison Himmelfarb YY11; Thomson et al. YY13).

First, the healthcare industry is changing due to the introduction of a new market system. In YYYY the Netherlands deregulated the healthcare industry by introducing the ‘Healthcare Insurance Act’ (HIA). This HIA enables Dutch patients to freely choose between healthcare insurance companies, leading to competition on a combination of price and quality of care (Gaynor, Ho, and Town YY15; Maarse, Jeurissen, and Ruwaard YY1Y; van Weel, Schers, and Timmermans YY1Y). This competitive environment is not only present in the Netherlands but also in countries like Germany, Norway, the Russian Federation, and Switzerland (Gaynor, Ho, and Town YY15; Günther et al. YY1Y; Thomson et al. YY13). Due to this change, healthcare insurers are faced with increasing numbers of churning customers (Günther et al. YY1Y; Hoffman and Lowitt YYYY; Thomson et al. YY13). After the introduction of the HIA in the Netherlands, an upward trend in switching customers is seen from YYY7 to YY13, moving from 3.5% to 7.Y%, while it is currently stabilising around Y.5% (Maarse, Jeurissen, and Ruwaard YY1Y; Romp, Merkx, and Vektis YY17).

(9)

9 also in other countries, like the United States (US), these numbers are shockingly high, as in YYYY in the US almost 7Y% of the adults was defined overweight or obese (Stephens, Allen, and Dennison Himmelfarb YY11). Several contributing factors to this unhealthy lifestyle are lack of physical activity, lack of fruit and vegetables intake, smoking, alcohol consumption, and eating too much and unhealthy food (Dute, Bemelmans, and Breda YY1Y; Hebden et al. YY1Y; Jayanti and Burns 199Y; Stephens, Allen, and Dennison Himmelfarb YY11). These factors are present in the Netherlands as for adults 1Y% drinks too much, Y5% smokes, 33% has too little physical activity, and over 9Y% eats too little fruit and vegetables (Rijksinstituut voor VolkXezondheid en Milieu n.d.). Nowadays governments, healthcare professionals, and society are more and more aware of this unhealthy lifestyle and its consequences, and as a result they demand more attention to preventive health actions, whereby a role for companies is identified as well (Ricciardi et al. YY13; Rijksinstituut voor VolkXezondheid en Milieu n.d.; Stephens, Allen, and Dennison Himmelfarb YY11).

Third, in the last years internet searches for health related issues have increased tremendously, in YY1Y 5Y-YY% of US adults (including seniors) and over YY% of Dutch adults searching for health information online (Van De Belt et al. YY13; Korda and Itani YY13). Besides, the use of mobile apps for health purposes has increased tremendously, since thousands of apps for physical activity, nutrition, and health improvement exist and are used by society (Hebden et al. YY1Y; Middelweerd et al. YY1Y; Stephens, Allen, and Dennison Himmelfarb YY11). Health apps and/or the internet are possibly more effective than real-life interventions to change behaviour, as they cost less (or nothing) and are available at all times (Azar et al. YY13; DeSmet et al. YY1Y; Korda and Itani YY13; Stephens, Allen, and Dennison Himmelfarb YY11).

(10)

10 While X decided to introduce the X platform, in the literature not a lot is known about these interventions, their effectiveness, and the impact of marketing activities on these interventions (Hebden et al. YY1Y; Hoying, Melnyk, and Arcoleo YY1Y; Smith et al. YY1Y). For example, Harwood and Garry (YY15) state that engagement platforms are found to be effective at increasing customer engagement. However, the extent to which the firm can influence this engagement is a still unexplored, but important area of research (Harwood and Garry YY15). Besides, the effects of health-related apps are still unclear, as several researchers identified that a lot of these apps do not use behavioural change theories and consumers find it hard or impossible to decide which health-related app is likely to be effective (Azar et al. YY13; Dute, Bemelmans, and Breda YY1Y; Pagoto et al. YY13; West et al. YY1Y).

A first step will be taken to explore this broad unfamiliar topic of online health interventions by looking at the effect of direct marketing on participant’s engagement with a health program. Besides, a direct impact of customer characteristics, relational characteristics, and mass media on the health program engagement (HPE), and a possible moderating effect of these on the marketing-engagement relationship will be investigated. The specific research questions addressed in this paper are:

RQ1: What is the effect of direct marketing on a participant’s HPE?

RQY: What is the effect of customer characteristics, relational characteristics, and mass media on a participant’s HPE?

RQ3: What is the effect of customer characteristics, relational characteristics, and mass media on the direct marketing-HPE relationship?

To allow for the possibility of investigating the research questions identified, data of a large health program was needed, which was received from the X platform of X. On the generated dataset an empirical research was performed consisting of an exploring and a modelling stage in which several Poisson models were specified, estimated, and compared.

(11)

11 program and prevent them from falling back into old habits (DeSmet et al. YY1Y; Korda and Itani YY13; Smith et al. YY1Y).

The results of this study show that direct marketing, customer characteristics, and relational characteristics have a positive impact on HPE. On the other hand, mass media is not found to have a significant impact on HPE. Moreover, customer characteristics and relational characteristics are found to have an effect on the direct marketing-HPE relationship.

As a result, this research contributes to the literature by being one of the first large online health programs to be investigated and contributing to the gap on the impact of marketing activities on HPE, as discussed before. Additionally, the impact of customer characteristics, relational characteristics and mass media on HPE is considered. As the HPE topic is new, these effects are also barely investigated in the literature. Finally, this research takes an integral approach by including moderating effects. Next to the theoretical implications, this study contributes to practice by providing insights for managers and health program developers. Based on the results of this study, new health programs could be developed and current ones could possibly be improved. These health programs will hopefully result in a healthier society in the future.

(12)

12

2 Background

2.1 Literature review

Since the health intervention topic is relatively new, a literature review was performed to reach better understanding on the topic (Hebden et al. YY1Y; Hoying, Melnyk, and Arcoleo YY1Y; Smith et al. YY1Y). Articles that did not discuss the effects of one or more intervention(s) were not included in the overview provided in the two tables which can be found on the following pages. Table 1 provides details on five well-investigated real-life interventions; while Table 2 summarizes nine meta-analyses regarding health interventions and apps that were discovered during the literature review.

