Product variety and vertical differentiation as a tool for revenue increase of mobile application from with-in app sales of virtual products

Размещено на http://www.allbest.ru/

Размещено на http://www.allbest.ru/

NATIONAL RESEARCH UNIVERSITY HIGHER SCHOOL OF ECONOMICS,

ST. PETERSBURG

DEPARTMENT MANAGEMENT

Product variety and vertical differentiation as a tool for revenue increase of mobile application from with-in app sales of virtual products: Topface company case

Graduation qualification thesis

Major course: «Management and International Business»

M.V. Isacov

Assistant professor

E.A. Antipov

PhD in Economics

Saint-Petersburg 2018



Annotation

The problem is that people in Russia do not pay for intangible products like movies, music, mobile applications and different products in mobile applications. (Lee I., 2017). One of the most popular marketing strategy is differentiation (Chamberlin, 1933). Vertical differentiation of products distributes products in the industry market in accordance with their quality (Dennis Z.Y., 2012, pp. 314-328). In the product line literature, authors (Moorthy K.S. ,Png I.P., 1992) (Moorthy K.S., 1984) conclude that a product should have inferior quality for consumers with a low valuation and superior quality for consumers with a high valuation. The hypothesis: vertical differentiation can serve as a tool to increase the number of purchases and total revenue from sales of virtual products within the mobile application. To test our hypothesis is the main goal of this study. Accepting the hypothesis will allow us to say that vertical product differentiation is a suitable tool for increasing revenue from purchases within an application that uses mixed monetization model.

Methods of the research are experiment and statistical analysis. During the experiment, new users were randomly divided into 3 groups and the experimental group was entered a list of 4 products, and the two control groups had only one option - a monthly subscription. During the statistical analysis data distribution was checked using Shapiro-Wilk and Kolmogorov-Smirnov statistical criteria. Due to tests results the probability that we will make a mistake if will reject the null hypothesis is 0.0%.

The difference of sample medians between Eperimental group and Control Groups is significant because p-value of Kruskal-Wallis H-test is 0.0046. The differences of sample means between Eperimental group and merged Control Groups is significant because p-value of Mann-Whitney rank test is 0.0010. Difference in revenue for subscription for premium account between experimental and control groups is 30.36% (210669 cents - Experimental Group, 177247 cents -Control Group №1, 145956 cents - Control Group №2).

Based on the above, we can argue that there is a significant difference between the sample means of the experimental and control groups and our hypothesis that vertical differentiation can serve as a way to increase the revenues of the mobile application from the sale of virtual products in the mobile application is confirmed.



Table of contents

Annotation

Introduction

1. Theoretical background of mobile app monetization and vertical differentiation

• 1.1 The main concepts
• 1.2 Mobile app monetization
• 1.3 Basic ways to monetize a mobile application
• 1.3.1 Paid applications
• 1.3.2 Free application with advertising
• 1.3.3 Freemium
• 1.3.4 With-in app sales of virtual products
• 1.4 Previous studies

2. Research problem, methodology and design

• 2.1 Formulation of the research problem
• 2.2 Methodology and design of research
• 2.3 Company's description

3. Results of the experiment

• 3.1 Changes in the user's behavior
• 3.2 Data description
• 3.3 Checking data distribution using Shapiro-Wilk and Kolmogorov-Smirnov statistical criteria
• 3.3 Checking the equality of samples mean and median using the Kruskal-Wallis and Mann-Whitney non-parametric tests

Conclusion

References



Introduction

Mobile devices are becoming more popular every day around the world and ownership traditional mobile phones, smartphones and other mobile devices. Currently, the planet has more mobile devices than people (International Telecommunication Union, 2014). These personal devices become primary means of access to the Internet in developed countries as a right to participate in a mobile subscription tariffs are almost saturated both in developed countries and in developing countries. According to a recent report by ComScore (2014), consumers in the US conducted almost 60% of the time online on mobile devices. Another study Movable Ink on US that 65% of marketing letters are open on a mobile device (Burdge, B. , 2014).Among low-income countries and in rural areas with weakened infrastructure, traditional mobile phones are the most popular communication devices belonging to most of the population. In Africa, 66% of the population owns a mobile subscription (International Telecommunication Union, 2014). Mobile phones offer affordable, instant communication and mobility for all users, regardless of income or previous access to fixed telephone lines. Undoubtedly, mobile devices offer unique become an important channel for marketers around the world.

Mobile applications are software products designed specifically for mobile devices, smartphones, tablet computers or other mobile devices. Mobile applications are distributed through application stores: Apple App Store, Google Play, Windows Phone Store, etc. Mobile applications help to solve various tasks: from mobile banking and social networks to highly specialized functions. They are designed to make life easier for users of mobile devices, as well as to diversify it.

Mobile applications have surely entered the everyday life (eMarketer, 2014) and have become one of the most promising markets of the modern world. According to App Annie (AppAnnie, 2016), the mobile applications industry has created a whopping $ 41.1 billion in gross annual revenue in 2015 and $101 billion in 2020 is predicted. For comparison, gross annual revenue of oil market in 2015 has created approximately $156,6 billion.

