How to motivate and activate users of health apps by email Investigating the effectiveness of six interventions on promoting health program usage

(1)

How to motivate and activate users of health apps by email

Investigating the effectiveness of six interventions on promoting health

program usage

(2)

How to motivate and activate users of health apps by email

Investigating the effectiveness of six interventions on promoting health

program usage

Master Thesis Marketing Management University of Groningen Faculty Economics & Business

Department Marketing

PO Box 800, 9700 AV Groningen (NL)

Joost de Graaf

Verlengde Frederikstraat 8a, 9724 NE Groningen (+31) 6 23 04 87 27

j.de.graaf.7@student.rug.nl

S2335298

1st supervisor:

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

2nd supervisor:

Dr. J.T. Bouma (j.t.bouma@rug.nl)

External supervisor (Healthcare insurance company): Dr. E. H. Noppers (noppers.e@healthcare insurance company.nl)

(3)

Management summary

Nowadays, health care insurer Healthcare insurance company is focusing a lot on their healthy lifestyle program Health app (English translation: …). With approximately 55.000 active users, the Health app app is one of the leading health apps in The Netherlands. Monthly, around 2000 new people start using the Health app app. Healthcare insurance company already has discovered that 75% of the new users stop using the Health app app during the first month. Moreover, they know that if users track activity in the Health app app two times, the chance of long term use is very high. But how can Healthcare insurance company motivate and activate users to track activity in the Health app app twice?

This two by three study investigates six methods to motivate and activate new users by email to track their activity in the Health app app. Activation methods are a text only email or an email with a video. The motivation methods are explaining the fitscore option in the Health app app, explaining the option to score points and get rewards in the Health app app and explaining the option to take on social challenges in the Health app app. All the new users between October 24th and November 14th are selected for this research. This leads to six different groups of 330 or 331 members and a control group of 104 members. All the new users, except the members of the control group, received an email on November 21th or November 22th with one of the two activations combined with one of the thee motivations. After following the new members two and a half weeks, the dependent variables, a) amount of tracked activities of the new members and the b) tracked energy expenditure of the new members are measured.

No big differences appear in the amount of tracked activities and tracked energy expenditure. All the different groups, motivation methods and activation methods have no significant difference comparing on both dependent variables. The small difference found, hints in the direction that text only emails are more effective than emails with an embedded video. However, the difference was not significant. Thereby, comparing the different motivation methods on their mean of both dependent variables, this research hints that emails with promoting the rewarding system are the most effective, followed by emails with promoting the social challenges and the emails with promoting the fitscore.

(4)

to new users sooner after registration leads to more tracked physical activity, but sooner sending an email to new users not necessarily leads to more tracked energy expenditure. After testing the model, this research discovers that young people are less sensitive for the different activation and motivation methods than the other life phase groups.

(5)

Preface

This is my master thesis. It is completed now in on January 10th. I, Joost de Graaf, have enjoyed this period, and I would like to thank Healthcare insurance company a lot. They gave me the opportunity to research a very practical problem in a field that I find extremely interesting. I would like to specifically thank Ernst Noppers. The cooperation with him was very good. He helped me a lot, he taught me a lot and we discussed a lot about this research. Ernst was always available for questions and knows all the ins and out of academic research. I could not wish for a better supervisor. I would also like to thank Koert van Ittersum. He guided me through the complete process very well. He was also always available for questions and feedback and always responded very quickly. Thereby, I would like to thank Jelle Bouma, director of the RUGCIC. He made the first contact with Healthcare insurance company and made it possible to write my thesis at this company. Lastly, I would like to thank my parents and girlfriend for supporting me a lot. They were always there for me.

(6)

1. Introduction

In 2016, almost 50% of the adults were overweight in The Netherlands. Compared to 33% in 1900, this is a huge increase (Volksgezondheidszorg, 2018). European adolescents and students tend to have low levels of physical activity and to eat unhealthy foods. Thereby, the prevalence of overweight and obesity has increased, which poses a public health challenge (Dute, Bemelmans and Breda, 2016). The absence of sufficient less physical activity has been a problem for years. For example, only 20% of college students participate in physical activity according to the norm (Douglas et al., 1997). The current norm of healthy movement is 150 minutes a week of moderately intensive exercise on different days in the week (Gezondheidsraad n.d.).

Evidence shows that physical activity improves mental and physical health, for example, cycling (Biddle & Asare, 2011; Food & Food, 2001). Oja et al. (2011) have reviewed 16 different studies about cycling. They all show a positive relationship between cycling and health and functional benefits for young boys and girls. Cycling also leads to improvements in cardiorespiratory fitness and disease risk factors as well as significant risk reduction for all- cause and cancer mortality and for cardiovascular, cancer, and obesity morbidity in middle- aged and elderly men and women (Oja et al., 2011).

Given this evidence, one would assume that society in general would keep the norm of physical activity. However, this is clearly not the case at this point in time. In developed countries, children, adolescents and (young) adults are more and more at risk of becoming obese and/or living unhealthy, resulting in the emergence of several illnesses (Hebden et al., 2012; Dute, Bemelmans and Breda, 2016; de Souza Andrade et al., 2018). Colditz (1999), researched the direct costs of inactivity and obesity. According to his research, physical inactivity and obesity accounts for 9,4% of the national health care expenditures in the United States.

This all leads to new trends. The first one is that companies are focusing more on health and try to change behavior of their employees and customers to make them more physically active and to help them make healthy choices (Malik et al., 2014; Sallis et al., 2015). Thereby, the second trend is that mobile phone apps seem to be a promising monitoring tool in health promotion strategies (Dute, Bemelmans and Breda, 2016).

(9)

days, more energetic employees, etc. (Van den Heuvel et al., 2005). An industry wherein these kinds of promotions are very interesting, is the world of healthcare insurers. One of their goals is to make their insured people healthier. This leads to less healthcare costs, which makes healthcare payable (Healthcare insurance company n.d.).

One of the leading healthcare insurers in this field is Healthcare insurance company. Healthcare insurance company is, with 2.1 million insured people, the fourth Dutch healthcare insurer. One of their three main goals in the Healthcare insurance company 2022 plan is to make their insured people healthier. They promote physical activity and a healthy lifestyle among other things via their Health app (English translation: ‘HealthyTogether’) application for mobile phones. This program, that was started as a loyalty program in 2013, helps people making healthy choices. It provides tips about physical activity, food and relaxing. Thereby, it has an online coach who supports the user to reach personal goals and the app has an option to track physical activity. In the app a lot of different kinds of activities can be tracked, for example, hiking and cycling. Other physical activity tracking apps, like Strava and Runkeeper can be connected to the Health app app.

