The business of teaching languages through apps: Machine learning driving competitive advantage in the mobile-assisted language learning industry

The business of teaching languages through apps:

Machine learning driving competitive advantage in the

mobile-assisted language learning industry

Author: Luca De Bortoli Student number: S4104749 Email: l.de.bortoli.1@student.rug.nl

Supervisor: P.J. Marques Morgado Co-assessor: H. Reijn

Faculty of Economics and Business University of Groningen

Duisenberg Building, Nettelbosje 2, 9747 AE Groningen, The Netherlands P.O. Box 800, 9700 AV Groningen, The Netherlands

http://www.rug.nl/feb

TABLE OF CONTENT

Competitive advantage in the MALL industry ... 5

Effectiveness of the method ... 6

Gamification features ... 6

Main players in the industry ... 6

Rosetta Stone ... 7

Duolingo ... 7

Product management in the MALL industry ... 9

Machine learning in the MALL industry ... 10

Theory ... 13

Sustained competitive advantage ... 13

ABSTRACT

Mastering a foreign language has been reputed a desirable skill all throughout history, but it has never been as important as it is today. The globalized world we live in and the rise of new media and technology has made it clear that communication with other people is essential. More and more people have been interested in learning a new language and recent software firms have decided to turn this interest into a business by creating language courses accessible through mobile devices. This newly born industry is known in the literature as the MALL (mobile-assisted language learning) industry. While researchers in psycholinguistics and pedagogy have long been debating about the efficiency of digital learning methods and tools, the business research environment has so far been silent in front of this rising phenomenon.

The main question that will be addressed throughout this paper is whether the introduction of machine learning in product management processes leads the firm to a competitive advantage. Moreover, the paper questions what the sources of competitive advantage in the MALL industry are by focusing on two market leaders representing different strategies, Duolingo and Rosetta Stone. The research-based view and the stakeholder theory will be the theoretical aids that are going to help channel non-business-related literature towards a business reasoning.

INTRODUCTION

“The limits of my language are the limits of my world” - Ludwig Wittgenstein (1889-1951)

Recent technological advances have showed us that the world can be a small space, so small we can hold it in one hand. Smartphones have changed the way we look at life. A rise in interconnectivity and the possibility to reach each other digitally without constraints of time and space has led us to newer needs for communication.

In a virtually limitless world, foreign languages seem to be the last border that separates people living in different realities. Some innovative firms have seen a breach in this border and have been able to combine new technological frontiers with the needs of the modern person.

Mobile-assisted language learning (MALL) or m-learning is the act of learning languages through the assistance of mobile devices (Lu, 2008; McCarty, Sato, & Obari, 2017). Such tools are practical, flexible, mobile, individually customizable and always available, qualities which sound more appealing than a traditional, pricey language course in a classroom.

McCarthy et al. (2017) stress the relevance of the industry. In the current environment, the advantages of the portability of such tools, combined with the relatively low prices of devices, has made MALL more attractive as compared to the usual, classroom setting. New companies were born to satisfy the needs of a new kind of learners that is attracted by the possibility of learning “anywhere, at any time”, by oneself and according to one’s mobility and needs (Bidin & Ziden, 2013; Böhm & Constantine, 2016; Kukulska-Hulme & Shield, 2008; Milutinović, Labus, Stojiljković, Bogdanović, & Despotović-Zrakić, 2013).

The size of the MALL industry does catch one’s attention. Adams (2019) reported that the online language-learning industry, with a focus on mobile apps, is generating $6 billion in revenue, a figure that is expected to rise to $8.7 billion over the next five years. In 2016, almost 10% of the over a million apps listed on online stores were educational: 1000-2000 of those are focused on language learning; the compound annual growth rate of the industry over five years is 11.1% (Heil, Wu, Lee, & Schmidt, 2016).

originate from and what is the role played by product management in achieving a superior market position.

The analysis of the literature and of a market leader’s annual reports (Rosetta Stone, Inc. Years 2009-2019) suggest that the industry is currently dominated by companies that are not explicitly profit driven. The main goal of big players in the MALL industry is, according to the literature, user acquisition. The companies that hold a superior position in the market are the ones which manage to acquire more users. A bigger pool of users usually translates into higher revenues originating from subscription plans, advertising or even private sponsoring.

The extant literature suggests that an effective product and gamification features added to the courses tend to attract more users, therefore the research will investigate these factors. These assumptions will be evaluated in combination with commonly known business theories: the resource- and stakeholder-based views.

Machine learning is a branch of AI (Artificial Intelligence) which refers to the ability of a software or system to improve its performance over a period of time (Saloky & Šeminský, 2019). The machine, by acquiring data through the interaction with its users, is able to provide specific data in return, coherent with the inputs provided (Byron, 2020). These premises show that the use of machine learning may have an impact on the success of a product: could this technology help firms acquire a competitive advantage?

The activity that this paper will investigate is product management. It consists of a set of processes that govern a product from its origins to its delivery to the market to generate the most significant value possible (Ebert, 2007). It was deemed relevant because choosing a broad, nuanced spectrum seemed more fitting in the analysis of an industry whose borders are rather blurred, given its recent establishment.

