App ratings in different stages of an app ventures life cycle

(1)

App ratings in different stages of an app

ventures life cycle

August 2nd, 2018 Hendrik Scheffer UvA 10300678 VU 2084503 MSc Entrepreneurship Supervisor:

(2)

Preface

The copyright rests with the author. The author is solely responsible for the content of the thesis, including mistakes. The university cannot be held liable for the content of the author’s thesis.

(3)

Abstract

The main purpose of this paper is to find a life cycle for app ventures and to see whether there are differences in the ratings of apps in the different stages of their life cycle. The trajectory of the life cycle and the position of an app in its life cycle is relevant to its customers, its

developers and the app store provider. This contributes to the entrepreneurial decision

whether or not to continue development of this app. The ratings in each stage are relevant for the findability of the app in the app store, as a higher rating is favorable for all parties. The main findings are that for 47% of the apps the life cycle consist of a startup stage, a rapid growth stage, followed by a decline stage and that for 40% of the apps the life cycle consists of a startup stage, directly followed by a decline stage. Both of these life cycles did not include a maturity stage. The part of the life cycle of an app venture with positive growth is on average only 2.2 years long. Also, the average ratings in each stage were found to be significantly different: startup stage 4.22 stars, rapid growth stage 4.08 stars and decline stage 4.03 stars. In both life cycles each further stage in the life cycle got a lower rating. This implicates that app developers need to move quickly and therefore customers and app store providers should anticipate this short life cycle.

(4)

1 Introduction

An app venture can be compared to a firm, as it has customers, has value and is an

entrepreneurial venture. However, does it also have a comparable life cycle, or does it differ on that aspect? What is the general trajectory of this life cycle, would it look like the life cycle of the firm model? Are there differences in the rating of an app in the different stages of its life cycle?

1.1 Research questions

The main question researched is:

To what extent are apps rated differently in the stages of their life cycle?

This question can be divided into two sub questions:

• What is an app?

• What is the life cycle model?

1.2 Contribution to literature

Apps are a new subject in academic research as they only existed in their current form for just over 10 years. Therefore, a lot of aspects of apps have not yet been researched. One of these aspects is an entrepreneurship perspective on the stages of the life cycle of an app venture, which is the main subject of this thesis. In previous research properties of organizations related to certain stages of the organizational life cycle. In this research a property of apps, the customer rating, in connection with the stage of the life cycle will be researched .

1.3 Relevance

There are several stakeholders for apps. The obvious ones are the customers and the

developers/publishers. Next to that, apps contribute to the app ecosystem of the platform and thus to the app store provider and device manufacturers.

(8)

1.3.1 Customers

For customers the importance of the life cycle of apps is relevant for the apps they use and consider using. This consists of three aspects:

• Does the app get updates with new features?

• Does the app still get support, so it keeps on working?

• In case of an app with collaboration or competition elements: does the app have

enough users?

1.3.2 Developers/publishers

For the developers/publishers their apps generate direct or indirect value in terms of direct income or indirect income, branding and customer service. However, maintaining an app costs time and/or money. Thus, an entrepreneurial decision must be made each time between:

• continuing active support, including development of new features,

• continuing maintenance support, to keep the app working, but without new features, • stopping support at all.

Knowing an estimate of the total life cycle of an app and its current position in this life cycle would make this decision easier and more profound.

1.3.3 App store provider and device manufacturers

App store providers generate money by selling apps through the app stores. Device

manufacturers earn money by selling smartphones, tablets, smartwatches etc. Both of these stakeholders need to have enough supported quality apps available for the platform. Without these apps the platform will collapse and this revenue stream will stop for these companies. Lee, Kim and Hong (2017) found that, according to developers, platform providers need to, amongst others, provide assurance of adequate financial returns to developers. Therefore, it

(9)

seems essential for these companies to optimize the life cycle of the apps on their platform and thereby triggering the developers to actively support their apps.

1.4 Outline

In chapter Theoretical Framework the theoretical background of the life cycle of the firm model and apps will be described. Also, the stages of the life cycle, that will be researched amongst the life cycle of app ventures, will be defined. In chapter Hypotheses the hypotheses to the main research question will be described. The research design and the methodology of the data collection and analysis will be explained in chapter Methods & Fieldwork. In chapter Results & Testing of Hypotheses the results of the research and the hypotheses testing are described and explained. In chapter Discussion the results are discussed. The last chapter, Conclusion, summarizes the conclusions of this research, gives answer to the research question and discusses the limitations and recommendations.

(10)

2 Theoretical Framework

This paper focusses on two main constructs: apps and the life cycle of the firm model.

2.1 Apps

App is short for application. However, there are some small differences between apps and traditional applications. Traditional applications are (computer) programs, they generally are not checked for malicious behavior by a third party before publishing and they are marketed to the users through a number of channels. While apps are generally published through an app store, which provides checks against malicious behavior and most of the time is the only method of distribution of this app.

Each app is made for a specific platform. Although there are tools to provide cross platform support (Rieger & Kuchen, 2018), they only can build apps for multiple platforms from a single code base. This still results in different apps for each platform. The most known mobile platforms currently are iOS, Android, Windows and BlackBerry. Most apps are built to work on either the iOS or Android platform (Godwin-Jones, 2011; Shuler, Levine, & Ree, 2012).

2.1.1 App ventures

Every app that is published becomes part of the relevant app ecosystem, whether it is made for profit or not, just like any company becomes part of the economy when it is founded. The people who download the apps are the customers of these apps. The other apps in the same app store that provide similar functionality can be seen as the competition of an app.

There are multiple ways for app developers and publishers to create value for themselves with apps. They can generate direct income by selling the app for a fixed fee in the app store or by adding in-app-purchases to enable users to purchase additional features or by adding in-app-advertising to earn money by showing advertisements to the users (Hao,

(11)

Guo & Easley, 2017). Another way is generating indirect income from apps (Gill, Sridhar & Grewal, 2017), such as a companion app for a smart device, for example smart thermostat Toon (Eneco, 2018), where the income comes from selling the smart device. An app can also be part of a marketing campaign, such as the D66 Nu app (D66, 2017), or provide other forms of branding value, such as the Coca-Cola Human Rights app (The Coca-Cola Company, 2015), or be part of the customer service of a company (Ehrenhard, Wijnhoven, van den Broek & Stagno, 2017), such as the PostNL app (PostNL Holding B.V., 2018).

2.1.2 App stores

Some platforms allow developers and app publishers to distribute their apps to the customers directly, without an app store, for example on Android. However, this also means that the app developers and app publishers need to find their own way to reach the customers.

Nowadays, all platforms are equipped with an app store to give the users a single place to find apps. Apple and Google opened their own app stores, the Apple App Store and the Google Play Store (Tong, She & Chen, 2015). Three other known app stores are the Windows Store, BlackBerry World and the Amazon Appstore (Wan, Zhao, Lu & Gupta, 2016). Next to that, there are a lot of smaller ones, which partly target the same platforms (Müller, Kijl, & Martens, 2011).

App stores earn a share of the revenue of the app sales and the in-app-purchases for the app distribution services they provide to app developers and app publishers. Although app stores do also benefit from having more quality apps in the app stores in other ways. For example, via increased device sales in case the app store provider is also a mobile phone manufacturer (Huang, 2016).

(12)

To ensure app quality in the app stores, the app store providers review each app on malicious behavior and on their content guidelines. Part of this process is automated and part of this is done manually. This differs per app store.