If the users are physical active and they track this in the app or a connected app, they earn points. If they are insured with Healthcare insurance company, they get with these points discounts on products or services. If they are not insured with Healthcare insurance company, they can use the app, but get other, less attractive discounts.

Another important implemented option is the fitscore. This score expresses the health of the users with a number between the 1 and 1000. The higher the score, the fitter the user is supposed to be. Users can improve their score by tracking physical activity or chatting with the online coach. During these chats, the coach asks the users for example questions about their lifestyle, eating pattern and mental health.

(10)

As said above, Healthcare insurance company is one of the companies who is focusing on health. Not only the health of their employees, but mainly the health of their insured people. They already made an intervention (Health app ) to improve the health of their customers. But the next problem is activating and motivating people to use the app on their mobile phone. At this moment the app has approximately 55.000 active users. Approximately 2000 new users are joining every month. Healthcare insurance company knows that if the users track two times an activity, they have a very big chance to become an active user. But at this moment, the problem is that only 24,1% of the new users use the app more than one month. Different aspects influence the duration of app usage, but because Healthcare insurance company has just started with the Health app app, they do not exactly know what influences the usage of their app. According to Allen et al. (2013) users value accountability and structure the most. Thereby, users suggest a stronger emphasis on exercise and additional feedback. Adding to that, Bond et al. (2014) reported that real-time smartphone display and feedback significantly increased their motivation to engage in physical activity. Healthcare insurance company wants to increase the usage of the Health app app. But what kinds of interventions should the company apply to increase the usage? And how should Healthcare insurance company activate and motivate the Health app users more? This research tests interventions to increase a) the amount of tracked activities in the Health app app and b) the tracked energy expenditure in the Health app app of the new users. The goal of this research is to find out what interventions work for activating and motivating new users to increase the amount of tracked activities and the tracked energy expenditure, so using the Health app app. This leads to the following research question:

RQ1 ‘’What email interventions can Healthcare insurance company best use to increase a) the

amount of tracked activities and b) the tracked energy expenditure of new users in the Health app app?’’

This research includes the following sub questions:

SQ1 ‘’Does explaining the social challenge option in emails increase a) the amount of tracked

activities and b) the tracked energy expenditure of new users in the Health app app?’’

SQ2 ‘’Does explaining the fitscore in emails increase a) the amount of tracked activities app

and b) the tracked energy expenditure of new users in the Health app app?’’

SQ3 ‘’Does explaining the rewarding system in emails increase a) the amount of tracked

(11)

Thereby, Healthcare insurance company uses only text in emails to explain options in the app at this moment. Healthcare insurance company employees have a strong suspicion that emails with a video are a more effective tool to explain options in the app than text only emails. Besides that, evidence shows that video tutorials should be more effective than text explanations (Shyu, 2000; Zhang et al., 2006). The current research investigates what manner of providing information more effective is. A stepwise explanation in text or a stepwise explanation with a video tutorial. This leads to the next sub question:

SQ4 ‘’Does a video tutorial in emails increase a) the amount of tracked activities and b) the

tracked energy expenditure of new users in the Health app app compared to text only emails?’’

Lastly, every new user of the Health app app is added to an email flow. In this flow Healthcare insurance company sends emails to the new users 1, 3 and 4 weeks after their registration in the Health app app. They have the suspicion that they have to activate the new users by an email as soon as possible after their registration in the app. This research investigates this suspicion, with the last sub question.

SQ5 ‘’Does sending an email to new users sooner after registration in the Health app app

leads to more a) tracked activities and b) tracked energy expenditure of new users in the Health app app?

This research is a big field experiment. Because of the scope of the complete research, this research tests three interventions to motivate people. These interventions are motivating people by explaining the content of the fitscore, by explaining rewarding users and by explaining using the social challenges. This research tests to activating methods, users activating with text emails and users activating with emails with a video.

(12)

2. Theoretical background

2.1 Physical activity

A lot of research is done concerning physical activity. Most reports show the positive impact of physical activity on health. In the current research, physical activity is defined as any bodily movement produced by skeletal muscles that results in energy expenditure. According to Caspersen et al. (1985) physical activity has four elements, it has bodily movement via skeletal muscles, it results in energy expenditure, the energy expenditure varies continuously from low to high and it is positively correlated with physical fitness. A lot of different methods identify the different types of physical activity. The simplest categorization identifies the physical activity that occurs while sleeping, at work and at leisure (Montoye, 1975). Being physically active leads to a higher physical fitness. Being physically fit has been defined as the ability to carry out daily tasks with vigor and alertness, without undue fatigue and with ample energy to enjoy leisure-time pursuits and to meet unforeseen emergencies (President's Council on Physical Fitness and Sports, 1971).

2.2 Mobile phone apps and health

Mobile phone applications seem to be a promising tool to improve people’s health. This can be done by, for example, increasing physical activity. Mobile phone apps are software applications that make it possible to run on smartphones. Smartphones are widely owned, can be used everywhere and at every time. This makes it possible to reach a lot of people. Also, it makes it possible to promote a healthy lifestyle and to help people when it comes to physical activities (Van Rooij & Schoenmakers, 2013). The newest technologies with constant GPS and internet connection make it possible to track activities everywhere (Dute et al., 2016), what makes these interventions available for a large group of people.

(13)

Moreover, the so called tele-health and other mediated approaches to health behavior change provide an empirically supported, convenient, and potentially lower-cost alternative for reaching large proportions of the public over a longer period (King et al., 2013). Multi- component interventions appear to be more effective than stand-alone app interventions, however, this remains to be confirmed in controlled trials (Schoeppe et al., 2016).

According to Kilpatrick et al. (2005), several motivations influence physical activity. All these motivations are described later. Recent research of Van de Wal (2018) commissioned by Healthcare insurance company, shows that three specific motivations influence physical activity. These are social, health and financial motivations. Van de Wal (2018) explains that the social motivation is doing something physical with others or competing with others. The health motivation is that people want to be physical active because they want to experience the health benefits, like feeling fitter, less sick, etc. The last driver is the financial motivation. People who are driven by this motivation like being physical active because they can earn points for discounts.