Based on the literature review, it can be concluded that intervention studies in general found positive results for the intervention for a broad number of factors related to improvement of people’s lifestyle. However, the meta-analyses show mixed results with several studies finding only small or no significant improvements. Besides, the meta-analyses indicate several factors that can enhance intervention effectiveness, like individual tailoring of the intervention and message, interventions focused at a specific health improvement, and personal goal-setting (DeSmet et al. YY1Y; Dute, Bemelmans, and Breda YY1Y; Fry et al. YY1Y; Korda and Itani YY13; Olsen and Nesbitt YY1Y).

2.2 X

(13)

13

Table 1: Review of 5 intervention studies

Intervention Authors What Aim of intervention Target group and size Immediate or retained results Results/Findings COPE Healthy Lifestyles TEEN program (COPE: Creating Opportunities for Personal Empowerment TEEN: Thinking, Emotions, Exercise, and Nutrition) Hoying, Melnyk, and Arcoleo (YY1Y) Melnyk et al. (YY13) Feasibility and efficacy of COPE Healthy Lifestyles TEEN program. Efficacy of COPE Healthy Lifestyles TEEN program. Alter faulty thinking and dysfunctional beliefs. Appalachian early adolescents (13/1Y years old; n=YY)

Appalachian middle adolescents (1Y-1Y years old; n=779) Intervention + control group

Pre- and post-intervention measures

Immediately after intervention + after Y months

Improvement found for: - Students’ anxiety - Depression

- Disruptive behaviour - Self-concept scores

- Healthy lifestyle behaviour scores Compared to control group:

- Immediate post-intervention outcomes ○ Higher number of steps per day ○ Lower BMI

○ Higher scores on social skills ○ Lower alcohol usage

- Y months post-intervention outcome: lower BMI, resulting in a smaller proportion of overweight teens CHIP (Complete Health Improvement Program) Kent et al. (YY13) Kent et al. (YY15) Long-term effectiveness of CHIP intervention. Influence of gender on CHIP intervention outcomes. Promote healthy diet, physical activity, and stress management techniques, in order to reduce the risk of chronic diseases. (n=YYY) (n=9Y5; 317 men, YYY women )  3 years after intervention

3Y days after start intervention and 3Y days after finishing intervention

Sustained improvement found in: - BMI

- Diastolic blood pressure (DBP) - Total cholesterol (TC)

- Triglycerides (TG)

- Improvement of all biometrics for men and women - Greatest reductions found for participants with highest baseline for BMI, systolic blood pressure (SBP), blood lipids, and FPG

(14)

14

Table 1: Review of 5 intervention studies (continued)

Intervention Authors What Aim of intervention Target group and size Immediate or retained results Results/Findings CHIP (Complete Health Improvement Program) (continued) Kent et al. (YY1Y) Influence of religion on CHIP intervention outcomes. Promote healthy diet, physical activity, and stress management techniques, in order to reduce the risk of chronic diseases. Total (n=7,17Y): Seventh-day Adventist (SDA) (n=1,5Y3) + non-SDA (n=5,YY3)

3Y days after start intervention/after intervention

- CHIP program effective for both groups - Non-SDA group larger reductions in: ○ BMI

○ Pulse ○ Blood lipids

- Majority of non-SDA group in highest risk classification showed an improvement of YY% or more, while only some in the SDA group showed this improvement

Lighten Up to a Healthy Lifestyle program (LU) Weight Watchers program (WW) Cobiac, Vos, and Veerman (YY1Y) Cost-effectiveness of intensive weight loss programs.

Weight loss via nutrition and physical activity.

Weight loss via low-calorie diet and physical activity. Overweight adults (Survey Y months (n=Y3Y); 1Y months (n=Y3) Overweight adults (n=119) After Y month intervention + follow-up after 1Y months After finishing Y month intervention

- Both programs produced small improvements in population health ○ Increase in intake fruit and vegetables (LU)

○ Increase in weekly walking time (LU) ○ Reduction in BMI (LU + WW)

- Cost-effectiveness ratios were high for both programs Looma Healthy Lifestyle Rowley et al. (YYYY) Effectiveness and longevity of community-directed program. Prevent primary and secondary obesity, diabetes, and cardiovascular disease. Aboriginal community in north-west Western Australia: High-risk individuals (n=Y9) + cross-sectional community samples (n=YYY) High-risk: followed Y years Cross-sectional: followed Y years (n=1Y5) and Y years (n=13Y)

- High-risk cohort:

○ Protection from increases in plasma glucose and triglycerides

○ No sustained weight loss

(15)

15

Table 2: Review of 9 meta-analyses

Authors Aim of meta-analysis

How Results/Findings

Brusse et al. (YY1Y)

Overview of social media and apps for health promotion interventions.

Analysis of 17 intervention studies, 7 systematic reviews, 5 Australian projects with significant social media health components, and Y mobile apps. Studies were categorized in:

- Smoking cessation - Sexual health

- Apps and social media programs with Indigenous focus

- Evidence of benefit social media and mobile apps is limited and scattered

- Most findings regarding smoking cessation relate to behaviour outcomes, but results are mixed

- Findings on sexual health relate to knowledge, attitude, behaviour, and health but results for each are mixed

Chou et al. (YY13)

Overview of impact on and utility of Social media and Web Y.Y for health promotion.

Analysis of 51Y publications, classified as: - Commentaries and reviews (n=YY7) - Descriptive studies (n=Y13)

- Pilot intervention studies (n=3Y)

- Findings in commentaries and reviews:

○ Social media/participative Internet is a powerful tool for interactive health promotion, as it increases reach and interactivity, lowers costs, and allows for personalised messages

○ Web Y.Y increases audience participation

○ Social media efforts can be achieved at lower cost than traditional media

○ Effectiveness of social media for health promotion may differ across populations, socioeconomic status, and health literacy levels

- Findings in descriptive studies:

○ Platforms, like blogs and social networking sites, have the potential to improve health - Findings in intervention studies:

○ Feasibility of social media and Web Y.Y may still be limited

○ Usability of social media and Web Y.Y are more promising, but more are needed to strengthen the validity of the conclusions

DeSmet et al. (YY1Y) Overview of effectiveness of serious digital games.