There are different ways of monetization of mobile application. The most popular of them are: paid applications, freemium (free and app subscription for extra features), crowdfunding, advertisement and Free-to-Play. Free-to-Play refers to situation when mobile application, usually game application, can be downloaded free of charge and developers get money through selling different virtual products with-in the app that make easier for user to achieve his goals in the game, usually virtual money are designed. Hybrid models of monetization, such as built-in advertising and purchases from the application, are clearly gaining popularity in the business world.

The problem is that people in Russia do not pay for intangible products like movies, music, apps and different products in apps. That is why premium models, when users pay for the use of service are working poorly. There remains a model in which publishers can earn by displaying ads inside applications. However, advertising fees from applications in developed countries are higher (Lee I., 2017).

With growth of the market more and more entrepreneurs, developers and investors come to compete for consumer's money. That is why marketing in IT sphere and mobile applications' market is very important question nowadays. One of the most popular marketing strategy is differentiation. The term «differentiation» was proposed for the first time by Edward Chamberlin in 1933 (Chamberlin, 1933). Vertical differentiation of products distributes products in the industry market in accordance with their quality. Products in this situation satisfy the same need, but they are intended for consumers with different incomes. Proposal of vertically differentiated products attract people with different income levels and can make a market niche.

The hypothesis: vertical differentiation can serve as a tool to increase the number of purchases and total revenue from sales of virtual products within the mobile application.

The research question: Is it possible to use vertical differentiation of the product to increase revenue from the sale of virtual goods in a mobile application.

Research problem is that Russian users do not pay for using the application and rarely make purchases with-in application.

We have chosen a subscription for premium account as a product. It is so-called Freemium. More than half (61%) of application development companies recommend to monetize the application using the freemium model, in which the application can be downloaded for free, and users are later offered paid upgrades to access additional features or the ability to make purchases to the application (Delgado, 2017). So we can state that validity of the research is proved.

Picture №1 - Recommended App Monetization Strategies (2017)



Therefore, the study of the validity of the marketing concept of vertical differentiation for this particular case seems to be an interesting and valid problem. This study will help to test the effectiveness of this strategy for virtual products for increasing the revenue. In this particular case, it is premium subscription.

Objects, field of study, main goal

Field of study of this scientific paper is marketing and product differentiation.

To test our hypothesis is the main goal of this study. Accepting the hypothesis will allow us to say that vertical product differentiation is a suitable tool for increasing revenue from sales within an application that uses the Freemium monetization model with subscription.

Design of research

Design of this research is case study. The study is conducted on the application with hybrid monetization model from the Dating segment. For differentiation, a subscription to premium account have been chosen. During the experiment, parts of new users were randomly divided into 3 groups and the experimental group was entered a list of 4 products (subscription for a week, a subscription for a month, a subscription for 3 months, a subscription for 6 months), and two control groups had only one option - subscription for a month.

Methodology of research

Field experiment, statistical analysis.

To reach our main goal we should complete some objects:

a) to conduct an experiment;

b) to collect the data;

c) to analyze the data;

d) to make a conclusion.



Theoretical background of mobile app monetization and vertical differentiation

1.1 The main concepts

This work is based on two directions - monetization of mobile applications and product differentiation. Firstly, we will discuss monetization theme, it's topical issues and articles. Then will recall insights from product variety and vertical differentiation literature. Our paper is at the junction of these two directions, therefore we need to review both of them.

The main concepts of our research are:

- mobile application;

- monetization;

- vertical differentiation.

A mobile application is a program developed for a mobile device. Mobile applications usually are alternatives for desktop applications that run on desktop computers and with web applications. The shortening of mobile application - "app" in 2010 was included in the "Word of the Year" list of the American Dialects Society. (2011) In high technology and marketing industries monetization refers to forcing assets to create income. Sites and mobile applications that make profit are usually monetized through advertising, subscription fees, or (in the case of mobile) purchases in the app. In the music industry, monetization occurs when the designer posts the video on the Internet and ads are displayed. (Wixen, Randall W. , 2005)

Vertical differentiation of products distributes products in the industry market in accordance with their quality. Products in this situation satisfy the same need, but they are intended for consumers with different incomes. Proposal of vertically differentiated products attract people with different income levels and can make a market niche.



1.2 Mobile app monetization

It was discovered (Kim, M., Kim, J., Choi, J., Trivedi, M., 2017) that there are several significant factors that have strong impact on with-in mobile app possession. These factors are the experience of working on the Internet and mobile experience, browser behavior for non-commercial applications helps to understand the ownership of trading applications, as it reflects user preferences for acquiring more applications. Moreover, possessing decisions were explained by digital experience (ie, experience on the Internet and mobile experience) and viewing information from trading applications, while other factors have little predictive value. So, we can say that the much time a person spends in the Internet and uses his mobile phone, the much money he will potentially spend in different mobile applications.