Healthcare insurance company has added in the Health app app interventions that tap in to the motivations described in the research of Van der Wal (2018). In the current research we link social motivation to the social challenges option in the Health app app, the health motivation to the fitscore option in the Health app app and financial motivation to the points what can be earned with physical activity. These points can be redeemed in the Health app web shop. These three motivation techniques and aspects are linked to relevant theories and are explained later.

2.3 Health promotion interventions

(14)

Another model what is used is the transtheoretical model (Prochaska & Velicer, 1997). This model estimates an individual readiness to adopt a new healthy behavior and guides the individual during the process with strategies.

Abraham and Michie (2008), have researched and developed 26 different types of behavior change techniques. These techniques are based on six different behavior change theories. These theories are: information-motivation-behavioral skills model, theory of reasoned action, theory of planned behavior, social cognitive theory, control theory, and operant conditioning. Abraham and Michie (2008), have described what types of theories are efficient to use in different kinds of apps.

Abraham and Michie (2008) discussed a lot of different apps in their article. For the current research the most comparable, is the Ak-Shen app. One of the theories they use in this app is the theory of self-monitoring. This app provides adolescents specific physical activity challenges. The app supports the user to activate it and to track their physical activity with GPS. Thereby, they add the theory of social comparison. The app makes it possible to share and compare their physical activity with other members. The effectiveness of these techniques is not measured. This is measured in the current research.

2.4 Social challenges and social motivation

(15)

under some circumstances, people prefer comparing themselves to others who are similar on attributes related to the dimension under evaluation (Taylor & Lobel, 1989).

Thereby, Festinger (1954) argued that people strive to be more capable than other persons with whom they can compare themselves. Other researchers, like Wheeler (1966) have argued that people prefer comparing with others who are slightly better than himself or herself. So, comparing with others should be done in groups wherein they can compare themselves with others who are the same in performance or a little bit better (Taylor & Lobel, 1989).

Franken & Brown (1995), have found that people select competitive situations for different reasons. For example, people join competition to increase their performance and have little or no regard for winning. For others winning is the most important thing. And for the last group competition provides the motivation for putting forth effort which presumably will lead to improved performance in the future.

Based on the literature described above, social challenges can improve the usage of the Health app app. This leads to the first hypothesis:

H1: Social challenges motivate users to track a) more activities and b) more energy consuming activities in the Health app app.

2.5 Gender differences in motivation

(16)

H2: The effect size of using social challenges as motivation tool for tracking a) more activities and b) more energy consuming activities in the Health app app is bigger for men than for women.

2.6 Health motivation and physical activity

In this research, the fitscore option in the Health app app is seen as a part of the intrinsic motivation. This is because of the fact that feeling healthy or feeling unhealthy is something that only the participating person can feel and experience. Deci and Ryan (1985) made a distinction between intrinsic and extrinsic motivation. Intrinsic motivation refers to doing something because it is inherently interesting or enjoyable for yourself. Extrinsic motivation refers to doing something because it leads to a separable outcome (Deci & Ryan. 1985). According to the cognitive evaluation theory, intrinsically motivated behavior is behavior which allows a person to feel competent and self-determined (Deci & Ryan, 1985). Individuals are said to be intrinsically motivated when they engage in an activity in the absence of extrinsic rewards or constraints (Nicholls, 1984, 1989). In task involvement, the individual engages in the activity for its own sake and experiences it as an end in and of itself (Nicholls, 1989), which is a fundamental element of intrinsic motivation (Deci & Ryan, 1985; Deci, Vallerand, Pelletier and Ryan, 1991). According to Richard et al. (1997), Deci, (1971) and Harackiewicz (1979) one of the ways to increase intrinsic motivation is to get positive feedback based on your actions.

Vallerand et al. (1992), have presented three types of intrinsic motivation. These are intrinsic motivation to know, intrinsic motivation to accomplish things and intrinsic motivation to experience stimulation. Intrinsic motivation to know can be defined as performing an activity for the pleasure and the satisfaction that one experiences while learning, exploring, or trying to understand something new. Intrinsic motivation to accomplish things can be defined as engaging in an activity for the pleasure and satisfaction experienced when one attempts to accomplish or create something. And lastly, intrinsic motivation to experience stimulation occurs when someone engages in an activity to experience stimulating sensations derived from one’s engagement in the activity (Pelletier et al., 1995).

(17)

The fitscore should increase the physical activity of the user because it increases the intrinsic motivation and it offers feedback in graphs. This leads to the next hypothesis:

H3: The fitscore motivates users to track a) more activities and b) more energy consuming activities in the Health app app.

2.7 Rewarding and physical activity

Rewarding is an effective tool to motivate people. Rewards are most effective if the reward fits with the interests of the person who must be motivated (Allen, 2006). Two types of rewarding are described in the literature. These are post-rewarding and pre-rewarding (Hartmann & Viard, 2008; Kopalle et al., 2012; Lewis, 2004). Researchers provide evidence of pre-reward effects in the short-term, when members must reach a spending threshold during a time-limited period to obtain a prespecified reward. The study provides support for positive post-reward effects when customers do not face binding deadlines and can choose the redemption timing and amount (Dorotic et al., 2014).

This study links rewarding as a part of extrinsic motivation. This is because extrinsic motivation shows behaviors that are engaged in to an end and not for their own sake (Deci, 1975). Originally, it was thought that extrinsic motivation referred to behavior that could only be prompted by external circumstances, for example rewards. However, Deci and Ryan (1985) and more researchers, like Ryan, Connell and Grolnick (1990), have discovered that different types of extrinsic motivation exist. These types are external regulation, introjection and identification. External regulation refers to behavior that is controlled by external sources, such as material rewards or constraints imposed by others (Deci & Ryan, 1985). With introjection, the formerly external source of motivation has been internalized such that its actual presence is no longer needed to initiate behavior. The last one is identification. This is when the individual values and judges the behavior as important, and that’s the reason why he or she performs it (Bartholomew et al., 2011).

Rewarding is supposed to be an effective tool to motivate users of the Health app app. The following hypothesis tests whether it really leads to more users:

(18)

2.8 Direct marketing

Direct marketing makes it possible for firms to target specific groups of potential consumers. Because it is possible to make it personal, direct marketing is more focused and sharper than the traditional ways of marketing (Kolsaker, Gortz and Gilbert, 2016; Feld et al., 2009). Currently, one of the most important ways of direct marketing is emailing. In 2009, direct emailing was responsible for one third of the direct marketing expenditures in most countries (Feld et al., 2009). According to the research of Nauta (2018), emails are an effective tool for increasing health program engagement. Specifically, informative emails have the biggest effect according to her paper.