Meta-analysis of 5Y serious digital games promoting a healthy lifestyle. Y categories of health behaviours included:

- Healthy diet and physical activity - Healthy responsibility/maintenance - Social behaviour

- Mental health promotion

- Serious games have a small positive effect on healthy lifestyles and determinants of a healthy lifestyle (especially knowledge)

- Long-term outcomes were maintained for all outcomes, except behaviour - No differences in appeal of serious games is found regarding age or gender - Duration of the intervention did not impact the effectiveness

- The best serious games are:

(16)

16

Table Y: Review of 9 meta-analyses (continued)

Authors Aim of meta-analysis How Results/Findings Dute, Bemelmans, Breda (YY1Y) Explore how mobile apps can contribute to the promotion of healthy lifestyles.

Analysis of 15 studies, including 1Y unique apps. Categorization apps: - Healthy nutrition

- Physical activity

- Prevention of overweight

- Most apps were part of a prevention program - Promising techniques for apps are:

○ Self-monitoring and providing feedback on performance ○ Specific goal-setting combined with personal feedback messages

- A social function can possibly enhance users’ motivation by providing rewards and opportunities for social comparison

Fry et al. (YY1Y) Determine association between implementation of community-based health improvement programs and county-level outcomes.

Available data for period YYYY-YYYY was used as basis (before implementation health improvement program). Four programs, performed in 39Y counties, were analysed.

- Modest evidence for improvement in health outcomes by health improvement program - Modest reductions in percentage of population reported being in poor or far health/being overweight or obese, although not significant

- Programs that focused on a specific health outcome (e.g. weight) showed greater changes in health outcomes, but this might be at expense of other health outcomes (e.g. smoking)

Kok, Van Den Borne, and Mullen (1997) Overview of effectiveness of health education and promotion. Analysis of Y1 articles: - 1Y primary invention articles

- 7 secondary prevention and/or patient education articles

- Interventions generally have quite substantial effects ○ Effect size (ES) is Y.YY for primary inventions ○ ES is Y.Y9 for secondary preventions/patient education - Determinant effectiveness is planned and systematic application

- Increase in effectiveness can be achieved with learning principles like rewards and feedback

- Too few interventions focus on facilitating desired behaviour (e.g. reminders, financial stimuli, skills improvement)

- Effectiveness is influenced by: ○ Quality of the intervention ○ Relevance

(17)

17

Table Y: Review of 9 meta-analyses (continued)

Authors Aim of meta-analysis How Results/Findings Korda and Itani (YY13) Overview of effect new/social media on health promotion and behaviour change.

Broad environmental scan; no details provided.

- More than half of the adults in the US searches for health information - Effect of age:

○ Older people (> 3Y years) are more likely to participate in online wellness programs

○ Younger people (< 3Y years) are more likely to participate in social networking and blogging sites - Effect of gender:

○ Women are searching more for information on symptoms, treatments, medications

○ Men are searching more for information on vitamins, health insurance providers, physicians - Social media can have a positive effect on health knowledge, behaviour, and outcomes - Health interventions that have a strong theory base have a greater impact

- Web-based interventions are found to have a small, positive effect on empowerment - Most promising ways to deliver the message

○ Using tailored messages

○ Applying multiple, complementary delivery modes ○ Encourage engagement with web-based applications Olsen and Nesbitt (YY1Y) Overview of the effectiveness of health coaching.

15 articles were analysed. - Significant improvements found for (but only in YY% of the studies): ○ Nutrition behaviour

○ Physical activity behaviour ○ Weight management behaviour ○ Medication adherence

- The following features seem to increase program effectiveness: ○ Goal setting

○ Motivational interviewing

○ Collaboration with healthcare providers ○ Program durations of Y to 1Y months West et al.

(YY1Y)

Overview of health and fitness apps.

A total of 3,33Y apps were reviewed.

- Apps that cost > $Y.99 (compared to < $Y.99) are:

○ More scored as intending to promote health or prevent disease ○ More credible/trustworthy

○ More used personally or recommended

- Common topics in apps are: healthy eating, physical activity, personal health & wellness

- Less common topics in apps are: substance abuse, mental and emotional health, violence prevention and safety, sexual and reproductive health

(18)

18

3 Theoretical framework

Health promotion is defined by Michael P. O’Donnell (YYYY) as “the science and art of helping people change their lifestyle to move toward a state of optimal health”. Optimal health in this context is specified as “the process of striving for a dynamic balance of physical, emotional, social, spiritual, and intellectual health and discovering the synergies between core passions and each of those dimensions” (O’Donnell YYYY; World Health Organization YY1Y). Considering this definition, health promotion goes further than only educating people on their health, as it also includes an attempt to make people adopt healthy behaviours (Donovan YY11). Changing one’s lifestyle, the goal of a health program, can be achieved through “a combination of efforts to enhance awareness, increase motivation, build skills and, most importantly, to provide opportunities for positive health practices” (Craig Lefebvre and Flora 19YY; Donovan YY11; O’Donnell YYYY; World Health Organization YY1Y).

As shown in Background, there is a limited amount of literature available on the topic of health programs, especially on health program engagement (HPE); therefore the theoretical sections will partly use literature regarding loyalty programs. It is assumed that these topics are highly related because, a health program can make customers more loyal and prevent customers from churning (Couper et al. YY1Y; Hoffman and Lowitt YYYY). Besides, loyalty and health programs have a lot of similar characteristics like point collection for specific, favourable behaviour and redemption of points via rewards. There is also limited information on the direct marketing form emails, since emails increased in importance only recently. Therefore, another type of direct marketing, direct mailings, will be used since a similar effect is expected (Cheung YYYY; Feld et al. YY13; Kolsaker, Görtz, and Gilbert YY1Y). Of course, there are also differences between direct mailings and emails, like how the marketing communication is opened, but these distinctions are small (Cheung YYYY; Feld et al. YY13).

(19)

19 and (3) relational characteristics, which will be examined in section five. Finally, the last section will present the conceptual model based on the hypotheses in the other sections.

3.1 Health program

In this article a loyalty program definition is chosen that can also be applied to a health program; health programs are “long-term oriented programs that allow consumers to accumulate some form of program currency, which can be redeemed later for free rewards” (Dorotic et al. YY1Y; Dorotic, Bijmolt, and Verhoef YY1Y; Liu and Yang YYY9, p. 9Y). In the application to health programs customers accumulate the currency based on healthy behaviour, while in the case of a loyalty program currency is based on purchases (Dorotic et al. YY1Y; Dorotic, Bijmolt, and Verhoef YY1Y; Helf and Hlavacs YY1Y; Liu YYY7; Stephens, Allen, and Dennison Himmelfarb YY11).