The app stores contain over 4 million apps (Paget & Frosch, 2016). Therefore, the most difficult part of the app creation is effectively maintaining the appearance of the app in the app store (Han Rebekah Wong, 2012).

2.1.3 App Reviews

All popular app stores provide their users with functionality to leave a review for every app they downloaded, so they can give feedback to the developers and inform potential app downloaders about the quality of this app (Maalej, Kurtanović, Nabil & Stanik, 2016; Zhang, Huang, Jiang, & Hu, 2017). Reviews generally contain a text part, the review contents, a grade for the app, the rating and the date of writing. Most app stores enable users to express this rating on a scale from 0 to 5, mostly represented by 0 to 5 stars. These app stores show these reviews to all visitors of the apps information page and also show the average rating and total number of reviews. According to Carter and Yeo (2016) people use reviews as guide when considering downloading an app. Jung, Baek and Lee (2012) found that customer ratings critically affect product survival when the price is zero. Although a study by Datta, Kajanan and Pervin (2013) suggests users do not always know how to interpret the reviews and ratings. For app developers and publishers, it is advised to get the highest rating possible. And even if an app starts with a low rating, increasing this will result in high increased downloads (Apptentive, 2015).

2.2 Life cycle of the firm model

The life cycle of the firm model has been described by many authors and exists in many variations.

(13)

2.2.1 Basis of the model

Downs (1967) researched the life cycle of government bureaus. He suggested that they had three stages in their life cycle: struggle for autonomy, where the legitimacy is established, rapid growth, where organization expands, and deceleration, where the organization is formalized.

Greiner (1972) identified five phases in the evolution of organizations and identified differences for each phase in, amongst others, management style, organization structure and control system. Greiner also states that between every two stages there was some form of crisis, which prevented the company from moving to the next stage, and revolution, where a crisis is overcome by an organizational change. He did link the phases to the overall size of the company, but not to the amount of growth in relation to the time.

Adizes (1979) found that organizations move through the stages as a result of changes in a combination of producing results, acting entrepreneurially, administering formal rules and procedures and integrating individuals into the organization. These factors also explain a possible decline of an organization. Therefore Adizes also added five more phases to the model of Greiner, which describe the downfall of an organization to the death of this organization. He also related the stages to the age of the organization.

2.2.2 Relation between growth and time

Churchill and Lewis (1983) researched amongst small businesses. They based their model on the model of Greiner (1972) but replaced the measure size with the combination of size, complexity and diversity. They also added the relation between growth and time to the different stages of an organization. They start with an Existence stage, where the company just had been founded and is growing very slowly. In the next stage, the Survival stage, the company size is taking off and the growth is increasing. This is followed by the Success stage, where the company size keeps growing, and the growth rate is at its maximum. In the

(14)

Take-Off stage the company is still growing but the growth rate is decreasing. And in the last stage, the Resource maturity stage, the company reaches its maximum size and the growth

disappears. They described each stage by five management factors: managerial style, organizational structure, extent of formal systems, major strategic goals and the owners involvement in the business. They did not research the demise of the organization after they reached the resource maturity stage.

Scott and Bruce (1987) researched the life cycle of small businesses and found the stages inception, survival, growth, expansion and maturity. They based this model on the model of Greiner (1972), however they included a relation between growth and time. They proposed a model for small business owners to plan for future growth, as, based on the model of Greiner (1972), moving to a next stage in the life cycle is prevented by a crisis and has to be overcome by some sort of revolution.

Mitra and Pingali (1999) describe a similar life cycle with existence, survival, success/growth and resource mature as stages based on the model by Churchill and Lewis (1983). They also suggest small and medium-size enterprises have a different life cycle than large firms. They based did on their own research amongst automobile ancillary firms in India. For these firms they indeed found a different life cycle, which is high growth, existence, survival, growth / disengage and resource mature.

Beverland and Lockshin (2001) researched the life cycle for small New Zealand wineries. Next to a startup, expansion and growth stage, they included a pre-birth stage, which they based on the research by O’Farrell and Hitchens (1988). This pre-birth stage happens before the company is actually launched. For each of the stages they found the average length, key focus, way of production, way of marketing, distribution method, staff functions and their main challenge.

(15)

Masurel and Van Montfort (2006) researched the life cycle model amongst small professional service firms, in particular architects. They distinguished four different stages, startup, growth, maturity and decline. They found that diversification in sales, the

differentiation in labor force, and the level of labor productivity increase in the startup, growth and maturity stage, and that they decrease in the decline stage.

Yan and Zhao (2010) found that the measurement of firm life cycle stages can also be done by comparing a firms status with its own historical overall status. Yan and Zhao decided to require stages to have a minimum length of two years. They defined the stages birth, growth, maturity, revival and decline. However, they decided to choose the IPO date as the start of the growth stage and did not look any further into the birth stage. Secondly, they defined decline stage as not after revival, but after maturity. This also means that once a firm moves out of the maturity stage it either moves into revival or into decline. Thus there are two life cycles a firm could follow.

Figure 1 General life cycle of the firm trajectory

2.2.3 Critique on the differences in the models

O’Farrell and Hitchens (1988) argued that most of the research is about the internal dynamics of the firms, while they underestimate external factors, which also are important. They also note that small firms are different than large firms, and therefore might experience a different life cycle than large firms.

(16)

Phelps, Adams and Bessant (2007) found that most of the life cycle research finds different number of stages and most of them suffer from being linear, unidirectional, sequenced and deterministic. They note that each businesses is different and therefore there are different life cycles possible. They found six ‘tipping points’, major transitions in the organizations as a solution to overcome a crisis, just like the crises of the model of Greiner (1972). They state that firms can cross each of the stages in their own order and that most growing companies will encounter all these crises at some point, this also shows a lot of resemblance with the model of Greiner (1972).

Jaafar and Halim (2016) also found that life cycle research uses several different numbers of stages and also that a the used measures vary a lot. They propose a life cycle classification method and emphasized the use of multiple financial proxies. Also they deprecate the use of the age of a firm and the dividend payout ratio.

2.2.4 Integration of the models

Although all models are created for different situations, they all follow a similar pattern (Quinn and Cameron, 1983). Which consists of an initial stage of slow growth, followed by a stage with rapid growth, which on its turn is followed by a stage of slow growth. Some of the models also include one or more stages of decline afterwards. There also are a few who include a stage before the founding of the firm (O’Farrell & Hitchens, 1988; Beverland & Lockshin, 2001). This general trajectory is shown in Figure 1. The models differ from 3 stages to 10 stages (Hanks, 1990).

The measure used to analyze which stage a firm is in also differences between the different studies, from a simple measure, such as revenue, to more complex measures, such as a combination of size, complexity and diversity (Churchill & Lewis, 1983).

(17)

Another difference between the studies is that some studies think of progress through the stages as a linear process, where the order of the stage is predefined. While other studies think of this process as companies being able to move through all the stages in their own particular order, where stages may be skipped or stages accessed multiple times (Hanks, 1990).

Most of the studies also connect properties of the organizations to the different stages. However, a study by Dodge, Fullerton and Robbins (1994) states that those properties are merely related to the state of the competition instead of the stage in the life cycle. Levie and Lichtestein (2010) found that coupling these properties to different stages could act as a barrier to advancement of research instead of help to advance the research into the growth of entrepreneurial organizations.