Research of Van Diepen, Donkers and Franses (2009) and Boiarsky (2015) shows that direct mailing has the largest impact in the first week after sending. According to Van Diepen, Donkers and Franses (2009) the effect of emailing reduces to 74.3% in the second week and to 55.2% in the third week. But overall, emails have a direct positive impact on usage incidence. The fifth hypothesis tests if this is the case in this situation too:

H5: Activating users soon after downloading the Health app app leads to a) more tracked activities and b) more tracked energy consuming activities in the Health app app.

It is beneficial for loyalty program providers to leverage the information they have and target members with personalized emails (Blattberg et al., 2008; Lewis, 2004). Heijltjes, Muste, and Van Eeden (2016) investigate the effect of emailing on different days. They conclude that Monday, Wednesday and Friday are the best days to send an email. These days have the largest open and click rates.

(19)

H6: Video tutorials in emails lead to a) more tracked activities and b) more tracked energy consuming activities in the Health app app compared to text only emails.

2.9 Conceptual model

On the next page, in figure 1, the conceptual model is shown. This conceptual model is based on the hypotheses in the theoretical background. The independent variables are promoting health motivation, promoting financial motivation and promoting social motivation. According to the hypothesis, this research expects that all the motivations have a positive effect on the dependent variable, the a) amount of tracked activities and b) the tracked energy expenditure. The moderator in this research is the gender of the user. One of the hypotheses expects that men are more sensitive for social challenges than women. Another independent variable is the video or text explanation in the emails. This research expects that videos in emails are more effective than text only emails.

2.10 Conceptualization

In this research, the concept ‘a) amount of tracked activities’ is defined by the number of times a user tracked their physical activity in the Health app app. For the concept ‘b) Tracked energy expenditure’ all the activities the users track are measured. These activities are expressed in points.

(20)

Figure 1: Conceptual model

2.11 Summary

Below, the different motivation, theories and tools are shown.

Motivations (Van de Wal, 2018) Relevant theories Tools in Health app application 1. Health motivation Intrinic motivation Fitscore

Feedback

2. Financial motivation Extrinsic motivation Reward in points for discounts Rewarding systems

3. Social motivation Social comparison Social challenges Competition

Table 1: Overview theoretical background

Promoting health motivation (Fitscore) Video tutorials compared to text only Promoting financial motivation (Rewards) Amount of tracked activities Tracked energy expenditure Promoting social motivation (Social challenges)

Gender Days between

(21)

3. Methodology

Because this research is looking for main effects, it makes use of quantitative data. This gives the research the main effects and not the underlying reasons or needs.

This research has a two by three experiment design. The independent variables are the motivations (promoting health motivation, promoting financial motivation and promoting social motivation) and the activations (by an email with text or an email with video). Also, the activations are independent variables in the conceptual model. Lastly, gender is expected to have a moderating effect when it comes to promoting social motivation. The dependent variable is a) the amount of tracked activities in the Health app app and b) the tracked energy expenditure in the Health app app.

3.1 Population

For this quantitative field study, all the new users who are started using the Health app app between October 24th and November 21th are selected. This group is excluded from other emailing campaigns. Otherwise this group receives different emails that motivate and activate them, which would interfere with this research. The new users receive the email with the activation and the motivation on Wednesday November 21th or Thursday November 22th. Because of problems with the server, this research is forced to make use of two batches. The first batch of 1533 people will be sent on Wednesday. The participants in this batch are started using the Health app app between October 24th and November 14th. The second batch of 433 people will be sent on Thursday. The participants in this batch are started using the Health app app between November 15th and November 21th. Thereby, participants younger than 18 years old, users who passed away and users without an email address are excluded. Participants who have unsubscribed from emails are logically excluded too.

In total, this research makes use of 2089 participants. 5% do not receive an email, because this is the control group. The other 1985 participants are randomly assigned to one of the six groups. So, every group has 330 or 331 members.

(22)

Table 2: Experiment groups

So,

- group 0 is the control group

- group 1 is activated by a text email and motivated by the health motivation, - group 2 is activated by a text email and motivated by the financial motivation, - group 3 is activated by a text email and motivated by the social motivation, - group 4 is activated by a video email and motivated by the health motivation, - group 5 is activated by a video email and motivated by the financial motivation and - group 6 is activated by a video email and motivated by the social motivation.

As said, a control group of 5% is used. This control group is used to compare the control group with the other groups and to ensure the effect occurs because of the different motivations and activations. The control group consists of 104 participants.

For group three and six this research creates a challenge in the Health app app. The name of the challenge is ‘Kom in Beweging’. This challenge starts at the day of emailing. To make sure that the challenge is as accessible as possible, in the challenge energy expenditure is the measure in the ranking and all the automatic tracked activities count. The duration of the challenge in the app is four weeks, but the activities of the users are followed two and a half weeks after sending the email. This is determined because of the effect of an email which, according to Van Diepen, Donkers and Franses (2009) is the biggest in the first two weeks, and because of the time wherein this research should be done.

(23)

(24)

because of the options GIFs offer. A GIF is with less than 1 mb relatively small, it is basic, it opens and starts directly when the email is opened, it can be posted in a loop and it has almost no waiting or loading time. So, embedding a GIF in an email is easy to use and to open. A disadvantage is that adding audio is impossible. But this is an advantage too, because of the high open rates of Healthcare insurance company on mobile devices. It is not desirable that an email makes noise when it opens on a mobile phone. The different emails are added in the appendix.

After sending the emails, the activities of the users are followed. The activity is measured by the amount of tracked activities and the amount of tracked energy expenditure in de Health app app in the two and a half weeks after sending the email with one of the motivations and activations.

The personal characteristics that are available for the analysis of all the users are UserID, motivation method, date of subscribing and the batch. More information is available for some users. These users are insured with Healthcare insurance company. Of this group the date of birth, gender, stage of life group, mentality profile, vitality profile and province are available. The total output file is presented in the appendix.

No information about the participants is available to the researcher that could lead to identifying participants with respect to personal information (e.g. name, email, address).