Health programs are often developed in response to consumer needs, aiming to improve or control unhealthy behaviour(s) (Craig Lefebvre and Flora 19YY; World Health Organization YY1Y). A good program, being it a loyalty or health program, meets five important specifications. First, good participation requirements, which relates to automatic or voluntary enrolment and the convenience during the program, like automatic gathering of points, where higher convenience is found to increase the program’s appeal (Liu and Yang YYY9). Second, an appropriate point structure, which concerns how points can be collected and used (Liu and Yang YYY9). For this specification, it is important that the points needed for a reward are not set too high and point collection and amount is not random (Dorotic, Bijmolt, and Verhoef YY1Y; Liu and Yang YYY9). The third specification is rewards, which concerns the choice and availability of rewards. A good program should have a good reward ratio and a broad number of reward options should be available (Leenheer et al. YYY7; Liu and Yang YYY9). Fourth, the program should have ongoing, tailored marketing efforts, like personalized mailings (Dorotic, Bijmolt, and Verhoef YY1Y). Finally, the program should be long-term oriented, where both parties (the firm and the participant) should invest (Dorotic, Bijmolt, and Verhoef YY1Y).

3.2 Health program engagement (HPE)

(20)

20 important measure to consider, as it is found to have positive effects for the firm employing the health program as well as for the participant. For the firm, higher HPE will result in higher customer retention and positive customer actions, like word-of-mouth, resulting in financial and non-financial benefits (Bijmolt et al. YY1Y; Couper et al. YY1Y; van Doorn et al. YY1Y; Evanschitzky et al. YY1Y; Hoffman and Lowitt YYYY; Liu YYY7). For the health program participant, higher HPE is related with an improvement in the participant’s health, social achievement, and an increase in self-esteem (Couper et al. YY1Y; Harwood and Garry YY15; X X n.d., YY1Y).

HPE can be operationalised using the Recency, Frequency, and Monetary value (RFM) model (Bijmolt et al. YY1Y; Cheng and Chen YYY9). In the RFM model, customers are scored based on their value to or engagement with the program, and based on these scores customers can be grouped for, for example, direct marketing campaigns (Bijmolt et al. YY1Y; Cheng and Chen YYY9). Recency refers to “the interval between the time that the latest consuming behaviour happens and present”, or relating this to a health program it can be defined as the interval between the time of the last login or healthy behaviour and the present (Bult and Wansbeek 1995; Cheng and Chen YYY9, p. Y17Y). Frequency, on the other hand, is defined as “the number of transactions in a particular period”, translating this to a health program would entail the number of logins and activities in a particular period (Bult and Wansbeek 1995; Cheng and Chen YYY9, p. Y17Y). Finally, monetary value refers to the “consumption money amount in a particular period”, translating this to a health program perspective would entail consumption via point collection resulting in point redemption (Bult and Wansbeek 1995; Cheng and Chen YYY9, p. Y17Y). A shorter interval for recency, higher frequency, and higher monetary value will lead to higher scores indicating higher customer value and engagement (Cheng and Chen YYY9). The main disadvantage of the model is that it does not account for double counting , as recency, frequency, and monetary value are not always independent (Bult and Wansbeek 1995). However, this model is still highly appropriate as it is an easy tool for measuring engagement by including the most important categories (Bult and Wansbeek 1995; Cheng and Chen YYY9). Below the identified components of engagement with a health program, using the RFM model, will be discussed.

3.2.1 Logins (Recency & Frequency)

(21)

21 Wansbeek 1995; Cheng and Chen YYY9; Couper et al. YY1Y). Logins are found to be especially effective in case of gamification, which entails adding game elements like rewarding additional logins with extra points, including achievements, and sharing the participant’s status with its social network (Anderson, Burford, and Emmerton YY1Y; Harwood and Garry YY15; Helf and Hlavacs YY1Y).

3.2.2 Point collection (Monetary value)

In this article, participants of a health program earn currency, or points, for specific health-related behaviour, therefore it relates to the monetary value of the RFM model (Dorotic, Bijmolt, and Verhoef YY1Y; Helf and Hlavacs YY1Y). Although when a participant collects points these do not have immediate value, point collection is of high importance due to its psychological meaning to the participant (Liu YYY7). Gamification is also effective in enhancing point collection (Helf and Hlavacs YY1Y).

3.2.3 Point redemption (Monetary value)

Point redemption is another important component of HPE, because it is part of the monetary value of the RFM model. In this article, point redemption is defined as the conversion of points collected to a (free) reward (Liu and Yang YYY9). Rewards can be, for example, discounts or gifts (Dorotic, Bijmolt, and Verhoef YY1Y). Point redemption may increase HPE due to a stronger relation between the participant and the program, and an increase in the attitudinal commitment of the participant (Dorotic, Bijmolt, and Verhoef YY1Y).

(22)

long-22 term health program is considered, it is expected that the long-term reward behaviour effect is dominant and thus a positive effect HPE should be expected.

3.2.4 Activities (Recency & Frequency)

Activities in this paper are defined as a health program action taken by the health program participant, which could be, for example, a health related activity like exercising, eating healthy, answering a program related question, or uploading a food picture (Adams, Katz, and Shenson YY1Y; Boswell, Kahana, and Dilworth-Anderson YYYY; von Bothmer and Fridlund YYY5). Several researchers indicate that increased variety and simplicity of activities is related to higher HPE (Hebden et al. YY1Y; Kim, Kim, and Wachter YY13; Lentferink et al. YY17). Additionally, the literature review in Background showed that the overall goal of activities in a health program is to achieve behavior change causing the health program participant to improve his/her lifestyle. The literature also highlighted the importance of activities as a way of health program participants to show their engagement with the program (Harwood and Garry YY15; Kim, Kim, and Wachter YY13; Lentferink et al. YY17). Therefore, activities are the final important measure of engagement as it influences recency and frequency.