(18)

3 Hypotheses

The goal of this research is to get more insight into the life cycle and rating of an app and whether it fits the life cycle of the firm model. Sub question 1, What is an app?, and 2, What is the life cycle model?, are answered in the previous chapter about the theoretical framework.

For this research the following stages are used: startup stage, starting at the release of an app lasting for all the consecutive years with slow growth. Rapid growth stage, the years with rapid growth after the startup stage. Maturity stage, slow growth after the rapid growth stage. Decline stage, with negative growth. Most literature uses growth stage as name for the rapid growth stage (Scott & Bruce, 1987; Masurel & Van Montfort, 2006), however, in both the startup stage and the maturity stage there also is growth. So to make this distinction more clear the name rapid growth stage is used, this name is also chosen by Downs (1967).

3.1 Life cycle of an app venture

Apps have a lot of similarities with businesses. As described in the previous chapter, they have customers, they have a business model and they can have revenue. Also, every now and then another app suddenly becomes very popular and after a while its popularity decreases. Examples are Wordfeud and Pokémon Go. These hyped apps seem to follow a pattern. They start very small, as they get published as new apps. Next, they enjoy some sort of growth until they get popular and grow very fast. After this the growth has to reach a limit since there is a limit to the number of smartphone users. Then they fall into decline and another app becomes very popular. This pattern shows a lot of similarities with the life cycle of the firm model.

This leads to the first hypothesis:

(19)

3.2 Main research question

The main research question is:

To what extent are apps rated differently in the stages of their life cycle?

The stages of the life cycle of the firm model consists of the startup stage, rapid growth stage, maturity stage and decline stage. It is expected that the change in growth will also provide a hint about the rating of the app in the different stages.

In the first stage, the startup stage, an app just has been released to the app store. It might have some issues or is lacking some features, what could influence the rating negatively. Thus, the rating in this stage is presumably not at its peak. Next, the app enters the rapid growth stage. One possibility is that the developers have implemented the feedback of the initial users, which might cause the app to enjoy higher growth and attain higher ratings. Which naturally results in the next sub-hypothesis.

Sub-hypothesis 2a: rapid growth stage has a higher rating than startup stage.

When the app reaches the maturity stage, it could be that the app does not bring enough value to the users to keep growing at the previous rate and the rating might stop growing. This results in de second sub-hypothesis.

Sub-hypothesis 2b: maturity stage has a higher rating than rapid growth stage.

When the app reaches the decline stage, this means that users are leaving, either because the app does not give any value to them anymore or they might be dissatisfied because of a change to the app. In both cases the rating would be influenced negatively, which leads to the last sub-hypothesis.

(20)

Overall, these sub-hypotheses suggest that the rating is positively related to the size of the app venture. This leads to the following hypothesis:

Hypothesis 2: An app venture has different ratings in each of the stages of its life cycle and there is a positive relation between the rating and the size of the app venture.

(21)

4 Methods & Fieldwork

In this chapter the research design, research context, data collection, the measures and the data analysis are discussed.

4.1 Research design

The life cycle model allows for both a qualitative and a quantitative approach. However, in this study a quantitative approach has been chosen, as some of the data from the app stores is publicly accessible. This quantitative approach uses the available data to find an answer to the research question. As some of the data of the BlackBerry app store, BlackBerry World, is publicly available, this data can programmatically be gathered for all apps in BlackBerry World. BlackBerry World contained 228,847 apps at the time of data collection. To be able to measure stages in the life cycle of apps based on the number of reviews, only the apps with at least a 1,000 reviews were included in the analysis of this study. This resulted in 1,786 eligible apps, which were included in this study in order to garner a representative sample of the total population of apps with measurable life cycles as expressed in number of reviews over time.

4.2 Research context

This research uses the actual data from BlackBerry World. Users place these reviews to give feedback to the developers of the apps and to inform potential clients of the apps about the quality of the apps.

4.3 Data collection

The goal is to collect all reviews for each app from the app store. BlackBerry has a way to access the data of the app store via a publicly available application programming interface (API). An API is an abstract way to programmatically get data and interact with an external

(22)

software solution. The API of the app store is amongst others capable of listing all apps per developer and getting all reviews per app.

To be able to analyze the data it first needs to be collected. To do so a program is written, which first gets the list of apps and their details per developer. The program then saves per app the name, the category, the publisher and the price into a database. Next for each app in the database the reviews are requested and also saved in a separate database table.

4.4 Measures

The life cycle of the firm model uses two variables: time and size of the firm. Time will be measured in years. Size of an app venture will be operationalized as number of reviews of an app per year, as Nayebi, Cho and Ruhe (2018) state that app reviews are used to analyze the evolution of apps. An example of such a life cycle can be seen in Figure 2, where a graph of the number of reviews and the growth in the number of reviews per year are shown for the app Flashlight by The Jared Company.

For this research the following stages are used: startup stage, starting at the release of an app lasting for all the consecutive years with slow growth. Rapid growth stage, the years with high growth after the startup stage. Maturity stage, slow growth after the rapid growth stage. Decline stage, with negative growth. According to the life cycle model is the year with the highest absolute growth part of the rapid growth stage. Half of this highest absolute growth is therefore a logical boundary of the rapid growth stage and thereby the cutoff point between high and low growth. Therefore, slow growth is defined as a growth of less than 50% of the growth in the year with the highest growth. High growth is defined as growth of at least 50% of the growth in the year with the highest growth. To clarify, this means that a year with an absolute growth of precisely 50% of the highest absolute growth is part of high growth and thus part of the rapid growth stage. And an absolute growth of precisely 0% of the highest

(23)

absolute growth, which is equal to an absolute growth of 0 reviews, is part of low growth. In a study by Yan and Zhao (2010) they chose two cutoff points, which equally distributed the key variable, at 33% and 67%. Therefore, in this study the cutoff point will also equally distribute the key variable and as one cutoff point will be used, this cutoff point is at 50%. Moving this cutoff point to for example 40% or 60% would result in possibly different lengths of stages, but for most of the apps the same stages will be measured. For example in Figure 2, where the number of reviews and the growth in number of reviews for the app Flashlight by The Jared Company is shown, the app has very high growth in the years 2011 and 2012 and

considerable less in 2013. This example shows that a cutoff point of 50%, which makes 2013 part of slow growth and 2011 and 2012 part of high growth, would really fit.

Besides the life cycle of the firm model, the rating in each stage of the life cycle of an app is of interest, as this is a performance indicator. When giving a review, users are able to rate the apps on a scale from zero to five stars, where they also are able to give half stars. A higher rating implies that the app is performing better. This is operationalized as number of half stars, which is the same as multiplying the rating with two. This results in an integer scale from zero to ten.

(24)

Figure 2 Graph of the number of reviews (the blue bars) and the growth of the number of reviews (the red line) compared to the previous year for the app Flashlight by The Jared Company.

4.5 Data analysis

To be able to use all the individual reviews with their date and rating in the life cycle of the firm model, the gathered data has to be converted to the measure of the model, which is the number of reviews per year. To do so, per app all the reviews in each year are counted to come up with the number of reviews for each app for each year. Also, the average rating of each app in each year is calculated. This analysis is done by a self-written computer program, included in Appendix 1.

Although the life cycle of the firm model uses size of the firm, it actually is looking into the growth (and negative growth/decline) of the size of the firm. Thus, per app the growth of the number of reviews compared to the previous year is calculated for each year.