3.2 Analysis

(25)

4. Results

4.1 Open rates

The first available results are the open rates of the different emails. The unique open rates of the different emails are approximately the same. The mean is 47.5%, the minimum is 42.1% and the maximum is 49.5%. The total number of emails send is 1976. This leads to 939 unique opens. Per test group are on average 157 participants who have opened. See appendix 3 for the complete table.

4.2 Tracked activities in the Health app app

Figure 3: Differences between groups

(26)

60 50 40 30 20 10 500 4.2 Data exploration

The first step is filtering the outliers from the data set. This is done by making a scatter plot of the dependent variables. The outliers give distorted picture of the data, so these values will be deleted. Thereby, it is unlikely that the users use the Health app app so much because of only one email. The scatterplot of the amount of energy expenditure is shown below. In this graph, it is clear that this data set has two outliers. These are red circled in the graph and deleted in the dataset.

Figure 4: Data set tracked energy expenditure

The scatter plot of the amount of tracked activities in the Health app app is shown below. In this graph are no outliers.

Figure 5: Data set amount of tracked activities

(27)

Thereby, in the data set no missing values are found. However, for some variables, like gender, age and lifestyle the data are only available for Healthcare insurance company customers. This information is available for 581 users. Incorrect values are not found.

Two boxplots are shown below in figure 6 and 7. These graphs show the dispersion of the dependent variables. The first boxplot shows the data of the amount of tracked activities. The second boxplot shows the data of the tracked energy expenditure. The boxplots show with all the data points above zero, the active users. Also, the complete box and the minimum and maximum non-outlier values are all 0, for every group.

(28)

Figure 7: Boxplot tracked energy expenditure

(29)

Group N a) amount of tracke d activitie s b) tracke d e ne rgy e xpe nditure Me an 0 Control group 104 0,31 4134 1 Text Health 330 0,5 11106 2 Text Financial 331 0,92 12867 3 Text Challenge 331 0,67 13682 Ave rage Te xt 992 0,70 12552 4 Video Health 331 0,53 10265 5 Video Financial 330 0,63 5654 6 Video Challenges 330 0,5 8621

Ave rage Vide o 991 0,55 8180

Ave rage He alth 661 0,52 10686

Ave rage Financial 661 0,78 9261

Ave rage Challe nge s 661 0,59 11152

Total Ave rage 661 0,63 10366

Table 3: Means of the different groups

Group N a) amount of tracke d activitie s b) tracke d e ne rgy e xpe nditure Std. De v 0 Control group 104 2,05 3,1854 1 Text Health 330 2,86 60990 2 Text Financial 331 4,61 65923 3 Text Challenge 331 3,26 69465 Ave rage Te xt 992 3,57 65459 4 Video Health 331 2,99 62316 5 Video Financial 330 4,58 32086 6 Video Challenges 330 2,60 40461

Ave rage Vide o 991 3,39 44954

Ave rage He alth 661 2,92 61653

Ave rage Financial 661 4,59 49005

Ave rage Challe nge s 661 2,93 54963

Total Ave rage 661 3,48 55207

(30)

4.3 Distribution

Discovering the type of distribution is an important step when choosing the right data analysis methods. Because the data contain a lot of 0s, the data set is probably not normally distributed. This is tested by a Q-Q plot and a histogram of both dependent variables.

This leads to the following plots.

(31)

Figure 9: Q-Q plot tracked energy expenditure

(32)

Figure 10: Histogram amount of tracked activities

(33)

The data in both histograms are not close to the normal distribution. So, this research concludes that the dependent variables are not normally distributed. For further analysis, this research makes use of non-parametrical tests.

4.4 Hypotheses testing

According to Figure 3, emailing seems to have no effect on tracked activity. However, there may exist differences in tracked activity between groups 1 to 6. To test the hypothesis given a non-normal distribution, this research makes use of a Kruskal Wallis test. This is a that is used to compare three or more groups with continuous, not-paired data. To test two groups, this research uses a Mann-Whitney U test. Both methods compare the mean rank of the different groups. First the complete model is tested. After that, the model is divided into all small parts, to test the different groups separately. This is statistically not the best method, but due to limitations of non-parametric tests this research is forced to do so.

4.4.1 Hypothesis 1, hypothesis 3 and hypothesis 4

Hypothesis 1, 3 and 4 are all about the different motivation techniques. It is therefore decided to group these hypotheses together in the remainder of this section.

The significance of the complete model, so comparing all the seven groups, is 0.582 for the variable ‘b_tracked_energy_expenditure’ and 0.510 for the variable ‘a_amount_tracked_activities’. The mean ranks of the seven different groups are comparable. These are between the 1027 and 1070 for both variables. Meaning that no differences are found between the seven groups.

This research also tested the three motivations and two activations separately. This leads to the following results for the activations. The Kruskal Wallis test has for activation or no activation a significance of 0.582 for the variable ‘b_tracked_energy_expenditure’ and 0.529 for the variable ‘a_amount_tracked_activities’. The mean ranks are comparable for the four groups too. This means that no differences are found between the activations.

(34)

All the motivations are also individually tested and compared to the other groups, including control group. The results score on significance all above the 0.05. The significance for the fitscore is 0.346 and 0.369. The significance for the rewards is 0.525 and 0.569. And the significance for the social challenges is 0.760 and 0.742.

Lastly, all the groups are compared individually with every other group. This leads to the following two tables for both dependent variables.

Amount of Tracked Activities

Group 0 1 2 3 4 5 6 0 x 1 0,818 x 2 0,335 0,272 x 3 0,802 0,968 0,294 x 4 0,682 0,381 0,051 0,365 x 5 0,761 0,452 0,066 0,437 0,887 x 6 0,741 0,883 0,343 0,918 0,309 0,371 x

Table 5: P-values testing difference in amount of tracked activities between groups

Tracked Energy Expenditure

Group 0 1 2 3 4 5 6 0 x 1 0,794 x 2 0,333 0,301 x 3 0,799 0,988 0,308 x 4 0,717 0,38 0,057 0,37 x 5 0,768 0,433 0,069 0,426 0,911 x 6 0,706 0,89 0,373 0,906 0,305 0,352 x

Table 5: P-values testing difference in tracked energy expenditure between groups

4.4.2 Hypothesis 2

(35)

model, only the interaction effect is tested. This leads to a significance of 0.034 and 0.031. Where the mean rank for men is around the 110 and for women around the 100, meaning that differences are found between men and women.