An additional important distinction is the difference between activities in real-life, online (via a website), or via an app as these result in different activities and engagement (Kim, Kim, and Wachter YY13). Researchers indicate that online health programs have several advantages compared to real-life programs: they have large reach, can be realised at low cost, personalisation is relatively easy, and they are available at all times (Azar et al. YY13; Couper et al. YY1Y; DeSmet et al. YY1Y; Korda and Itani YY13; Stephens, Allen, and Dennison Himmelfarb YY11). There is, however, one main disadvantage of online health programs: they are not sticky, resulting in a large proportion of participant drop-out, which has an influence on the health program effectiveness (Couper et al. YY1Y).

3.3 Direct marketing

(23)

23 (Cheung YYYY; Feld et al. YY13; Kolsaker, Görtz, and Gilbert YY1Y). There are several benefits of using email communications for the company and the receiver as it permits two-way communication; it allows for convenience, because both parties do not have to be present at the same time; it is a relatively cheap compared to other communication tools; has a large reach; and is often not found to be intrusive (Ellis-Chadwick and Doherty YY1Y; Huang and Shyu YYY9). However, companies often overestimate the impact of email campaigns on HPE, as Danaher and Rossiter (YY11) discovered that companies underestimate the easiness of email rejection, while overestimating the trustworthiness of an email. Therefore, studying the real effect of emails on HPE is an important research topic.

It is of high importance that a company develops an integrated marketing communications (IMC) program (Batra and Keller YY1Y; Danaher and Rossiter YY11). The American Marketing Association defines IMC as “a planning process designed to assure that all brand contacts received by a customer or prospect for a product, service, or organization are relevant to that person and consistent over time” (American Marketing Association YY17; Batra and Keller YY1Y). Several so-called C’s enable a company to establish such an IMC: (1) consistency or commonality, which is about in how far the messages on different channels have the same message and thus reinforce each other; (Y) complementarity, which entails that different communications should complement each other, which is in contrast with consistency and, therefore, balance between these two is needed; (3) cross-effects, which entails that prior exposure to message on one channel can enhance the effect of a message on another channel; (Y) coverage, which entails the part of the target group that is reached by via a message; (5) cost, which entails that the most effective and efficient IMC should be developed at the lowest cost possible; (Y) contribution, which entails the reach and effect of a specific channel that cannot be reached with any other channel; and (7) conformability, which is about whether or not a message works for everyone at every moment in time (Batra and Keller YY1Y; Danaher and Rossiter YY11).

3.3.1 Email recipient actions

(24)

24 communication depends on attention-grabbing factors, which are the sender and whether the communication headline can grab the recipients attention (Ellis-Chadwick and Doherty YY1Y; Feld et al. YY13; Rettie YYYY). When an email communication is opened this indicates some interest of the recipient for the communication, which might have a positive outcome on HPE, leading to the following hypothesis:

Hypothesis 1: When a customer opens an email this will have a positive effect on (a) total HPE and its components: (b) activities, (c) logins, (d) point collection, and (e) orders.

After opening the email, the ultimate goal of the email is to get a response from the recipient, which will be measured via a click in the email on a graphic or hyperlink (Cheung YYYY; Feld et al. YY13). Following van Diepen, Donkers and Franses (YYY9), it is assumed that the response of a recipient to the email is immediately after opening. A click or response is only achieved with attention-sustaining factors, like personalization, relevance of the message, and an appealing layout (Brandal and Kent YYY3; Rettie YYYY; Wilson, Hall-Phillips, and Djamasbi YY15). Besides, a response or click is found to increase awareness, enhance loyalty, and engagement (Brandal and Kent YYY3; Ellis-Chadwick and Doherty YY1Y). Therefore, it is expected that a click will have a positive effect on HPE, leading to the following hypothesis:

Hypothesis Y: When a customer clicks in an email this will have a positive effect on (a) total HPE and its components: (b) activities, (c) logins, (d) point collection, and (e) orders.

A final possible action taken by an email recipient is an unsubscription or opting out of the communication. Opt out is defined as an unsubscription from all future email communications. Opting out of email communications is done when the emails are perceived intrusive, irrelevant to the recipient, or annoying (Cases et al. YY1Y; Dorotic, Bijmolt, and Verhoef YY1Y; Feld et al. YY13). This might have a negative impact on attitudes towards the health program, HPE, and possibly even result in termination of the relationship (Cases et al. YY1Y). This leads to the following hypothesis:

Hypothesis 3: When a customer opts out of email messages this will have a negative effect on (a) total HPE and its components: (b) activities, (c) logins, (d) point collection, and (e) orders.

3.3.2 Goal of the email

(25)

25 the development of the customer relationship (Kivetz and Simonson YYY3). Using the literature the following categories are identified: welcoming, informative, reward, activation, and recruitment. First, when a participant enrolled in a health program it is important to make the participant feel welcome and appreciated, and inform him/her about the benefits of participation, like rewards and discounts, entertainment, social status, belonging, and a healthier live leading to less discomfort in the future (Dorotic, Bijmolt, and Verhoef YY1Y; Duffy 199Y; Kent et al. YY13; Leenheer et al. YYY7; Park and Kim YYY3).

Hypothesis Y: Welcoming emails will have a positive effect on (a) total HPE and its components: (b) activities, (c) logins, (d) point collection, and (e) orders.

Hypothesis 5: Informative emails will have a positive effect on (a) total HPE and its components: (b) activities, (c) logins, (d) point collection, and (e) orders.

When the customer actually participates in the engagement program it is important to keep the participant engaged. An email communication that can help keeping customer engaged are emails informing participants about their rewards, achievements, and points they collected (García Gómez, Gutiérrez Arranz, and Gutiérrez Cillán YYYY; McCall and Voorhees YY1Y; Park and Kim YYY3; Rettie YYYY). Besides, when customers seem to drop out of the health program, sending an activation email might involve them again, especially when inactivity is due to a participant forgetting about the program (Park and Kim YYY3). An activation email will re-attract or remind them to the program resulting in a positive effect on engagement, leading to the following two hypotheses (Kolsaker, Görtz, and Gilbert YY1Y; Park and Kim YYY3; World Health Organization YY1Y):

Hypothesis Y: Reward emails will have a positive effect on (a) total HPE and its components: (b) activities, (c) logins, (d) point collection, and (e) orders.

Hypothesis 7: Activation emails will have a positive effect on (a) total HPE and its components: (b) activities, (c) logins, (d) point collection, and (e) orders.

(26)

26 Hypothesis Y: Recruitment emails will have a positive effect on (a) total HPE and its components: (b) activities, (c) logins, (d) point collection, and (e) orders.