(25)

• The startup stage is per definition at the start as an app starts with 0 users when it gets

released to the app store. Detecting the end of the startup stage is more difficult. The startup stage ends when the rapid growth stage starts. So, the last year of the startup stage is the last consecutive year with growth lower than 50% of the highest positive growth. In the example of the app Flashlight by The Jared Company, see Figure 2, the app is in the startup stage is 2010, because in 2011 it enters the rapid growth stage. • It is easy to find a year that is part of the rapid growth stage: the year with the highest

positive growth is per definition part of the rapid growth stage. However, finding the start and end of this stage is harder. All years with a growth of at least 50% of the highest positive growth are part of the rapid growth stage. In the example of the app Flashlight by The Jared Company, see Figure 2, the app is in the rapid growth stage in the years 2011 and 2012, because 2012 has the highest growth and 2011 has a growth larger than 50% of the highest growth.

• The maturity stage is the stage with low positive growth after the rapid growth stage.

So, the start of the maturity stage is the year after the last year of the rapid growth stage. The maturity stage ends when the growth becomes negative. In the example of the app Flashlight by The Jared Company, see Figure 2, the app enters the maturity stage in 2013 as the growth is below half the top growth and above 0 and the app has already had a rapid growth stage.

• The decline stage is easily detectable because that stage can be detected by looking for

negative growth. Every year with negative growth is decline stage. In the example of the app Flashlight by The Jared Company, see Figure 2, the app is in decline stage in the years 2014 and beyond.

• Years that are not part of any of these stages will be labeled as other stage. In practice

(26)

app Cnectd Messenger by MCI Consultants, see Figure 3, the app is in an unknown stage in the year 2012, which is a year with very small growth between two stages of rapid growth and thus labeled as other stage.

Figure 3 Graph of the number of reviews and the growth of the number of reviews compared to the previous year for the app Cnectd Messenger by MCI Consultants.

For each stage of each app the average rating of all reviews in this stage is calculated. This is done by calculating the weighted average of the ratings for the years in each stage, where the weighting is the number of reviews in the respective year.

To discover a general life cycle of an app venture the found stages will be analyzed. All different occurring orders of stages are measured. For each of the different orders of stages the frequency of apps with this order is counted. For each of these orders of stages the average length of each stage is calculated. This results in all occurring orders of stages and their average lengths in the life cycle of an app venture and how often this order occurs.

The combination of the rating in each stage and the form of the life cycle of an app venture results in an answer to the main research question.

(27)

5 Results & Testing of Hypotheses

For this research there have been 18,652,626 reviews that have been measured. The gathered review data is summarized in Table 1.

Year Number of distinct apps with reviews

Number of Reviews 2009 1,337 37,204 2010 4,716 211,274 2011 14,284 928,794 2012 27,419 4,948,667 2013 36,644 5,079,422 2014 31,718 3,747,949 2015 24,864 2,297,278 2016 18,651 1,082,890 2017 11,984 285,630 2018 4,208 33,518 Total 74,472 18,652,626

Table 1 Summary of collected data, with number of reviews per year and the number of apps with reviews in that year

5.1 Life cycle of an app venture & Testing of Hypothesis 1

As a result of the analysis there are 23 different life cycles (orders of stages) found. The most frequent life cycles and their frequency are listed in Table 2. The full list of all life cycles can be found in Appendix 2.

As discussed in the theoretical framework the stages are: startup stage, starting at the release of an app lasting for all the consecutive years with slow growth. Rapid growth stage, the years with high growth after the startup stage. Maturity stage, slow growth after the rapid growth stage. Decline stage, with negative growth. Slow growth is defined as a growth of less than 50% of the growth in the year with the highest growth. High growth is defined as growth of at least 50% of the growth in the year with the highest growth.

(28)

Life cycle / order of stages (average length of stage in years) Frequency Startup (1.76) Rapid growth (1.45) Decline (3.66) 845

Startup (1) Decline (4.5) 711

Startup (1.33) Decline (1.47) Rapid growth (1.07) Decline (2.8) 75 Startup (1.14) Rapid growth (1.21) Maturity (1.36) Decline (3.36) 68 Startup (1.2) Rapid growth (1.2) Decline (1.67) Maturity (1.13)

Decline (2.33) 32

Table 2 Summary of orders of stages and the frequency of occurrence of these orders. Only the orders with frequency of at least 1% are listed. The full list can be viewed in Appendix 2.

The frequency of the life cycles table shows two most likely life cycles for an app venture. The first life cycle has a startup stage with an average length of 1.76 years, followed by a rapid growth stage with an average length of 1.45 years and concluded by a decline stage with an average length of 3.66 years. So, on average these apps have a lifespan of just over 3 years before they fall into decline. An example of such an app can be seen in Figure 4, where the graphs of the number of reviews and the growth of the number of reviews are shown for the app BlackBerry Messenger by BlackBerry Ltd. BlackBerry Messenger has a startup stage in the years 2009-2012, followed by a rapid growth stage in 2013 and a decline stage in 2014 and beyond.

The second life cycle has a startup stage of only 1 year in all cases and directly moves into the decline stage with an average length of 4.5 years afterwards.

(29)

5.1.1 Maturity stage

As can be seen in Table 2 and the full table in Appendix 2, the most frequent life cycle with a maturity stage occurred only 68 times. This is less than 4% of the total apps with at least a 1,000 reviews. Of the 1,786 apps with at least a 1,000 reviews only 125 (≈7%) have a maturity stage. In 113 of those maturity stages it only lasted for one year. In the other 12 cases it lasted 2 years. Thus, it is remarkable that the maturity stage did only occur in about 7% of the life cycles, while it was expected to be present in all of them.

5.1.2 Testing of Hypothesis 1

Hypothesis 1: It is expected that apps experience a life cycle that follows a sequence of startup, rapid growth, maturity, and decline.

Based on the results this hypothesis is not correct in two aspects: firstly, the maturity stage does only occur in 7% of the apps and thus is not part of the general life cycle of an app venture as explained in section 5.1.1. Secondly, there is not one life cycle that covers a majority of the apps. However, there are two life cycles that cover respectively 47% and 40% of the apps. So to summarize this: Hypothesis 1 is rejected. However, an app venture is found to have a life cycle of either startup stage, rapid growth stage and decline stage or startup stage and decline stage.

5.2 Ratings & Testing of Hypothesis 2

The found average ratings in each of the researched stages are listed in Table 3. Also worth noting is the fact that the average rating of all reviews of all apps with at least a 1,000 reviews was 8.00 half stars. Thus, the average ratings in the startup, rapid growth and decline stage are all above average, while the average rating of the maturity and the other stage were below this rating.

(30)

Stage Average rating Standard Deviation 95% CI Apps with this stage Reviews in Stage Startup 8.44 1.14 [8.39, 8.49] 1,786 4,665,812 Rapid growth 8.16 1.23 [8.09, 8.24] 1,073 5,983,628 Maturity 7.51 1.63 [7.23, 7.80] 125 291,516 Decline 8.06 1.12 [8.01, 8.12] 1,784 5,282,857 Other 7.58 1.63 [6.14, 9.01] 5 258,908

Table 3 Stages of the life cycle of an app venture and the average rating in them

Aside from the other stage all the average ratings of each stage only fit in their own 95% confidence interval, which means that the ratings in each of the stages are significantly different on the 95% significance level.