In general, market research of mobile applications tells us that this is one of the most promising areas, and also recently becomes the most popular source of media among users.

In many cases, the mobile application is not part of the marketing strategy, but the basis of the business strategy, the largest, if not the only, income item (Cossa, 2015).

A market research of J'son & Partners Consulting, published in March 2013, is one of the most frequent studies on the analysis of the volume of the mobile applications market. According to J'son & Partners Consulting, in 2012 the mobile applications market in the world amounted to $ 7.83 billion and by 2016 will be $ 65.79 billion. These figures differ slightly from different researchers, but it remains obvious that the mobile market is really large. The study also has a detailed description of each business model: What is it remarkable about? And how effective is it in practice? What are its main advantages and disadvantages in the implementation? (ConsultancyManagement, 2014)

Very important thing for a mobile app monetization is the attitude to this particular app, its brand and image. Researchers from USA (Yang, 2013) made a model to predict young users' attitude to the application. The model was tested by a web survey of 555 American college students in the winter of 2011. The results showed that the moods and intentions of young American consumers predict the use of mobile applications. Perceived pleasure, utility, ease of use and subjective norm appear as important predictors of their mobile applications. Perceived behavioral control, usefulness and use of mobile Internet predict their intention to use mobile applications. Their use of mobile applications is determined by perceived utility, intent to use, mobile Internet use, income and sex. We can assume that the situation among Russian users is the similar, because the main part of popular applications are global and globalization trend is very strong nowadays.

Many marketing experts believe that the mobile device is an extremely promising marketing tool for overcoming the main problems associated with obtaining time and attention of consumers. Aggressive companies create and implement their own applications to remain competitive in the global market. During the implementation phase, the consumer response is very positive, as many people signed the mobile shopping application, and the number of users visiting the website via the mobile application was also very high. The real purchases and transaction rarely occur through a mobile application in which the amount of actual purchases and transactions is lower in comparison with sales recorded through the company's official website. (Musa, R. ; Saidon, J.; Harun, M.H. ; Adam, A.A. ; Dzahar, D.F. ; Haussain, S.S. ; Lokman, W.M., 2015)

The findings of the research on users behavior with the online survey with 218 respondents showed that the functions of mobile applications are the most influential predictor of consumer attitudes, while security and confidentiality are the least influential predictor. Also, it was found that the intentional behavior of the main consumers is the exchange of information. These statements prove the arguments about globalization trends, which were mentioned before.

Moreover, mobile applications on the person's device can say a lot about the person. This statement was tested and approved in a very interesting research. (Unal P., Temizel T.T., Eren P.E., 2017) Scientists specifically made a mobile application to conduct an experiment, willing to learn is there a connection between personality and applications that person downloads on his mobile phone. It was found that good faith is positively associated with the adoption of recommended applications and the acceptability is related to the preferences of the editor's selection applications. In addition, this study explores user-owned applications and the composition of applications by category and their relationship to personal characteristics. It is shown that the number of applications belonging to the user and their category differ depending on gender and personality. The presence of such applications and the number of applications belonging to certain categories increase the likelihood of adopting the recommended applications.

1.3 Basic ways to monetize a mobile application

1.3.1 Paid applications

Perhaps, the most understandable version of monetization - the user pays once for downloading your application. You, as the owner, receive income immediately from download. Users of paid applications are usually more loyal - they are easier to keep. With paid downloads, your task is to get as many buyers as possible, but it's rather difficult. People are reluctant to pay for what they have not tried. Demand and free application-analogs, which provide similar opportunities, are brought down. Yet, despite the difficulties, in 2017, paid applications brought their owners $ 29 billion (AppTractor, 2016).

1.3.2 Free application with advertising

Advertising is one of the most common ways to generate income from the mobile app. There are no limitations on downloading, but the goal of the developer is to get as many users as possible. Analyzed data about purchases is provided to advertisers, who want to pay for advertising. The problem is that most of people attitude badly to advertisement. It can make using the app uncomfortable and even can lead to deleting the app. As was figured out, irritation has a strong and significant negative effect on attitude to the mobile advertisement. Turkish researchers Gцkhan Aydin and Bilge Karamehmet made a research (Aydin G., Karamehmet B., 2017) where they compared users' attitude to the advertisement from SMS and in mobile applications. The study stated that entertainment has strong positive effect on the people's attitude to the advertisement. In comparison with the SMS advertisement, informativeness has much less positive effect, even less that credibility. As was mentioned, irritation had a significant negative effect, but only for mobile application advertisement, because SMS ads do not disturb as much as mobile ads do. Development of mobile advertising as fun and pleasant as possible, perhaps, may help to overcome the general negative attitudes.

1.3.3 Freemium

Freemium - the application has a set of basic and featured functions that users can purchase.The goal of this model is to attract as many people as possible with the basic functions. After a period, some users want more and start to buy additional features. The free version is a fundamental investment, in which it is necessary to invest a part of (future) income from a paid one, it also needs to be promoted and maintained in the current state. A free application advertises the user the idea of ??how it will be good with a paid version.