To test hypothesis 5, this research makes use of a regression analysis. The regression between the variable the amount of days between registration and email and the variable amount of activities is tested, besides that the other dependent variable ‘b_tracked_energy_expenditure’ is tested with regression too. This leads to a significance of 0.031 for the variable ‘a_amount_tracked_activities’. The t-value for this variable is -2.160. The other dependent variable, ‘b_tracked_energy_expenditure’, tests a significance of 0.165 and a t-value of -1.387. Meaning, the fewer days between registration and email, the more tracked activities.

Hypothesis 6 states that video emails are more effective than text only emails. This leads to the following results. The Kruskal Wallis test gives, for the analysis with the control group, a significance of 0.311 for the variable ‘a_amount_tracked_activities’ and for the variable ‘b_tracked_energy_expenditure’ 0.318. The mean ranks are comparable. These are for both variables between 1035 and 1054. This is tested with the control group.

(36)

4.5 Hypotheses overview

Hypothesis a) Sig. b) Sig. Rejected/accepted

H1: Social challenges motivate users to track a) more activities

and b) more energy consuming activities in the Health app app. 0,76 0,42 Rejected

H2: The effect size of using social challenges as motivation tool for tracking a) more activities and b) more energy consuming activities in the Health app app is for men bigger than for

women.

0,034 0,031 Accepted

H3: The fitscore motivates users to track a) more activities and b)

more energy consuming activities in the Health app app. 0,346 0,369 Rejected H4: Post-rewarding motivates users to a) more activities and b)

more energy consuming activities in the Health app app. 0,525 0,569 Rejected H5:Activating users soon after downloading the Health app app

leads to a) more tracked activities and b) more tracked energy consuming activities in the Health app app.

0,031 0,165 Accepted / Rejected

H6: Video tutorials in emails lead to a) more tracked activities and b) more tracked energy consuming activities in the Health app app compared to text only emails

0,311 0,318 Rejected

Table 6: Overview hypotheses

4.6 Ranking the groups

Although no significance is discovered, this research ranks the different motivation methods and activation methods. Because no strong evidence exists, this is shortly described below. Ranking is done by taking in account the means of a) and the means of b) of the different groups. Means are chosen because they express the impact better and they are more meaningful values. This leads to the following results. Firstly, observing all the groups leads to the following results.

Group Mean a) Mean b)

2 Text Financial 0,92 12867 3 Text Challenge 0,67 13682 5 Video Financial 0,63 5654 4 Video Health 0,53 10265 1 Text Health 0,5 11106 6 Video Challenges 0,5 8621

(37)

For the activation, text or video only emails, it appears that emails with only text more effective.

Activation Mean a) Mean b)

Average Text 0,70 12552

Average Video 0,55 8180

Table 8: Means different activation methods

For the activation methods, financial motivation, so rewarding (financial) seems to be the most effective. Followed by social challenges and health (fitscore).

Motivation Mean a) Mean b)

Average Financial 0,78 9261

Average Challenges 0,59 11152

Average Health 0,52 10686

Table 9: Means different motivations

4.7 Model free evidence

In this part, possible differences between groups are tested. During discovering the data is found that there could be a difference between the groups if this research sorts on other variables. One of the variables used is life phase. The life phase is only available for users who are Healthcare insurance company customer. This leads to a data set of only 581 users. But, this is enough to analyze the data. This variable is tested with the variable tracked physical activity or no tracked physical activity. Possible differences are found in the life phase group young people and the other groups.

Family Young people Other Senior

Not-tracked 155 109 85 162

Tracked 25 5 13 27

Total 180 114 98 189

Table 10: Different life phases and activity

(38)

5. Conclusion

The research question of this research: ‘’What email interventions can Healthcare insurance

company best use to increase a) the amount of tracked activities and b) the tracked energy expenditure of new users in the Health app app?’’ will be answered in the conclusion. The five

sub questions help to answer the research question. These sub questions are answered by testing hypotheses.

As stated in the previous section, four out of six hypotheses are rejected. Only the second and the a) part of the fifth hypothesis are accepted. The research questions, sub questions and hypotheses are discussed below. Comparing the different groups and testing them on significance, gives as result that there is no significant difference between any of the groups. This is shown in Table 5 and Table 6. However, this research answers the research question and sub question in the next part.

5.1 Sub question 1

The first research question is ’does explaining the social challenge option in emails increase a)

the amount of tracked activities and b) the tracked energy expenditure of new users in the Health app app?’’ Testing the related hypothesis, H1: social challenges motivate users to track a) more activities and b) more energy consuming activities in the Health app app, shows that

social challenges do not increase the amount of tracked activities and do not increase the amount of tracked energy expenditure in the Health app app. Because, both tests score not significant. A hypothesis tests not significant, if the score is higher than the assumed significance level of 0.05. The score of the hypothesis on the a) part, so the amount of activities, is 0.76 and the score on the b) part, so the energy expenditure, is 0.42.

So, part a) of sub question 1, does explaining the social challenge option in emails increase the amount of tracked activities of new users in the Health app app? No. And part b) of the first sub question, does explaining the social challenge option in emails increase the tracked energy expenditure of new users in the Health app app? No.

The second hypothesis is also related to sub question 1. The second hypothesis is, H2: the effect

size of using social challenges as motivation tool for tracking a) more activities and b) more energy consuming activities in the Health app app is bigger for men than for women. So,

(39)

(40)

scores on significance 0.034 on part a) and 0.031 on part b). This leads to significant results. So, this research concludes that men are more motivated by challenges than women.

The second sub question is ‘’does explaining the fitscore in emails increase a) the amount of

tracked activities app and b) the tracked energy expenditure of new users in the Health app app?’’. The third hypothesis, H3: the fitscore motivates users to track a) more activities and b) more energy consuming activities in the Health app app, supports answering this sub question.

The hypothesis test scores not significant. The significance scores are 0.346 on part a), fitscore motivates users to track more activities in the Health app app and 0.369 on part b), the fitscore motivates users to track more energy consuming activities in the Health app app. For this hypothesis, no significant difference is discovered.

So, part a) of sub question 2, does explaining the fitscore in emails increase the amount of tracked activities of new users in the Health app app? No. And part b) of sub question 2, does explaining the fitscore in emails increase the tracked energy expenditure in of new users the Health app app? No.