For both of the frequent occurring life cycles of apps the ratings decline as the app progresses through the stages. This contradicts the suggestion that apps get rated better over time, due to for example updates or the lack of updates.

5.2.1 Testing of Hypothesis 2

The researched hypothesis is:

The ratings in each of the stages are significantly different. They show a downward

movement when chronologically progressing through the stages of both the found life cycles. Therefore Hypothesis 2 is rejected. However, it is found that an app rating has different ratings in the stages of its life cycle and that this rating declines with each next stage.

Hypothesis 2 also has three sub-hypotheses, these will be tested below.

(31)

This hypothesis turned out to be wrong, as the rating in the rapid growth stage is significantly lower than the rating in the startup stage. Thus, this hypothesis is rejected.

Sub-hypothesis 2b: maturity stage has a higher rating than rapid growth stage.

This hypothesis is rejected as the maturity stage has a lower rating than the rapid growth stage.

Sub-hypothesis 2c: decline stage has a lower rating than maturity stage.

(32)

6 Discussion

In the previous chapter both Hypotheses were tested and rejected. In this chapter possible causes are described.

6.1 Life cycle of an app venture

As described in the previous chapter Hypothesis 1 was rejected. This was partly caused by the maturity stage that did not occur in the life cycle of the majority of the apps. It might be that this stage did not appear because of the speed of changing hypes, apps are subject to hypes and this influences a lot of people to move from one app to another in a short period of time. This might also be influenced by the ease of switching from one app to another as the barrier to download a new app and remove the old one is very low.

Another difference between Hypothesis 1 and the results is that there are two major life cycles instead of one. This also matches the research by Levie and Lichtestein (2010) and Mitra and Pingali (1999), where they state that different life cycles can occur and that number of stages can differ. Approximately half of the apps with at least a 1,000 reviews went from startup stage directly into decline stage, possibly because the app did not get enough support from the developers or because the app was built for a time specific event, such as the

Olympics. The other approximate half of the apps went from startup stage into rapid growth stage and finally into decline stage, these apps possibly got longer support from the

developers or were relevant for a longer period.

6.2 Ratings

Hypothesis 2 was rejected because the ratings did not correlate with the stage as expected. The ratings declined as the went to a next stage. This might be because the users are more forgiving in the early stages of an app and later get fed up with bugs that have not been fixed

(33)

or features that have not been improved or did not get updates at all or that the new updates also introduced new problems.

Another finding was that the average rating of the other stage also fits in the

confidence interval of the maturity stage. Therefore, the other stage and the maturity stage do not have a significant difference in rating. This can be explained by the fact that there are several similarities between the maturity stage and the other stage: the other stage is a stage of slow growth between two stages of rapid growth and therefore also is a stage of slow growth after a stage of rapid growth just like the maturity stage. In other words, the other stage is similar to the maturity stage except for the fact that is does not end in decline.

Another remarkable fact is that the maturity stage and the other stage both have a lower rating than the other stages. However, it might be that the apps with a maturity or other stage are rated lower through all of their stages, which would explain the low average rating in the maturity and other stage. And as these apps are only a small portion of the apps, they only influence the average ratings for the startup stage, the rapid growth stage and the decline stage marginally compared to the apps without maturity or other stage.

(34)

7 Conclusion

The goal of this research was to find a life cycle for app ventures and to see whether there are differences in the ratings of apps in the different stages of this life cycle.

7.1 Research questions

The first hypothesis is:

Hypothesis 1: It is expected that apps experience a life cycle that follows a sequence of startup, rapid growth, maturity, and decline.

The hypothesis turned out to be somewhat right: there is not one life cycle that fits all apps, there are two life cycles that cover most apps. However, these life cycles did not match the life cycle that is proposed in Hypothesis 1. Therefore Hypothesis 1 is rejected. As follows from our results, the two most prevalent types of life cycle consist of startup stage, rapid growth stage and decline stage (47% of the apps) and startup stage and decline stage (40% of the apps). In this research the maturity stage appeared to be absent for the majority of apps as it only occurred for 7% of the apps. In other words, the hypothesis underweighted the facts that apps, like companies, often fail, and that not all life cycles are similarly distributed in terms of their stages.

The second hypothesis is:

The first part of the hypothesis turned out to be correct, in this study was found that an app venture has a different rating in the different stages of its life cycle. The second part of the

(35)

hypothesis, however, was incorrect. The rating was not positively related to the size of the app venture.

The main research question is:

To what extend are apps rated differently in the stages of their life cycle?

The average ratings in each stage were found to be significantly different: startup stage 8.44 half stars, rapid growth stage 8.16 half stars and decline stage 8.06 half stars. Converted to normal stars the startup stage has 4.22 stars, the rapid growth stage has 4.08 stars and the decline stage has 4.03 stars. In short, for each further stage the rating decreases, which seems strange given that apps become better over time, as they get updated by their developers.

7.2 Limitations & Recommendations for future research

The largest limitation of this study consists of the fact that, while studying apps and their life cycles, the best measurable variable to gauge said developments turned out to be reviews per year. Possible measurable variables such as the number of downloads over time, number of views in the app store, number of minutes of usage per month etcetera are unavailable because they consist of sensitive competitive information. Therefore, the resulting sample does not quite represent all 288,000 apps in the BlackBerry app store. It does represent the apps that are relevant in the course of answering our research question, because the sample size of 1,786 apps with more than a 1,000 reviews does represent every app that has had the opportunity to experience a life cycle at all and it does represent the apps that get ratings at all. The apps that are not included in this research either get close to no downloads at all or get only downloaded by a limited number of users but did not get a 1,000 reviews. When an app gets no downloads at all, it does not experience a life cycle. When an app gets

(36)

downloaded by only a limited number of users and did not get a 1,000 reviews, then the app is probably represented by the apps included in this study. However, these cases could be

researched more in dept in a future study.

The number of reviews is not the perfect measure for app size, since writing reviews are onetime actions, while app size is about the actual usage over a period. Reviews and especially the number of reviews are influenced by problems within the app or requests for reviews by the developer. It is recommended to repeat this research with minutes of usage in a certain period as measure for app size.

While Nayebi, Cho and Ruhe (2018) state that app reviews are often used to analyze the evolution of apps, reviews are only snapshots of usage. The reviews do not perfectly match actual usage. However, the number of reviews over a period seems to approach the actual usage. It would be recommended to do more research into this connection.

This research only used data from BlackBerry World and does not include data from the other app stores. It is recommended to repeat this research for other app stores to research the generalization of the found life cycles and the ratings in each stage.

For practical reasons a timespan of a year is chosen as measure for time in this research, however that may have caused measurement errors. Yan and Zhao (2010) required life cycle stages to be at least 2 years long, although they measured time in quarters, so this may not be a problem. However, it is recommended to repeat this research with time measured in a smaller timespan, for example months. However, this also requires a more accurate measure for app size.

This research measured absolute growth of each app and compared that to the maximum absolute growth in a year for that app. A cutoff point of 50% of this maximum absolute growth was used to make a distinction between high and low growth. It is

(37)

recommended to repeat this research with variations of this cutoff point, just like in the research by Yan and Zhao (2010). Also, this research could have been done with relative growth. It is recommended to do a future research into using relative growth in the life cycle model and repeat this research with relative growth as growth measure.

This research used the fact that apps remain the same app over time. Updates to the apps might change the app to such an extent that you could argue that it has become another app. It is recommended to do a study into the changes of apps over time.