1.3.4 With-in app sales of virtual products

If a company uses this model, the application is a channel for sales or online storefronts. Owner of the app can sell virtual or real products or services, or even access to content or new functions inside the application. Since the introduction of the model of purchases within the application in 2009, more companies are using the approach - this model of monetization accounts for more than 50% of mobile revenues for 2017. (AppTractor, 2016)

1.4 Vertical differentiation

One of the most popular marketing strategy is differentiation. Term «differentiation» was proposed for the first time by Edward Chamberlin in 1933 (Chamberlin, 1933). Vertical differentiation of products distributes products in the industry market in accordance with their quality. Products in this situation satisfy the same need, but they are intended for consumers with different incomes. Proposal of vertically differentiated products attract people with different income levels and can make a market niche.

Product differentiation has been extensively studied by the example of FMCG, automobiles and computers by many scientists (Dennis Z.Y., 2012, pp. 314-328) (Moorthy K.S. ,Png I.P., 1992) (Moorthy K.S., 1984).

Vertical differentiation in recent years dramatically changed air travelling industry. Low-cost flies became extremely popular, and account for a large share of Western civil aviation, particularly in Europe. Carriers offer flights of different quality and can sign agreements with suppliers of goods and services at the destination. Customers take care of home destination and the quality of the flight. The thing is, that only low-income travelers fly with low-cost airlines, while no-frills carriers are more likely to act as a platform than older airlines (Serio, L.; Tedeschi, P.; Ursino, G., 2018).

For example, Dennis (Dennis Z.Y., 2012, pp. 314-328) stated that the optimal number of product variants of a firm is determined by a parameter that depends on the size of the market and the costs associated with the operation, such as the cost of installing and storing stocks. This parameter can be interpreted as the relative cost of production technology. In the other words, less expensive manufacturing technology provides a solid opportunity to produce more product variants. More expensive production technologies lead to less variety of products, lower product quality and less market coverage (ie, to the percentage of the covered market). The difference is that we deal with IT sphere, where manufacturing costs is very low, literally it comes down to the price of a couple of hours of development. Therefore, we can assume that for IT sphere function will be different in comparison with automobile or FMCG industry. In this case, consumers' demand and purchasing power is more important, because to add some products to a product line is very simple.

Moorthy (1984) and Moorthy and Png (1992) concluded that the product should be of poor quality for consumers with low income and excellent quality for consumers with high scores. It can be obvious for someone, however it was revealed that perceived confidence had a stronger impact than the estimated price of the intention to purchase both potential and regular customers of the online store (Hee-Woong K. Yunjie X. Sumeet G., 2012). The interesting thing is that the estimated price had a stronger impact on the decision to buy repeat customers compared to potential customers. Perceived trust has had a stronger impact on the decision to buy potential customers compared to repeat customers.

Certainly, increasing company's profit through product differentiation, offering products with different prices and quality for groups with different income is good for company. However, Zanchettin and Mukherjee (Zanchettin, P.; Mukherjee, A., 2017) state that differentiation reduces social welfare. They show that this social cost of vertical integration is most likely to arise in innovative and competitive industries, and that competition for an upstream mining upstream integration channel is reliable for competition in production.

1.5 Previous studies

While analyzing sources and previous studies we faced the problem of few number of similar topic studies. It happens due to the very applied type of research. The main part of the articles about mobile app monetization is devoted to the analysis of existing ways to monetize a mobile application. Almost all authors write about the same ways, some authors talk about a number of basic mistakes that can be encountered in developing the path. We could not find any disputes between the authors, since almost the same paths are described, just the authors consider them in varying degrees of detail. There are some statistics about the most popular types of monetization of mobile applications - it is a free application with advertising. Nevertheless, the most promising is freemium as Michelle Delgado figured out in her research (Delgado, 2017). Often developers do not know how to properly present and promote their offspring, because it is not enough just to develop an application, we need to release it to the market in such a way that the largest number of people know about it and want to download it. Of course, the ultimate goal is to get profit from the project, to earn enough money to return the money invested and efforts, and further development of the product or other business. monetization differentiation revenue

We managed to find statistics, which indicates a significant growth in each category of mobile applications. Researchers conclude that mobile applications from marketing communications tools themselves become media distribution channels. In the US, the use of mobile applications already exceeds all other channels of media consumption - it is 82% of the total time (Forbes, 2015).



Research problem, methodology and design

2.1 Formulation of the research problem

The development of the mobile application market is facilitated, firstly, by the continuous growth of mobile devices' performance, and, secondly, by the growing availability of the Internet, including mobile ones, to increase the speed and stability of connections. This is critical, since most mobile applications require constant connection to the network. Most of the income of mobile application stores is still in games. The second direction, which actively developed, is shopping. The time spent by users in this type of app has increased by 30% over the year, first of all, at the expense of the USA. In the world market of mobile applications, one can note the development of countries that were formerly outsiders of the market: India, Indonesia, Brazil and Mexico. In 2016, China ranked first in revenue in mobile application stores. Thanks to the development of Internet networks and the penetration of the Internet, the popularity of streaming video applications has increased significantly; Here the growth of incomes was observed both due to the increase in the number of users, and due to their involvement.