The third sub question is ‘’does explaining the rewarding system in emails increase a) the

amount of tracked activities and b) the tracked energy expenditure of new users in the Health app app of new users?’’ This is related to the rewarding system in the Health app app. The

fourth hypothesis, H4: post-rewarding motivates users to a) more activities and b) more energy

consuming activities in the Health app app, helps answering the sub question. This hypothesis

(41)

The fourth sub question is, ‘’Does a video tutorial in emails increase a) the amount of tracked

activities and b) the tracked energy expenditure of new users in the Health app app compared to text only emails?’’ The sixth hypothesis, H6: video tutorials in emails lead to a) more tracked activities and b) more tracked energy consuming activities in the Health app app compared to text only emails, supports this sub question. This hypothesis tests with a score on part a) video

tutorials in emails leads to more tracked activities, 0.311 and on part b) video tutorials in emails leads to more tracked energy consuming activities, 0.318 not significant. Thereby, looking at the mean, it seems that only text is more effective than a video tutorial. This is not in line with the hypothesis. Logically, answering both parts of the sub questions leads to two rejections. A video tutorial in emails does not increase a) the amount of tracked activities and b) the tracked energy expenditure of new users in the Health app app.

The fifth sub question ‘’does sending an email to new users sooner after registration in the

Health app app leads to more a) tracked activities and b) tracked energy expenditure of new users in the Health app app?’’, is tested by the fifth hypothesis. The fifth hypothesis is, H5: activating users soon after downloading the Health app app leads to a) more tracked activities and b) more tracked energy consuming activities in the Health app app. This hypothesis is

tested by the regression between the days between registration and sending the email and the both dependent variables. A significant difference of 0.031 is found for part a) of this hypothesis. For part b) no significant difference is found. This score was 0.165.

So, for part a) of the sub question, does sending an email to new users sooner after registration in the Health app app leads to more a) tracked activities? Yes. But, for part b) of the sub question, does sending an email to new users sooner after registration in the Health app app leads to more b) tracked energy expenditure? No.

(42)

5.6 Research question

The research question is ‘’what email interventions can Healthcare insurance company best

use to increase a) the amount of tracked activities and b) the tracked energy expenditure of new users in the Health app app?’’, all the sub questions support answering the research question.

However, no significant differences between the groups, motivation methods and activation methods are found. Nevertheless, there is a significant difference between the effect between men and women for challenges. So, Healthcare insurance company should motivate men with challenges.

As said, no significant difference is found between the groups, activations and motivations. But simply answering the research question, taking the significance not into account and simply comparing means of different groups, activations and motivations, there is a slight indication that Healthcare insurance company should activate with text only emails and motivate with financial aspects.

(43)

6. Discussion

6.1 Limitations

Firstly, this research uses Kruskal Wallis tests testing single relationships, because the dependent variables are not normally distributed. The downside of this method is that this research could not test the full model with one single test. The best way to test is to test the complete model. But, this research did this because of the low significance values found by testing the complete model and because the non-normal distributed data set. It was not possible to add more independent variables in a Kruskal Wallis test. An advantage of this method is that it is possible to discover differences between groups for data that are not normally distributed.

Further, for the fifth hypothesis a linear regression analysis is done between the variable the amount of days between registration and email and the variable amount of activities. The regression coefficient was negative, meaning that people who received the email shorter after registration tracked more activities than people who receive the email later after registration. It is not certain that the increase in effect is due to the email. Other factors like recency of registration or willingness to use the app can be factors influencing the amount of tracked activities too.

Thirdly, the email is sent in two batches on different days of the week. The first one is send on Wednesday and the second one is send on Thursday. Because of the weather and other external factors, this can lead to differences in the amount of tracked activity and amount of tracked energy expenditure. Also, according to research, Wednesday is a better day to send emails. This can lead to higher open rates for the first batch. Because the first batch is started using the Health app app 4 to 1 week before sending and the second batch is started using the Health app app 1 week before sending, this can lead to differences.

(44)

differences occur between groups if the weather was better, because the one activation or motivation works better with good weather. This could lead to more significant results.

Another limitation of this study is that the challenge ‘Kom in Beweging’ in the Health app app was open for everybody. Within a week, a thousand Health app users were participating in the challenge. Participants in this research are not competing with comparable people, but are competing against the total group of users. It contains several fanatical users, which makes it relatively hard to reach the top of the ranking, which could lower motivations.

Thereby, this research cannot compare the receivers of the email and the joiners of the social challenge because of privacy rules. So, it is impossible to measure if the people who have got an activation by email with the social challenge, are really joining the social challenge ‘Kom in Beweging’ in the Health app app.

Furthermore, a push notification is sent to the Health app app users. This push notification contains the text: ‘’De uitdaging Kom in Beweging staat voor je klaar.’’ It is sent at 4:22 p.m. November 21th. This can lead to more people who use the challenges. Because they first get the push notification and a few minutes later the email. The email is sent at 4:30 p.m. It was not possible to switch off the notifications on the mobile phones. This can lead to more effect on emails with an activation on social challenges, because the members of this group are activated twice.

Also, the video emails do not contain a real video. The video mails contain a GIF without audio. Maybe videos with audio are more effective as tutorial. People may prefer the voice explanation in videos, which are not available in the GIFs. This can lead to less effect for emails with a GIF.

Lastly, the data for the model free evidence part are only available of Healthcare insurance company customers. But, Healthcare insurance company customers can be more activated than only by the email. This can lead to other results. Because current customers can be activated longer by other Health app promotions and are maybe more familiar with the concept.

6.2 Further research

(45)

(46)

groups want to be approached in different ways. The content of the emails can be interesting too. Further research in the exact reasons why users of the Health app app stop using the app after a month, is also interesting. Which factors contribute to this? An option to research this, is by forming focus groups.

Another direction for further research is what motivates users to use the app. This research uses the other way around, it tests motivations that are already available in the app. Perhaps, Healthcare insurance company has to implement other motivation techniques, fitting to the (latent) needs of the customers. A requirement for this is that Healthcare insurance company investigates what the latent needs are. This can be done in for example focus groups and in-depth interviews.

Thereby, because it seems that the financial motivation works best compared to the other two techniques, it is interesting for Healthcare insurance company to research if motivation with different customized rewards works. For example, in December the Xbox One was available for Healthcare insurance company customers in the web shop. Maybe, with promoting this, the amount of young adult users increases. While promoting for example theater tickets will increase the usage of generation 50+. In the theoretical background is this discussed as an option that could work.