3.4 Mass media

Although email is a new media channel that is extensively used, traditional mass media remains highly important as well (Danaher and Rossiter YY11). Mass media differs from emails because it allows a company to reach a large number of people at the same time (Dixon et al. YY1Y). Several different channels exist within the mass media, each with its own purpose. Television campaigns are used to attract attention and to keep people engaged to your program, radio campaigns enable your program to stand out, print campaigns can provide more details on your program, and internet campaigns can be more personal on several aspects as message content and timing (Batra and Keller YY1Y; Wang YY11).

Dixon et al. (YY1Y) and Donovan (YY11) found that mass media is also effective for health campaigns about topics like healthy weight, anti-smoking, and healthy lifestyle. These campaigns are especially effective when one discusses the problem (‘why’) and another the solutions (‘what’ and ‘how’) (Dixon et al. YY1Y). Dovey et al. (YY17) also indicate that television campaigns on healthy food are found to be effective, since they increase fruit and vegetable intake. Combined these findings lead to the following hypothesis:

Hypothesis 9: The mass media channels, (1) TV, (Y) radio, (3) print, and (Y) online will have a positive effect on (a) total HPE and its components: (b) activities, (c) logins, (d) point collection, and (e) orders.

(27)

27 Hypothesis 1Y: When the mass media channels, (a) TV, (b) radio, (c) print, and (d) online support the message on the health program conveyed via email, there is cross-channel integration, resulting in a positive effect of the mass media channels on the HPE-email relationship.

3.5 Customer characteristics

Customer characteristics are of large influence on loyalty programs, and likely health programs, being it on enrolment or participation (Liu and Yang YYY9). Segmentation of customers can be done on, for example, attitudinal factors, behavioural factors, or personal characteristics (Liu and Yang YYY9). Previous studies found effects based on different groupings of consumer characteristics, indicating the importance to treat customers as heterogeneous instead of one homogeneous group, while it matters less how you define customer heterogeneity (Craig Lefebvre and Flora 19YY; Liu and Yang YYY9). Currently, in health promotion programs the heterogeneity of the target group is not well investigated and people are treated homogeneous, which undermines the effectiveness of the health promotion (Craig Lefebvre and Flora 19YY). Therefore, including customer heterogeneity will have a positive impact on potential reach, effectiveness of the health promotion, receptivity of the target group, and engagement (Craig Lefebvre and Flora 19YY).

3.5.1 Additional customer characteristics

A large amount of literature on the direct effect of customer characteristics on health programs is available. However, not all literature has the same findings. For example, von Bothmer and Fridlund (YYY5) provide evidence that women are more engaged with health programs than men, while Kent et al. (YY15) found that men are more engaged with health programs. Additionally, being a couple is found to increase HPE (Kent et al. YY15). However, several articles find no effect for customer characteristics like age, gender, household size, and income (DeSmet et al. YY1Y; Leenheer et al. YYY7). Therefore, the following general hypothesis is presented for all often used customer characteristics (Adams, Katz, and Shenson YY1Y; Craig Lefebvre and Flora 19YY; Smedley and Syme YYY1):

(28)

28 Besides a direct effect of customer characteristics on HPE, customer characteristics are expected to also have a moderating effect on the email-HPE relationship. Based on Dorotic et al. (YY1Y) one would expect that when health program participants are older and/or wealthier the email-HPE relationship is strengthened. On the other hand, Dovey et al. (YY17) discuss a negative effect of age on the direct marketing-HPE relationship. Because, again current literature is not conclusive, a more general hypothesis is presented (Adams, Katz, and Shenson YY1Y; Craig Lefebvre and Flora 19YY; Smedley and Syme YYY1):

Hypothesis 1Y: The strength of the email-HPE relationship differs for participants with different (1) age, (Y) gender, (3) family composition, (Y) income level, (5) social class, and (Y) educational level.

3.5.2 Health characteristics

Because this article investigates a health program, a more appropriate consideration is a distinction in customer characteristics based on health characteristics. Motivication (a consultancy company) identified eight mentality groups based on status and social standards, which can be found in Appendix 1 (Motivication International YY1Y). To identify the mentality groups, several customer characteristics are taken into account, like educational level, gender, personal views, wealth, and origin (van Duuren YY15; Motivication International YY1Y; Zuiderduin YY1Y). These mentality groups are applicable to several industries, however, for the health industry an additional model is made, combining some of the eight mentality groups in five vitality groups: achievement-oriented socialisers, balance seekers, flexathletes, recreational movers, and sportwatchers (for more detail see Appendix 1). Because these vitality groups rely on several more general health characteristics, discussed in the previous section, two similar hypotheses can be stated:

Hypothesis 13: (a) total HPE and its components: (b) activities, (c) logins, (d) point collection, and (e) orders differ between the vitality groups.

Hypothesis 1Y: The strength of the email-HPE relationship differs between vitality groups.

3.6 Relational characteristics

3.6.1 Length of relationship

(29)

29 making customers more likely to participate in a health program of the company, leading to the following hypothesis (van Doorn et al. YY1Y; Dorotic et al. YY1Y):

Hypothesis 15: The length of the relationship with the health insurer has a positive effect on (a) total HPE and its components: (b) activities, (c) logins, (d) point collection, and (e) orders.

Huang and Shyu (YYY9) provide evidence that when a person is more loyal to the firm (due to a longer firm relationship) email campaigns are found to be more effective. Additionally, when the relationship with the firm increases the participant trusts the company more and has a positive attitude towards the firm, which both enhance email effectiveness (Cases et al. YY1Y). Combining these findings, the following hypothesis is derived:

Hypothesis 1Y: The length of the relationship with the health insurer has an enhancing effect on the email-HPE relationship.

The second relationship length investigated in this paper is the length of time a customer is enrolled in the health program. A high amount of health program literature indicates a high dropout rate of participants from a health program when time after enrolment increases (DeSmet et al. YY1Y; Kent et al. YY15; Melnyk et al. YY13; Smith et al. YY1Y). Reasons for this high dropout rate are: lack of interest or motivation, and reduction in perceived benefits (Couper et al. YY1Y; Smith et al. YY1Y). Additionally, to be thorough a moderating relationship of relationship length on the email-HPE relationship will be investigated as well, leading to the following hypotheses:

Hypothesis 17: The length of the relationship with the health program has a negative effect on (a) total HPE and its components: (b) activities, (c) logins, (d) point collection, and (e) orders.