7.3 Concluding remark

Apps do have a life cycle with a similar form as the life cycle of the firm model. The main three differences are: there are two possible life cycles for app ventures, the timespan of these life cycles is much shorter and there is no maturity stage present. Each stage of the life cycle has a different rating and these ratings are negatively related to the age of the app.

This is the first research in this direction and although there are conclusive results there is a lot of opportunity for future research.

(38)

8 References

Adizes, I. (1979). Organizational Passages: Diagnosing and Treating Lifecycle Problems of Organizations. Organizational Dynamics, 8(1), 2.

Apptentive (2015) Increasing Star Ratings Drives Mobile App Revenue. Retrieved from

https://go.apptentive.com/rs/170-TZF-108/images/IncreasedStarRatingsDriveAppRevenue.pdf

Beverland, M., & Lockshin, S.H. (2001). Organizational life cycles in small New Zealand wineries. Journal of Small Business Management, 39(4), 354.

Carter, S., & Yeo, A. C. M. (2016). Mobile apps usage by Malaysian business undergraduates and postgraduates: implications for consumer behaviour theory and marketing

practice. Internet Research, 26(3), 733-757.

Churchill, N. C., & Lewis, V. L. (1983). The five stages of small business growth. Harvard Business Review, 61(3), 30.

The Coca-Cola Company (2015) Coca-Cola Human Rights. Retrieved from https://play.google.com/store/apps/details?id=com.ko.humanrights

D66 (2017) D66 Nu App voor Android & iPhone. Retrieved from https://d66.nl/d66-app/

Datta, A., Kajanan, S., & Pervin, N. (2013). A mobile app search engine. Mobile Networks and Applications, 18(1), 42-59.

Davenport, T. H. (2013). Analytics 3.0. Harvard Business Review, 91(12), 64-72.

Dodge, H. R., Fullerton, S., & Robbins, J. E. (1994). Stage of the organizational life cycle and competition as mediators of problem perception for small businesses. Strategic

(39)

Downs, A. (1964) The Life Cycle of Bureaus. Inside Bureaucracy, 296-309.

Ehrenhard, M., Wijnhoven, F., van den Broek, T., & Stagno, M. Z. (2017). Unlocking how start-ups create business value with mobile applications: Development of an App-enabled Business Innovation Cycle. Technological forecasting and social change, 115, 26-36.

Eneco (2018). Eneco Toon. Retrieved from https://www.eneco.nl/energieproducten/toon-thermostaat/toon-app/

Gill, M., Sridhar, S., & Grewal, R. (2017). Return on engagement initiatives: A study of a business-to-business mobile app. Journal of Marketing, 81(4), 45-66.

Godwin-Jones, R. (2011). Emerging technologies: Mobile apps for language learning. Language Learning & Technology, 15(2), 2–11.

Greiner, Larry E. (1972) Evolution and revolution as organizations grow. Harvard Business Review, 50(4), 37.

Han Rebekah Wong, S. (2012). Which platform do our users prefer: website or mobile app?. Reference Services Review, 40(1), 103-115.

Hanks, S. H. (1990). The organization life cycle: Integrating content and process. Journal of Small Business Strategy, 1(1), 1-12.

Hao, L., Guo, H., & Easley, R. F. (2017). A mobile platform's in‐app advertising contract under agency pricing for app sales. Production and Operations Management, 26(2), 189-202.

Huang, H. C. (2016). Freemium business model: construct development and measurement validation. Internet Research, 26(3), 604-625.

(40)

Jaafar, H., & Halim, H. A. (2016). Refining the firm life cycle classification method: A firm value perspective. Journal of Economics, Business and Management, 4, 112-119.

Jung, E. Y., Baek, C., & Lee, J. D. (2012). Product survival analysis for the App Store. Marketing Letters, 23(4), 929-941.

Lee, S. M., Kim, N. R., & Hong, S. G. (2017). Key success factors for mobile app platform activation. Service Business, 11(1), 207-227.

Levie, J., & Lichtenstein, B. B. (2010). A terminal assessment of stages theory: Introducing a dynamic states approach to entrepreneurship. Entrepreneurship Theory and practice, 34(2), 317-350.

Maalej, W., Kurtanović, Z., Nabil, H., & Stanik, C. (2016). On the automatic classification of app reviews. Requirements Engineering, 21(3), 311-331.

Masurel, E., & Van Montfort, K. (2006). Life cycle characteristics of small professional service firms. Journal of Small Business Management, 44(3), 461-473.

Mitra, R., & Pingali, V. (1999). Analysis of growth stages in small firms: A case study of automobile ancillaries in India. Journal of Small Business Management, 37(3), 62.

Müller, R. M., Kijl, B., & Martens, J. K. (2011). A comparison of inter-organizational business models of mobile app stores: There is more than open vs. closed. Journal of theoretical and applied electronic commerce research, 6(2), 63-76.

Nayebi, M., Cho, H., & Ruhe, G. (2018). App store mining is not enough for app improvement. Empirical Software Engineering, 23(5) 1-31.

O’Farrell, P. N., & Hitchens, D. M. (1988). Alternative theories of small-firm growth: a critical review. Environment and Planning A, 20(10), 1365-1383.

(41)

Paget, L., & Frosch, D. L. (2016). What will it take to reduce the app gap?. Journal of General Internal Medicine, 31(12), 1408-1409.

Phelps, R., Adams, R., & Bessant, J. (2007). Life cycles of growing organizations: A review with implications for knowledge and learning. International Journal of Management Reviews, 9(1), 1-30.

PostNL Holding B.V. (2018) PostNL App. Retrieved from https://www.postnl.nl/campagnes/postnl-app/

Quinn, R. E., & Cameron, K. (1983). Organizational life cycles and shifting criteria of effectiveness: Some preliminary evidence. Management science, 29(1), 33-51.

Rieger, C., & Kuchen, H. (2018). A process-oriented modeling approach for graphical development of mobile business apps. Computer Languages, Systems & Structures, 53, 43-58.

Scott, M., & Bruce, R. (1987). Five stages of growth in small business. Long range planning, 20(3), 45-52.

Shuler, C., Levine, Z., & Ree, J. (2012). iLearn II: An analysis of the education category of Apple’s app store.

Tong, Y. X., She, J., & Chen, L. (2015). Towards better understanding of app functions. Journal of Computer Science and Technology, 30(5), 1130-1140.

Wan, J., Zhao, L., Lu, Y., & Gupta, S. (2017). Evaluating app bundling strategy for selling mobile apps: an ambivalent perspective. Information Technology & People, 30(1), 2-23.

Yan, Z., & Zhao, Y. (2010). A new methodology of measuring firm life-cycle stages. International Journal of Economic Perspectives, 4(4), 579-587.

(42)

Zhang, L., Huang, X. Y., Jiang, J., & Hu, Y. K. (2017). CSLabel: An Approach for Labelling Mobile App Reviews. Journal of Computer Science and Technology, 32(6), 1076-1089.