(47)

7. Recommendations

7.1 Next steps

This research investigates the best way to activate and motivate users to use the Health app app. No significant difference is discovered between groups, motivation methods or activation methods. But this research hints in the direction of motivating participants with a text only email. So, the recommendation for Healthcare insurance company is to use only text in emails to explain a function in the app.

Regarding the motivation methods, in general, results suggest that motivating people by using the financial aspect might be slightly more effective than the other motivation types. This is promoting the rewarding system in the Health app app. A side effect described in the literature is tested in this research too. This is that the effect size for the motivation social challenges is bigger for men than for women. This hypothesis tested significant. So, when promoting social challenges, Healthcare insurance company should focus on male participants.

Because the few to no differences in the amount of activities and the total energy expenditure between the control group and the 6 other groups together, this research strongly recommends reconsidering the use of one specific email to motivate and activate users to use the Health app app. This research recommends Healthcare insurance company to look for potential other options for activating and motivating users to track activity in the Health app app. Possible options to motivate and activate are discussed in the next section.

7.2 Potential new options to motivate and activate

In this part, some other options for activating and motivating app usage, found on websites and in articles, are discussed. On different websites and in articles in-app push notifications are mentioned as promising tools to motivate and activate people (Zadorozny et al., 2016; Cassidy et al., 2018). For Healthcare insurance company it is interesting to investigate this option and research the best way to make use of this.

(48)

A promising option that comes with the social challenges is make use of social sharing in the app. It should be easy to share your activities and your rewards on social media platforms like Facebook, Twitter and Instagram. This can motivate people as they want to be seen by others in a positive way (Fadıllıoğlu, 2018).

On different websites, emailing is mentioned as an option that should not be forgotten (Fadıllıoğlu, 2018). Especially emails with information about updates, news or special discounts or price promotions are interesting. The effectiveness of discounts and price promotions confirms this research partly. Thereby, it is important to target the right customers with email. According to the current research, emails are less effective for young people and should therefore not be send to young people.

(49)

References

Abraham, C., & Michie, S. (2008). A taxonomy of behavior change techniques used in interventions. Health psychology, 27(3), 379.

Ajzen, I. (2011). The theory of planned behaviour: reactions and reflections.

Allen, J. K., Stephens, J., Dennison Himmelfarb, C. R., Stewart, K. J., & Hauck, S. (2013). Randomized controlled pilot study testing use of smartphone technology for obesity treatment.

Journal of obesity, 2013.

Allen, T. D. (2006). Rewarding Good Citizens: The Relationship Between Citizenship Behavior, Gender, and Organizational Rewards 1. Journal of Applied Social Psychology, 36(1), 120-143.

Andrade, A. C. de S., Mingoti, S. A., Fernandes, A. P., de Andrade, R. G., Friche, A. A. de L., Xavier, C. C., Caiaffa, W. T. (2018). Neighborhood-based physical activity differences: Evaluation of the effect of health promotion program. PLoS ONE, 13(2), e0192115.

Armour, B. (2018). 5 Methods For Increasing App Engagement & User Retention. Retrieved December 20, 2018, from https://clearbridgemobile.com/5-methods-for-increasing-app- engagement-user-retention/

Bartholomew, K. J., Ntoumanis, N., Ryan, R. M., & Thøgersen-Ntoumani, C. (2011). Psychological need thwarting in the sport context: Assessing the darker side of athletic experience. Journal of Sport and Exercise Psychology, 33(1), 75-102.

Biddle, S. J., & Asare, M. (2011). Physical activity and mental health in children and adolescents: a review of reviews. British journal of sports medicine, bjsports90185.

Boiarsky, C. (2015). The impact of emailing and texting on effective written communication: Changes in reading patterns, convergence of subgenres, confusion between social and business communication. In Professional Communication Conference (IPCC), 2015 IEEE International (pp. 1-6). IEEE.

(50)

Boreham, C., & Riddoch, C. (2001). The physical activity, fitness and health of children.

Journal of sports sciences, 19(12), 915-929.

Caspersen, C. J., Powell, K. E., & Christenson, G. M. (1985). Physical activity, exercise, and physical fitness: definitions and distinctions for health-related research. Public health reports,

100(2), 126.

Cassidy, B. G., Warrick, P. S., & Carriere, L. M. (2018). U.S. Patent Application No.

10/026,101.

Colditz, G. A. (1999). Economic costs of obesity and inactivity. Medicine and science in sports

and exercise, 31(11 Suppl), S663-7.

Deci, E. L. (1971). Effects of externally mediated rewards on intrinsic motivation. Journal of

Personality and Social Psychology, 18, 105–115.

Dorotic, M., Verhoef, P. C., Fok, D., & Bijmolt, T. H. (2014). Reward redemption effects in a loyalty program when customers choose how much and when to redeem. International Journal

of Research in Marketing, 31(4), 339-355.

Douglas, K. A., Collins, J. L., Warren, C., Kann, L., Gold, R., Clayton, S., ... & Kolbe, L. J. (1997). Results from the 1995 national college health risk behavior survey. Journal of American

College Health, 46(2), 55-67.

Dute, D. J., Bemelmans, W. J. E., & Breda, J. (2016). Using Mobile Apps to Promote a Healthy Lifestyle Among Adolescents and Students: A Review of the Theoretical Basis and Lessons Learned. JMIR mHealth and uHealth, 4(2), e39.

Eagly, A. H., Karau, S. J., Miner, J. B., & Johnson, B. T. (1994). Gender and motivation to manage in hierarchic organizations: A meta-analysis. The Leadership Quarterly, 5(2), 135-159.

Ellis-Chadwick, F., & Doherty, N. F. (2012). Web advertising: The role of e-mail marketing.

Journal of Business Research, 65(6), 843-848.

How to motivate and activate users of health apps by email Investigating the effectiveness of six interventions on promoting health program usage

How to motivate and activate users of health apps by email

Investigating the effectiveness of six interventions on promoting health

program usage

How to motivate and activate users of health apps by email

Investigating the effectiveness of six interventions on promoting health

program usage

Management summary

Preface

Table of contents

1. Introduction

2. Theoretical background

3. Methodology

4. Results

5. Conclusion

6. Discussion

7. Recommendations

References