The greatest amount of time, however, users spent in social applications and instant messengers. Nevertheless, other categories of applications within the world market were actively growing. World market research shows that about half of mobile device owners download and install additional applications (not pre-installed on the device), two-thirds of this amount is used regularly by applications. Conducted segmentation of the market showed that the majority of mobile application users are men and women aged 25-30 years, married or married, have higher education.

In 2015, 50 million of the 82 million Internet users in Russia used mobile devices to access the network. Visiting the Internet from stationary computers and laptops with it steadily decreases. For 10% of the country's population in general or for 14% of the total number of Internet users, mobile Internet is the only way to access the network, that is, stationary devices are not used. At the end of 2016, the number of downloads of mobile applications grew by 15% compared to 2015. At the same time, users began to spend 25% more time in applications. This increase in popularity brought publishers 40% more revenue compared to last year.

The problem is that people in Russia do not pay for intangible products like movies, music, apps and different products in apps. That is why premium models, when users pay for the use of service are working poorly. There remains a model in which publishers can earn by displaying ads inside applications. However, advertising fees from applications in developed countries are higher (Lee I., 2017).

2.2 Methodology and design of research

On the basis of described theoretical aspect of monetization of mobile application and product variety and vertical differentiation, we can suppose that vertical differentiation can be useful for IT-companies which wants to increase its revenue, because it helped companies in automobile, FMCG and hardware industries as it was described in the theoretical part of paper. It is a reason why we think the hypothesis is relevant and should be tested.

On the basis of described theoretical aspect and our hypothesis, we suppose, that the most relevant research method in this case is a field experiment because it will give us a possibility to compare different metrics in experimental and control. To figure out if vertical differentiation is a good way to increase mobile app's revenue, we should conduct an experiment. After that, data will be collected, analyzed and we will reject or approve the hypothesis. Methods of the research are field experiment and statistical analysis. In details, to reach our main goal we should complete some objects:

- to make targeting while registration of new users in mobile application to label experimental and control groups;

- to run the experiment;

- to monitor if data is collecting properly during the experiment;

- to collect the data;

- to check metrics of interest;

- to check if collected data is enough;

- to collected necessary amount of data;

- to stop the experiment;

- to get all the data;

- to clean the data;

- to make necessary statistical tests, descriptive statistics;

- to make a conclusion on the basis on driven analysis.

We have conducted a field experiment to figure out if our hypothesis is true or not. During the experiment, new users were randomly divided into 3 groups: experimental group, control group №1, control group №2. The experimental group had a list of 4 products (subscription for a week, subscription for a month, subscription for 3 month and subscription for a year) and the two control groups had only one option - a monthly subscription.

Collected data are payments of participants of the experiment during the period of almost 2 month. These payments were grouped by users, so we could compare how much money people spent when they had different products and when they did not.

Design of this research is case study. We would describe our attempt to test a hypothesis that vertical differentiation can serve as a tool to increase the number of purchases and total revenue from sales of virtual products within the mobile application on the example of dating mobile application called “Topface” and people from Russian Federation who use Android-devices.

The experiment lasted for 2 months. It was about 13 000 of mobile users in each group. They are Android users from Russian Federation. Samples are independent because after the registration each user had got a target mark with of specific group. During the statistical analysis, data distribution will be checked using Shapiro-Wilk and Kolmogorov-Smirnov statistical criteria. The difference of sample medians between Eperimental group and Control Groups will be checked with Kruskal-Wallis H-test and Mann-Whitney rank test.

Due to the methods that were used and the design of research, resuts this paper have some limitations. Collected data is of high quality and gives us possibility to make good statistical and econometric analysis. Nevertheless, we made an experiment only in one mobile application which is from dating segment, therefore it is case-study research. Also, this experiment was conducted only for Android-OS users in Russia. Therefore, we cannot extrapolate our results to the whole general population - all mobile applications.

Data was collected using Kibana service, based on SQL queries. Files with payments are in csv format. They were analysed using Python programming language, with Jupyter Notebook programm. The link to git.hub with the whole script is represented in the attachements. Necessary scills of programing on Python to execute the analysis were got on DataCamp web site.

2.3 Company's description

Topface is an international dating service, one of the most popular dating services in Russia. 1.6 million of visitors every day, more than 100 million participants in the database, only half of them are from Russia. The service is available through the website, social networking applications VKontakte, Facebook, Odnoklassniki, as well as on Android and iOS devices.

The form of the organization is a limited liability company. The company employs about 60 people: large IT-department (developers, testers, devops, analysts), hr, financial and legal departments, top-management and event-managers. Topface has a strong corporate culture that provides employees with high quality of labour conditions, good compensations, free medical insurance and very comfortable and cosy atmosphere in the whole organization. Management of Topface company does care about his employees, because it is the main asset of the IT-organization. Therefore it regularly conducts staff interviews on job satisfaction, organizes visits to seminars on different topics in Russia and outdoor. Developers, testers and product-managers often participate in paid and free professional webinars.