(43)

9 Appendices

9.1 Appendix 1 PHP code for analysis

9.1.1 Detecting stages in the yearly review data per app <?

error_reporting(E_ALL); require_once('config.php'); $dontanalyse2018 = true; $treshholdrapidgrowth = 0.5;

function getWeigthedAverageRatingInPeriod($startyear, $endyear, $data){

if($startyear > 0&& $endyear>0 && array_key_exists($startyear, $data) && array_key_exists($endyear, $data)&& $startyear<=$endyear){

$res = 0; $reviews = 0;

$years=$endyear - $startyear + 1; for($i=$startyear;$i<=$endyear;$i++){

$res = $res + $data[$i]['AverageRating']*$data[$i]['NumberOfReviews']; $reviews = $reviews + $data[$i]['NumberOfReviews'];

} return $res/max($reviews,1); }else{ return 0; } } function getRatingsPerStage($data){ $result=['s'=>0,'r'=>0,'m'=>0,'d'=>0,'o'=>0]; $numberofreviews=['s'=>0,'r'=>0,'m'=>0,'d'=>0,'o'=>0]; $totalrating=['s'=>0,'r'=>0,'m'=>0,'d'=>0,'o'=>0]; foreach($data as $year=>$value){ $stage=$value['Stage']; $numberofreviews[$stage] = $numberofreviews[$stage] + $value['NumberOfReviews']; $totalrating[$stage] = $totalrating[$stage] + $value['NumberOfReviews']*$value['AverageRating']; }

foreach($result as $stage => $value){

$result[$stage] = $totalrating[$stage] / max($numberofreviews[$stage], 1); }

return $result; }

function analyzeApp($appid){

global $_db, $dontanalyse2018, $treshholdrapidgrowth; if($appid>=0){

$sql = 'SELECT * FROM `CumulatedYearResults` WHERE `idApps` = :appid

ORDER BY `CumulatedYearResults`.`Year` ASC '; $stmt = $_db->prepare($sql);

(44)

$stmt->execute(array(':appid'=>$appid)); $results = $stmt->fetchAll();

$c = $stmt->rowCount(); if($c>0){

if(!($c==1 && $dontanalyse2018 && $results[0]['Year'] == 2018)){ $data = array();

$years = array();

$previousreviews = $results[0]['NumberOfReviews']; $previousrating = $results[0]['AverageRating']; foreach($results as $row){

if(!($row['Year'] == 2018 && $dontanalyse2018 )){

$data[$row['Year']]['NumberOfReviews']=$row['NumberOfReviews']; $data[$row['Year']]['AverageRating']=$row['AverageRating']; $data[$row['Year']]['ReviewsGrowth']=$row['NumberOfReviews']-$previousreviews; $data[$row['Year']]['RatingGrowth']=$row['AverageRating'] - $previousrating; $previousreviews = $row['NumberOfReviews']; $previousrating = $row['AverageRating']; $years[]=$row['Year']; } } $firstyear = min($years); $lastyear = max($years); if($lastyear-$firstyear>=count($years)){ print_r($years); //die('jaar gat'); echo 'jaar gat';

for($i = $firstyear+1; $i <2018;$i++){ if(!array_key_exists($i, $data)){ $years[]=$i; $data[$i]['NumberOfReviews']=0; $data[$i]['AverageRating']=0; $data[$i]['ReviewsGrowth']=0 - $data[$i-1]['NumberOfReviews']; $data[$i]['RatingGrowth']=0 - $data[$i-1]['AverageRating']; if(array_key_exists($i+1, $data)){ $data[$i+1]['ReviewsGrowth']=$data[$i+1]['NumberOfReviews']; $data[$i+1]['RatingGrowth']=$data[$i+1]['AverageRating']; } } } } sort($years); ksort($data); echo '<pre>'; print_r($data); echo '</pre>';

(45)

echo 'lastyear:' . $lastyear . ' ';

$reviews = array_column($data, 'NumberOfReviews'); $top = max($reviews);

echo 'top:' . $top . ' ';

$topyear =$years[array_search($top,$reviews)]; echo 'topyear:' . $topyear . ' ';

$reviewsgrowth = array_column($data, 'ReviewsGrowth'); $topgrowth = max($reviewsgrowth);

echo 'top growth:' . $topgrowth . ' ';

$topgrowthyear =$years[array_search($topgrowth,$reviewsgrowth)]; echo 'top growth year:' . $topgrowthyear . ' ';

$topgrowth = max($topgrowth, 1); $bottomgrowth = min($reviewsgrowth); echo 'bottom growth:' . $bottomgrowth . ' ';

$bottomgrowthyear =$years[array_search($bottomgrowth,$reviewsgrowth)]; echo 'bottom growth year:' . $bottomgrowthyear . ' ';

if($bottomgrowthyear<$topgrowthyear){

//echo 'in statup/rapid growth or maturity stage'; $startupend = 0; $rapidgrowthstart = 0; $rapidgrowthend = 0; $maturitystart = 0; $maturityend = 0; $declinestart = 0; if($topgrowthyear<$lastyear){ if($data[$lastyear]['ReviewsGrowth']/$topgrowth <= $treshholdrapidgrowth){

echo 'maturity stage'; }else{

echo 'rapid growth over peak'; }

}elseif($topgrowthyear==$lastyear){

echo 'rapid growth or startup stage at peak'; }else{

die('cannot occur'); }

}elseif($bottomgrowthyear==$topgrowthyear){

echo 'growth constant, so invalid? / not enough data'; $startupend = 0; $rapidgrowthstart = 0; $rapidgrowthend = 0; $maturitystart = 0; $maturityend = 0; $declinestart = 0; }else{

echo 'decline stage'; $startupend = $firstyear; $rapidgrowthstart = $firstyear; for($i=$firstyear;$i<$topgrowthyear;$i++){ if($data[$i]['ReviewsGrowth']<$topgrowth*$treshholdrapidgrowth){ $startupend = $i; $rapidgrowthstart = $i + 1;

(46)

}else{ $i=$topgrowthyear; } } $rapidgrowthend = $topgrowthyear; $maturitystart = $topgrowthyear; for($i=$topgrowthyear;$i<=$lastyear;$i++){ if($data[$i]['ReviewsGrowth']>$topgrowth*$treshholdrapidgrowth){ $rapidgrowthend = $i; if(array_key_exists($i+1, $data)){ if($data[$i+1]['ReviewsGrowth']>=0){ $maturitystart = $i + 1; }else{//geen maturity $maturitystart = 0; } }else{ $maturitystart = 0; } }else{ $i=$lastyear+1; } } $maturityend = $maturitystart; $declinestart = $rapidgrowthend+1; if($maturitystart!=0){ for($i=$topgrowthyear;$i<=$lastyear;$i++){ if($data[$i]['ReviewsGrowth']>=0){ $maturityend = $i; if(array_key_exists($i+1, $data)){ $declinestart = $i + 1; }else{ $declinestart = 0; } }else{ $i=$lastyear+1; } } } echo ' ';

echo 'startupstart: ' . $firstyear . ' '; echo 'startupend: ' . $startupend . ' ';

echo 'rapidgrowthstart: ' . $rapidgrowthstart . ' '; echo 'rapidgrowthend: ' . $rapidgrowthend . ' '; echo 'maturitystart: ' . $maturitystart . ' '; echo 'maturityend: ' . $maturityend . ' '; echo 'declinestart: ' . $declinestart . ' '; }

$startuprating = getWeigthedAverageRatingInPeriod($firstyear, $startupend, $data);

$rapidgrowthrating = getWeigthedAverageRatingInPeriod($rapidgrowthstart, $rapidgrowthend, $data);

(47)