Hypothesis 1Y: The length of the relationship with the health program has a negative effect on the email-HPE relationship.

3.6.2 Intensity of relationship

(30)

30 YYY7; Liu YYY7; Park and Kim YYY3). Additionally, to be thorough a moderating relationship of relationship intensity on the email-HPE relationship will be investigated as well:

Hypothesis 19: The intensity of the relationship with the health insurer has a positive effect on (a) total HPE and its components: (b) activities, (c) logins, (d) point collection, and (e) orders.

Hypothesis YY: The intensity of the relationship with the health insurer has a positive effect on the email-HPE relationship.

3.7 Conceptual model

(31)

31

(32)

32

4 Data

4.1 Data description

During this descriptive quantitative research the main relationship of concern is the effect of emails on HPE. Hereby, the emails are categorized based on the email goals identified in Theoretical framework, and for each email category different actions are included. Besides, mass media, customer characteristics, and relational characteristics variables will be included to investigate their impact on HPE and the email-HPE relationship. Data from the X health program of X, will be used to study these relationships. The X program was developed to stimulate participants to improve their lifestyle (X X n.d., YY1Y). Shortly summarized, X is an online platform introduced in YY1Y, which was complemented with an app in October YY17. With X, participants can (1) enter activities, read articles, set goals, and include friends; (Y) answer questions and get information from an online coach; and (3) get a daily “fitscore” based on all kind of aspects measured by using the platform (X X n.d., YY1Y).

For a proper analysis of the effects on HPE a longer time period should be taken, as HPE will change slowly over time. Besides, when a short time period is taken a marketing effort could result in an increase in HPE, but it is unclear whether this effect is lasting (van Diepen, Donkers, and Franses YYY9; McCall and Voorhees YY1Y). Therefore, an observation period of 15 months was taken resulting in a panel structure in the data; more specifically weekly data from the 1 of January YY17 until 31 March YY1Y (exactly Y5 weeks) was collected.

4.2 Data preparation

(33)

33 Theoretical framework, resulting in all hand-made variables. Below, the data extraction process and the construction of the variables will be discussed.

4.2.1 Data extraction

During the extraction process, variables from different environments within the data warehouse were used. More specifically, four environments were used: one to extract email information, one to extract mass media information, one to extract all information gathered with the X program, and one with customer and relational information gathered from people who are insured at X. As stated before, from the tables within these environments only the variables that were needed for this research were extracted.

During this extraction process, several pre-selection criteria were included to ensure a relevant and useful dataset. The first criterion is already discussed in Data description, namely the observation period from the 1st of January YY17 until the 31st of March YY1Y. A second pre-selection criteria limited the X participants to participants that where a member of the X program for the complete observation period, meaning that people who dropped out before the 31st of March YY1Y or joined after the 1st of January YY17 were excluded from the dataset. This criterion was chosen to ensure complete longitudinal observations in the study, easing the investigation of relationships. A third criterion limited the dataset to ‘real’ X participants (as discussed in Data description). Because the X program activates every member on a health insurance policy, two additional conditions were included: (1) the X participant should have logged in at least once before the 1st of March YY1Y and (Y) the X participant should have accepted the conditions of participation before the 1st of March YY1Y.

(34)

34 4.2.2 Data transformation

During the extraction of data from the data warehouse of X, most variables were already adapted to a convenient format for the current research. In Table 3 an overview of the variables, their operationalisation and, if applicable, formula are given. The most important transformations to the variables and the reasons for transformation will be shortly discussed.

The first large data transformation concerns the HPE variables. For all HPE components identified in Theoretical framework a variable was created. It is important to note that the point redemption variable, as identified in 3.2.3, was operationalised by considering the amount of products bought instead of points that are redeemed. Additionally, total HPE is investigated by consideration of the engagement score variable. In the data warehouse of X daily records are found for all engagement variables, which were transformed into weekly observations to limit the size of the dataset and to reduce missings (as some days were missing due to a technical issue or error). Transformation to weekly observations for all HPE variables, except the engagement score, was done by taking sum of all daily observations as it is most logical to look at the total of, for example, the logins during a week. The engagement score, on the other hand, was calculated by taking the average of the daily records as this variable is a variable created by X which includes several components resulting in a score between Y and 1YY (see Appendix 0).

(35)

35 The third group of transformations were performed on the mass media variables. The mass media variables contained information for each campaign of a mass media type for each week. It was decided to take the total of all mass media campaigns of a specific type for each week in order to limit the variables in the dataset, since mass media variables are only included complementary to the email variables and are not the variables of main concern. The mass media types TV, radio, and print are accounted for in GRPs, gross rating points, which entails a reach of 1% of the target group, which X defined as everyone aged 1Y or older. There is no maximum to these GRPs as people can also be reached more than once. Online mass media, on the other hand, is counted based on the number of online impressions. Due to the large numbers of impressions in some of the weeks, it was decided to make an additional transformation to this variable and counting it per 1YY,YYY online impressions.

The customer and relational variables were transformed as well. A first transformation in these variables concerned a limitation to the amount of categories for each variable, as in the data warehouse of X a high amount of categories are present. For example, the variable family composition existed of 1Y categories and was reduced to Y categories. Besides, the character length of the categories was reduced as lengthy character categories increase the processing size of a dataset tremendously. The total amount of characters allowed in the extracted variables was 15 characters, but most variables are no longer than 1Y characters. It is important to mention the vitality variable separately, as this variable was hand-made based on the mentality group, as discussed in 3.5.2.

Finally, after all data was extracted from the data warehouse and transformed to the desired format the data has been anonymised, by changing the variable names and the IDs. This reduces the risk of a data leak and prevents tracing results back to specific X participants which is conform to privacy policies.

Table 3: Overview of used variables, including the type, operationalisation, and formula for each variable

Variable Operationalisation Formula

ID (i) Numerical variable indicating a X participant.

Week (t) Numerical variable indicating the week number for each week in the observation period from the 1st of January, YY17 until the 31st of March, YY1Y.