Picture №2 - Topface company advertisement



Results of the experiment

3.1 Changes in the user's behavior

Difference in revenue between groups (Experimental Group, Control Group №1 Control Group №2) in general: 17.65% (311016 275572 253112).

Difference in revenue between groups (Experimental Group, Control Group №1 Control Group №2) for organic users: 5.161% (159814 151011 152929).

Difference in revenue between groups (Experimental Group, Control Group №1 Control Group №2) for referral users: 34.55% (151202 124561 100183).

Difference in revenue per user between groups (Experimental Group, Control Group №1 Control Group №2) in general:

11.64% (711.7070938215103 680.0595744680851 601.2161520190024).

Difference in revenue per user between groups (Experimental Group, Control Group №1 Control Group №2) for organic users:

9.990% (680.0595744680851 634.5 602.0826771653543).

Difference in revenue per user between groups (Experimental Group, Control Group №1 Control Group №2) for referral users: 12.70% (748.5247524752475 728.4269005847954 599.8982035928144).

Difference in revenue for premium account subscribe between groups (Experimental Group, Control Group №1 Control Group №2) in general: 30.36% (210669 177247 145956).

Difference in revenue for premium account subscribe between groups (Experimental Group, Control Group №1 Control Group №2) for organic users: 19.52% (102558 93030 78576).

Difference in revenue for premium account subscribe between groups (Experimental Group, Control Group №1 Control Group №2) for referral users: 42.62% (108111 84217 67380).

Difference in amount of purchases for premium account subscribe groups (Experimental Group, Control Group №1 Control Group №2) in general: 34.75% (473 344 358).

Difference in amount of purchases for premium account subscribe groups (Experimental Group, Control Group №1 Control Group №2) for organic users: 28.85% (259 189 213).

Difference in amount of purchases for premium account subscribe between groups (Experimental Group, Control Group №1 Control Group №2) for referral users: 42.66% (214 155 145).

Difference in revenue for premium account subscribe with trial period only between groups (Experimental Group, Control Group №1 Control Group №2) in general: 10.59% (64543 64543 58037).

Difference in revenue for premium account subscribe with trial period only between groups (Experimental Group, Control Group №1 Control Group №2) for organic users: 4.044% (28758 28761 26519).

Difference in revenue for premium account subscribe with trial period only between groups (Experimental Group, Control Group №1 Control Group №2) for referral users: 15.98% (39028 35782 31518).

Difference in amount of purchases for premium account subscribe with trial period only between groups (Experimental Group, Control Group №1 Control Group №2) in general: -9.60% (80 91 86).

Difference in amount of purchases for premium account subscribe with trial period only between groups (Experimental Group, Control Group №1 Control Group №2) for organic users: -12.1% (36 42 40).

Difference in amount of purchases for premium account subscribe with trial period only between groups (Experimental Group, Control Group №1 Control Group №2) for referral users: -7.36% (44 49 46).

3.2 Data description

The experiment was held only for users from Russia who use Android devices. Therefore, collected data contains payments in “Topface” mobile application of users of Android devices from Russia who participated in the experiment. Also, we have drown payments of products which strictly contain strings “prem” (premium), “vip” (VIP) or “trial” (trial period for subscription) to filter necessary purchases for velid analysis. Then we recode labels of groups to numbers: “tg” to 0 (Experimental or Test group), “cg1” to 1(Control Group1) and “cg2” to 2 (Control Group2).

Picture №3 - Filtering the data using Python programming language.

Here is a chunk of data collected that was analyzed. You can see values of columns "payment ID", "payment date", "payment amount", "revenue in cents", "payment service ID", "payment provider", platform, place, country, "last app", "Product Id". There are columns:

a) “ref ID” for the ID that was given by referral agent,

b) “partner ID" for the ID of referral agent,

c) "ref date" for the date when client was involved,

d) "user ID" for the user ID,

e) "registration date" for the registration date,

f) "payment ID" for the ID of specific type of product,

g) "payment date" for the payment date,

h) "payment amount" for the payment date,

i) "revenue in cents" for the revenue in cents,

j) "payment service ID" for the payment service ID,

k) "payment provider" for the payment provider like “Money from GooglePlay”,

l) “platform” for the platform which was used for registration,

m) “place” for the specific place in the application from which the transition to purchase occurred,

n) “Country” for the country of user,

o) "last app" for the last app of user,

p) "Product Id" for the product ID.

Each line is for each purchase that was made by people involved in the experiment, who use Android devices in Russia. Grouping them for analyzing will help us to discover some links and differences between experimental (test) and control groups.