$declinerating = getWeigthedAverageRatingInPeriod($declinestart, $lastyear, $data);

echo 'startuprating: ' . $startuprating . ' ';

echo 'rapidgrowthrating: ' . $rapidgrowthrating . ' '; echo 'maturityrating: ' . $maturityrating . ' '; echo 'declinerating: ' . $declinerating . ' ';

$previousstage = 's';//s startup, r rapid growth, m maturity, d decline, o other $hashadrapid = false;

foreach($data as $key => $value){ $r = $value['ReviewsGrowth']; $s = 's';

if($r >= $topgrowth * $treshholdrapidgrowth){ $s = 'r'; $hashadrapid = true; }elseif($r<0){ $s = 'd'; }else{ if($hashadrapid){ $endofrapid = true;

for($i = $key+1; $i<=$lastyear;$i++){

if($data[$i]['ReviewsGrowth'] >= $topgrowth * $treshholdrapidgrowth){ $endofrapid=false; $i=$lastyear+1; } } if($endofrapid){ if($previousstage == 'd' && $r==0){ $s = 'd'; }else{ $s = 'm'; } }else{ $s = 'o'; } }else{ if($previousstage == 'd' && $r==0){ $s = 'd'; }else{ $s = 's'; } } } $previousstage = $s; $data[$key]['Stage'] = $s; } $newratings = getRatingsPerStage($data); print_r($newratings); $y=[];

for($i=2009; $i<=2018; $i++){

if(array_key_exists($i, $data)){ $y[$i]=$data[$i]['Stage']; }else{

(48)

$y[$i]=''; }else{ $y[$i]=''; } } } echo ' '; print_r($y);

$sqlu = 'UPDATE `Apps` SET `Done` = 1,

`StartupStart`=:startupstart,`StartupEnd`=:startupend,`RapidgrowthStart`=:rapidgrowthstart,`Rapidgro wthEnd`=:rapidgrowthend,`MaturityStart`=:maturitystart,`MaturityEnd`=:maturityend,`DeclineStart`= :declinestart,`StartupRating`=:startuprating,`RapidgrowthRating`=:rapidgrowthrating,`MaturityRating` =:maturityrating,`DeclineRating`=:declinerating,`2009`=:2009,`2010`=:2010,`2011`=:2011,`2012`=:2 012,`2013`=:2013,`2014`=:2014,`2015`=:2015,`2016`=:2016,`2017`=:2017,`2018`=:2018,`NStartupR ating`=:nstartuprating,`NRapidgrowthRating`=:nrapidgrowthrating,`NMaturityRating`=:nmaturityratin g,`NDeclineRating`=:ndeclinerating,`NOtherRating`=:notherating WHERE `idApps` = :appid';

$stmtu = $_db->prepare($sqlu);

$stmtu->execute(array(':startupstart'=>$firstyear, ':startupend'=>$startupend, ':rapidgrowthstart'=>$rapidgrowthstart, ':rapidgrowthend'=>$rapidgrowthend,

':maturitystart'=>$maturitystart, ':maturityend'=>$maturityend, ':declinestart'=>$declinestart, ':startuprating'=>$startuprating*100, ':rapidgrowthrating'=>$rapidgrowthrating*100,

':maturityrating'=>$maturityrating*100, ':declinerating'=>$declinerating*100, ':2009'=>$y[2009], ':2010'=> $y[2010], ':2011'=> $y[2011], ':2012'=> $y[2012], ':2013'=> $y[2013], ':2014'=> $y[2014], ':2015'=> $y[2015], ':2016'=> $y[2016], ':2017'=> $y[2017], ':2018'=> $y[2018],

':nstartuprating'=>$newratings['s']*100, ':nrapidgrowthrating'=>$newratings['r']*100, ':nmaturityrating'=>$newratings['m']*100,

':ndeclinerating'=>$newratings['d']*100,':notherating'=>$newratings['o']*100, ':appid'=>$appid)); echo ' success:' . $stmtu->rowCount();

}else{

echo 'geen data';

$sqlu = 'UPDATE `Apps` SET `Done` = 1 WHERE `idApps` = :appid'; $stmtu = $_db->prepare($sqlu); $stmtu->execute(array(':appid'=>$appid)); } } }else{

echo 'mag niet negatief'; } } if(isset($_GET['i'])){ $appid = (int)$_GET['i']; analyzeApp($appid); }else{

echo 'geen i'; }

?>

9.1.2 Analysis of orders of stages and their lengths <?

error_reporting(E_ALL); require_once('config.php');

(49)

switch($char){ case 's':

return 'Startup'; break;

case 'r':

return 'Rapid growth'; break; case 'm': return 'Maturity'; break; case 'd': return 'Decline'; break; case 'o': return 'Other'; break; } } $sql = 'SELECT *, CONCAT(`2009`, `2010`, `2011`, `2012`, `2013`, `2014`, `2015`, `2016`, `2017`, `2018`) as `order` FROM `Apps` WHERE `NumberOfReviews` >= 1000 ORDER BY `NumberOfReviews` ASC '; $stmt = $_db->prepare($sql); $stmt->execute(); $results = $stmt->fetchAll(); $c = $stmt->rowCount(); if($c>0){ $orders = array(); foreach($results as $row){ $order = preg_replace('{(.)\1+}','$1',$row['order']); if(array_key_exists($order, $orders)){ $orders[$order]['count'] = $orders[$order]['count'] + 1; $orders[$order]['lengthstrings'][] = $row['order']; }else{ $orders[$order]['count'] = 1; $orders[$order]['lengthstrings']=[]; $orders[$order]['lengthstrings'][] = $row['order']; } } uasort($orders, function($a, $b) { return $b['count'] - $a['count']; });

echo '<table>';

foreach($orders as $order => $value){ $res = []; foreach($value['lengthstrings'] as $val){ $k=0; for($i=0;$i<strlen($val);$i++ ){ $j=1; $char = substr($val,$i,1); while(substr($val,$i+1,1)==$char){ $j++; $i++; }

(50)

$res[$val][$k] = $j; $k++; } } print_r($res); $resstring = ''; for($i=0;$i<strlen($order);$i++){ $char = substr($order,$i,1); $lengths = array_column($res, $i);

$avglength = round(array_sum($lengths)/count($lengths),2); $resstring .= charToStage($char) . ' (' . $avglength . ') '; }

echo '<tr><td>' . $resstring . '</td><td>' . $value['count'] . '</td></tr>'; }

echo '</table>'; print_r($orders); }else{

echo 'no results'; }

?>

9.1.3 Analysis of the number of reviews per stage <? error_reporting(E_ALL); require_once('config.php'); $stages=['s', 'r', 'm', 'd', 'o']; $r=['s' => 0, 'r' => 0, 'm' => 0, 'd' => 0, 'o' => 0]; for($i=2009;$i<=2018;$i++){ $year = $i;

$sql = 'SELECT COUNT(*) as `c` FROM Reviews WHERE ìdApps` IN(SELECT ìdApps` FROM Àpps` WHERE `NumberOfReviews` >= 1000 AND `' . $year . '` = :stage) AND

YEAR(`Date`) = ' . $year . ''; echo $sql . ' '; $stmt = $_db->prepare($sql); foreach($stages as $stage){ echo $stage . ' '; $stmt->execute(array(':stage' => $stage)); $results = $stmt->fetchAll(); $d = $stmt->rowCount(); echo $d; if($d>0){ $c=$results[0]['c']; $r[$stage] = $r[$stage] + $c; } } } print_r($r); ?>

App ratings in different stages of an app ventures life cycle