IF year = ‘YY1Y’ then t=t+5Y Day (d) Auxiliary variable indicating a specific day in the observation

(36)

36

Table 3: Overview of used variables, including the type, operationalisation, and formula for each variable (continued)

Health program engagement (HPE) variables

Health program engagement (HPEite)

Dependent count variable, indicating the amount of engagement e for participant i, for each week of observation period t. Where e can be the number of logins on the internet, the number of collected points, the number of orders, or the number of activities (total, active, passive, X app, other).

𝐻𝑃𝐸𝑖𝑡𝑒 = ∑ 𝐻𝑃𝐸𝑖𝑑𝑒 7 𝑑=1 Health program engagement score (HPESite)

Dependent count variable, indicating the engagement score e for participant i, for each week of observation period t.

(The engagement score of each day is added together and then divided by the number of observations in the week, as sometimes a day is missing.) 𝐻𝑃𝐸𝑆𝑖𝑡𝑒 = ∑7𝑑=1𝐻𝑃𝐸𝑆𝑖𝑑𝑒 𝑑𝑡𝑜𝑡,𝑖𝑡 Email variables Email opens (EOimt)

Numerical variable indicating the amount of opens of email m by participant i in a week, for each week in observation period t. Where m can be a welcoming, points, recruitment, activation, informative, or birthday email.

𝐸𝑂𝑖𝑚𝑡 = ∑ 𝐸𝑂𝑖𝑚𝑑 7 𝑑=1 Email clicks (ECimt)

Numerical variable indicating the amount of clicks of email m by participant i in a week, for each week in observation period t. Where m can be a welcoming, points, recruitment, activation, informative, or birthday email.

𝐸𝐶𝑖𝑚𝑡= ∑ 𝐸𝐶𝑖𝑚𝑑 7 𝑑=1 Unsubscribe email (EUmt)

Binary variable indicating whether (1) or not (Y) participant i unsubscribed from the X mailings, for each week in observation period t. When subscribed in the previous week, the following week(s) will also get a value of 1.

Mass media variables

Gross rating points TV (GRP_TVt)

Numerical variable indicating the total GRPs for TV campaigns, for each week of observation period t. Where one GRP means a reach of one percent of the target group (1Y+).

𝐺𝑅𝑃_𝑇𝑉𝑡

= ∑ 𝐺𝑅𝑃_𝑇𝑉𝑑 7

𝑑=1

Gross rating points radio (GRP_radiot)

Numerical variable indicating the total GRPs for radio campaigns, for each week of observation period t. Where one GRP means a reach of one percent of the target group (1Y+).

𝐺𝑅𝑃_𝑟𝑎𝑑𝑖𝑜𝑡

= ∑ 𝐺𝑅𝑃_𝑟𝑎𝑑𝑖𝑜𝑑 7

𝑑=1

Gross rating points print (GRP_printt)

Numerical variable indicating the total GRPs for print campaigns, for each week of observation period t. Where one GRP means a reach of one percent of the target group (1Y+).

𝐺𝑅𝑃_𝑝𝑟𝑖𝑛𝑡𝑡 = ∑ 𝐺𝑅𝑃_𝑝𝑟𝑖𝑛𝑡𝑑 7 𝑑=1 Online impressions (Onlinet)

Numerical variable indicating the total online impressions per 1YY,YYY units, for each week of observation period t.

𝑂𝑛𝑙𝑖𝑛𝑒𝑡

=∑ 𝑂𝑛𝑙𝑖𝑛𝑒𝑑

7 𝑑=1

1𝑌𝑌 𝑌𝑌𝑌

Customer characteristics variables

Vitality (VITi) Categorical variable indicating the vitality group (see Appendix 1)

of participant i on the 15th of January YY1Y. With possible categories: achievement-oriented socialisers (AS), balance seekers (BS), flexathletes (FA), recreational movers (RM), and sportwatchers (SW).

Age (Agei) Numerical variable indicating the age (in years) of participant i on

the 15th of January YY1Y.

Gender (Genderi) Categorical variable indicating the gender (M/V) of participant i on

the 15th of January YY1Y. Family composition

(FCi)

(37)

37

Table 3: Overview of used variables, including the type, operationalisation, and formula for each variable (continued)

Customer characteristics variables (continued)

Income level (Incomei)

Categorical variable indicating the expected income level of participant i on the 15th of January YY1Y (based on postal code). With possible categories: minimum, lower than modal, modal, higher than modal, unknown.

Social class (SCi) Categorical variable indicating the expected social class of

participant i on the 15th of January YY1Y (based on postal code). With possible categories: A, B1, BY, C, D, unknown.

Educational level (Educationi)

Categorical variable indicating the expected educational level of participant i on the 15th of January YY1Y (based on postal code). With possible categories: LO, VMBO, HAVO, VWO, MBO, HBO, WO-PhD, unknown.

Relational characteristics variables

Relationship length X (RL_Mi)

Numerical variable indicating the amount of months participant i is insured at X since YYYY on the 15th of January YY1Y.

Relationship length X (RL_Xit)

Numerical variable indicating the amount of weeks participant i is a member of the X program, for each week in observation period t. (To create this variable the amount of weeks on 1-1-YY17 was taken for week 1, and the following weeks were calculated by adding the weeknumber minus one)

RL_Xit =

RL_Xon1-1-YY17 +

(t-1)

Relationship intensity X (RI_Mi)

Categorical variable indicating the intensity of the relationship between participant i and X on the 15th of January. With possible categories: 1 (basic health insurance only), Y (basic health insurance and one additional package; namely dentist or additional insurance), 3 (basic health insurance and both additional packages).

4.3 Descriptive statistics

4.3.1 Graphical examination data

CONFIDENTIAL + ANONIMISED

CONFIDENTIAL

+ ANONIMISED

The new role of health insurance companies in a

health-focussed society

Investigating the effectiveness of direct marketing on health

program engagement

CONFIDENTIAL

The new role of health insurance companies in a

health-focussed society

Investigating the effectiveness of direct marketing on health

program engagement

Management summary

Acknowledgements

Table of contents

1 Introduction

2 Background

2.1 Literature review

2.2 X

3 Theoretical framework

3.1 Health program

3.2 Health program engagement (HPE)

3.3 Direct marketing

3.4 Mass media

3.5 Customer characteristics

3.6 Relational characteristics

3.7 Conceptual model

4 Data

4.1 Data description

4.2 Data preparation

4.3 Descriptive statistics