Picture №4 - Descriptive statistics in Jupyter Notebook

There is a descriptive statistics of distribution of revenue in cents per each user here. We can see that there are 1010 unique participants in the experiment. Mean value is 528,6 cents (5,28$), minimum value is 81,0 cents(0,81$) and maximum value is 9244 cents (92,44$). Standard deviation is 658,10 cents. The first quartile is 271,25 cents, the second quartile is 363,5 cents and the third quartile is 606,0 cents.

Picture №5 - Histogram of revenue distribution per user in Jupyter Notebook

Here you can see a histogram that shows graphically how our data is distributed. It does not look like normal Poisson distribution (Graph №5). That is the reason we will test the null hypothesis that our data is drown from normal distribution.

Picture №6 - Normal Poisson distribution

Therefore, we will test the null hypothesis that our data is drown from normal distribution.

3.3 Checking data distribution using Shapiro-Wilk and Kolmogorov-Smirnov statistical criteria

The Shapiro-Wilk criterion is used to test the hypothesis H0: "the random variable X is normally distributed" and is one of the most effective criteria for verifying normality. Criteria that verify the normality of the sample are a special case of the criteria for agreement.

Picture №7 - Results of Shapiro-Wilk test

Function scipy.stats.shapiro() in python returns:

a) W : float. The test statistic;

b) p-value: float. The p-value for the hypothesis test.

It means the probability that we will make a mistake if will reject the null hypothesis is 0.0%, that means that our data is not distributed normally due to Shapiro-Wilk criterion.

Another way to check if our data is distributed normally is Kolmagorov-Smirnov criterion. The Kolmogorov-Smirnov criterion is used to test the hypothesis H0: "the random variable X has the distribution F (x)".

Picture №8 - Results of Kolmagorov-Smirnov test

scipy.stats.kstest() returns:

a) D : float. KS test statistic, either D, D+ or D-.

b) p-value : float. One-tailed or two-tailed p-value.

So the probability that we will make a mistake if will reject the null hypothesis is 0.0%, that means that our data is not distributed normally due to Kolmagorov-Smirnov criterion.

3.4 Checking the equality of samples mean and median using the Kruskal-Wallis and Mann-Whitney non-parametric tests

Box-plot graph shows us the difference between medians and means in Experimental(0), Control №1(1) and Control№2(2) groups. It is not obvious if there is a significant difference between samples mean and median, so we will implement statistical tests to discover it.

Picture №9 - Box-plot of groups in experiment

Since data is not distributed normally, nonparametric tests are used. A non-parametric version of ANOVA is the Kruskal-Wallis H-test. The Kruskal-Wallis H-test tests the null hypothesis that the population median of all of the groups are equal.

scipy.stats.kruskal() returns:

a) statistic : float. The Kruskal-Wallis H statistic, corrected for ties

b) pvalue : float. The p-value for the test using the assumption that H has a chi square distribution.

Picture №10 - Results of Kruskal test

Here is the result of Kruskal-Wallis H-test for Experimental group, Control Group №1 and Control Group №2. The difference of sample medians between Eperimental group and Control Groups is significant because p-value of Kruskal-Wallis H-test is 0.0046.

Picture №11 - Results of Kruskal test (2)

You can see that of the difference of sample medians between Eperimental group and merged Control Groups is significant because p-value of Kruskal-Wallis H-test is 0.0020 .

One of the most popular nonparametric tests for checking the equality of sample means is Mann-Whitney rank test. scipy.stats.mannwhitneyu() returns:

statistic : float. The Mann-Whitney U statistic, equal to min(U for x, U for y) if alternative is equal to None (deprecated; exists for backward compatibility), and U for y otherwise.

pvalue : float. P-value assuming an asymptotic normal distribution. One-sided or two-sided, depending on the choice of alternative.



Picture №12 - Results of Mann-Whitney test

Here you can see that the difference of sample means between Eperimental group and Control Group №1 is significant because p-value of Mann-Whitney rank test is 0.0162. The difference of sample means between Eperimental group and Control Group №2 is significant because p-value of Mann-Whitney rank test is 0.0008. The difference of sample means between Eperimental group and merged Control Groups is significant because p-value of Mann-Whitney rank test is 0.0010.

Based on the above, we can argue that there is a significant difference between the sample means of the experimental and control groups and our hypothesis that vertical differentiation can serve as a way to increase the revenues of the mobile application from the sale of virtual products in the mobile application is confirmed.

Limitations and future research avenues

Experiment can be repeated for users from other countries and of different applications, for different types of products (tangible and intangible, virtual and real). This can be a future work, continuation of the research started in this particular paper. Also, design of research can be switched to cross-sectional, to prove that vertical differentiation of products in mobile application is a good marketing strategy not only for apps from Dating segment, but also from others segments of mobile applications with Freemium model like education, sport, business applications, etc.



Conclusion

In conclusion, we will evaluate value of made research and speak about potential future work. We made an experiment to test our hypothesis. We have figured out that product variety and vertical differentiation can be a tool for revenue increase of mobile application from with-in app sales of virtual products. Conducting this experiment led to introducing 4 products for new Russian users and increase of real revenue without any increase of costs. We can say that implementing this strategy in IT sphere, where production of new products costs very few is very powerful and useful technique. Applications' owners can differentiate their products without any investments and improve their financial indicators: profitability index, return on investments, turnover ratio and so on. Therefore, we can say that this thesis enriches the literature of this field of science (Marketing in IT) with a new interesting case-study. It can be repeated for users from other countries and of different applications, for different types of products (tangible and intangible, virtual and real). This can be a future work, continuation of the research started in this particular paper. This finding enriches the existing literature by providing with an understanding of how to increase revenue from mobile app. Big amount of articles describe different monetization models and analyze which of them is better, but very few say how to improve the existing model. We have described a case of mobile app with the most recommended monetization model - Freemium, and our insights are illustrated through a numerical example. The experiment can be repeated. Furthermore, we suggest to test our hypothesis not only for Android users from Russia, but also iOS users and people from different countries.

We hope that our research will inspire other scientists to enrich this field of marketing with more researches and new insights. IT is very fast changing, profitable and promising sphere that share with itы models of project management with other industries. Therefore, we think it would be logical to bring something new in the marketing sphere from IT.

References

1. ICT Facts and Figures 2014 [Online] // International Telecommunication Union. - 2014. - 5 20, 2018. - http://www.itu.int/en/ITU-D/Statistics/Pages/stat/default.aspx. .

2. ""App" voted 2010 word of the year by the American Dialect Society American Dialect Society". [Online] // Americandialect.org. . - 8 1, 2011. - 4 2, 2018.

3. App Development Cost Survey [Online] // Clutch. - 2017.

4. AppAnnie [Online] // AppAnnie. - 2016. - May 2018. - https://www.appannie.com/ru/.

5. AppTractor Best practice of mobile app monetization [Online]. - 2016. - 1 16, 2018. - http://apptractor.ru/marketing-monetization.

6. Aydin G., Karamehmet B. A comparative study on attitudes towards SMS [Article] // Int. J. Mobile Communications. - 2017. - 5 : Vol. 15. - p. 514.

7. Burdge, B. New research shows mobile dominates desktops [Online] // MovableInk. - 2014. - 5 2, 2018. - http://blog.movableink.com/new-research-shows-mobile-dominates-desktops-with-65-of-totalemail-opens-in-q4-2013/.

8. Chamberlin E. Theory of monopolistic competition [Book]. - 1933.

9. ConsultancyManagement Mobile app market in Russia and worldwide [Online]. - 3 23, 2014. - 4 2018. - http://www.json.ru/poleznye_materialy/free_market_watches/analytics/.

10. Cossa Research of mobile app promotion market [Online]. - 2 23, 2015. - 4 2018. - http://www.cossa.ru/155/72460/ .

11. Delgado M. Launch, How to Make an App: From Brainstorm to Launch [Journal] // Clutch. - 10 10, 2017.

12. Dennis Z.Y. Product variety and vertical differentiation in a batch production system [Journal] // International Journal of Production Economics. - 2012, pp. 314-328. - 138 (2). - pp. 314-328.

13. eMarketer Worldwide Smartphone Usage to Grow 45% in 2014 [Online] // eMarketer. - 2014. - May 2018. - http://www.emarketer.com/Article/Worldwide-Smartphone-Usage-Grow-25-2014/1010920.

14. Forbes How Much Do Average Apps Make? [Online] // Forbes . - 8 10, 2015. - 4 15, 2018. - http://www.forbes.com/sites/tristanlouis/2013/08/10/how-much-do-average-apps-make/#4dd7e27212cb.

15. Hee-Woong K. Yunjie X. Sumeet G. Which is more important in Internet shopping, perceived price or trust? [Journal] // Electronic Commerce Research and Applications. - 2012. - 11 : Vol. 3. - pp. 241-252.

16. Kim, M., Kim, J., Choi, J., Trivedi, M. Mobile Shopping Through Applications: Understanding Application Possession and Mobile Purchase [Article] // JOURNAL OF INTERACTIVE MARKETING. - 2017. - 9 : Vol. 39. - pp. 55-68.

17. Lee I. Russia entered the top 5 in terms of downloads of mobile applications in the world [Journal] // RBK. - 2017.

18. Moorthy K.S. ,Png I.P. Market segmentation, cannibalization, and the timing of product introductions [Journal] // Management Science. - 1992. - 38 (3). - pp. 345-359.

19. Moorthy K.S. Market segmentation, self-selection, and product line design [Journal] // Marketing Science. - 1984. - 3 (4) . - pp. 288-307.

20. Musa, R. ; Saidon, J.; Harun, M.H. ; Adam, A.A. ; Dzahar, D.F. ; Haussain, S.S. ; Lokman, W.M. The Predictors and Consequences of Consumers' Attitude Towards Mobile Shopping Application [Article] // Fifth International Conference on Marketing and Retailing. - 2015. - Vol. 37. - pp. 447-452.

21. scipy.org scipy.stats.kruskal [Online] // scipy.org . - 2018. - 4 20, 2018. - https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.kruskal.html .

...