Within product management, the AI department plays a major role in the scope of this paper: it is in charge of machine learning. However, machine learning alone does not lead to a competitive advantage, supposedly. To ensure effectiveness and enjoyment in the learning process, product management has to collaborate with other departments that filter internal and external inputs – such as demand or feedback (Ebert & Brinkkemper, 2014).

Question:

How does machine learning contribute to product management in the achievement of competitive advantage in the MALL industry?

LITERATURE REVIEW

Competitive advantage in the MALL industry

This section aims at outlining the success factors of two big players in the MALL industry to identify their sources of competitive advantage.

In their 2019 annual report, Rosetta Stone admitted not being able to keep up with the competition. The industry easily accepts new entries, which can offer courses at lower prices and that attract bigger pools of customers. This last characteristic is highlighted in the report, which suggests that firms attracting more users will eventually gain a superior position in the market.

The non-business-related literature about firms in the MALL industry has been focusing mainly on two aspects: effectiveness of the method and presence of gamification features. Given the interest dedicated by researchers and the influence of their reviews on the common consumer, it is likely that firms will aim at improving such cores of their products.

Based on these assumptions, it is hypothesized that firms involved in the industry gain competitive advantage if their methods are deemed superior in effectiveness and, overall, fun and easy to use. Both features are likely to attract more users.

Most of the firms in the industry rely on profit made through subscription to their services or a “pay to unlock” feature, where the customer pays a flat fee to access more courses or levels. Another common monetization system is in-app advertising, which can be stopped through payment (Heil et al., 2016).

Product management has been chosen as an activity worth investigating because of its intrinsic flexibility: a product manager monitors the whole development of the product and is therefore the one who takes decisions, basing them on directives given by the different departments within the firm and by external stakeholders. Special attention will be dedicated to the AI-department within the product management processes.

Effectiveness of the method

The first assumed source of competitive advantage that will be discussed is the effectiveness of the method, namely the perceived improvement of a user in their target language after using a digital course.

A vast body of literature, mostly found in the fields of psycholinguistics and pedagogy, has been produced on the matter ever since the first appearance of MALL methods. Companies often fund research in order to scientifically support their claim of being superior as compared to their competition (duolingo.com; babbel.com; rosettastone.co.uk). A proven effectiveness is expected to attract more users and justify the firm’s existence.

According to Böhm (2016), students have shown more interest in using language learning apps when they identified a real advantage in using them to improve their language skills.

Lu (2008), at the early stages of the development of the MALL industry, has identified that young learners tend to appreciate the use of language learning software on their devices as they admit recalling vocabulary easily. The author supported the theory that mobile learning offers real learning gains and the characteristics of portability, mobility and immediacy typical of modern mobile devices particularly fit this kind of learners.

Gamification features

It is hypothesized that the introduction of gamification features in a language learning app or software is source of competitive advantage.

Gamification means that an app or software proposes learning under the use of game elements to create enjoyment (Huynh, Zuo, & IIda, 2016).

A strong component that gamification adds to the product is addictiveness: the user feels challenged, motivated and eager to achieve learning goals because the system will award him with advantages or in-game currency (Godwin-Jones, 2017). Rewards or penalties are supposed to enhance a user’s motivation and connect it more personally with the use of a specific app. (Milutinović et al., 2013)

Main players in the industry

Rosetta Stone

One of the most popular players is Rosetta Stone, an American company with more than 25 years of experience in the industry. The software has been made popular by its partnerships with companies, institutions and universities; notable examples are NASA, Calvin Klein, Fender among others (rosettastone.co.uk)

Users of its apps and websites can progressively learn one language through translation, repetition of vocabulary and grammar structures. Personal attention is dedicated to the user, who is invited to repeat the words and structures that the software perceives as weaker.

The revenue of Rosetta Stone in 2019 was $163m, with an 11% growth as compared to the previous year. However, the company experienced a first drop in revenues in 2013 (-7.8%), only in 2019 the decreasing trend has been reversed (orbis database, bvdinfo.com).

The prices of Rosetta Stone’s course are the highest on the market, which make it considerably less accessible than its main competitors (Nushi & Eqbali, 2018). Rosetta Stone makes profit through a licensing fee that can be valid up to 24 months. A lifetime subscription is also available. Prices vary from $180 to $300. Rosetta Stone offers live tutoring as well, that can be purchased in-app after signing up to the course (rosettastone.co.uk).

Duolingo

The US company Duolingo became a hit in 2013, when Apple named it App of the year thanks to its rising popularity (Munday, 2017). At its beginnings in 2011, venture capital was betting on its success with an approximate investment of $15 million (Garcia, 2013). The software now offers 94 language courses in 23 languages and counts more than 150 million users (2019).

Unfortunately, no certain financial information is available, considering that the company is private. Adams (2019) estimates a $36 million in revenues in 2018, a figure that has apparently been growing ever since the company’s appearance on the market. It is worth noticing that Rosetta Stone has started seeing a drop in its revenue around 2013, i.e. approximately when Duolingo became relevant competition.

Duolingo’s net worth is estimated at $700 million (Adams, 2019). According to the company’s CEO, the revenue in 2019 was expected to hit $86 million; in 2016, the company reported a revenue of $1 million. In 2020, also given the rise in users triggered by the covid-19 pandemic, the revenue will be well beyond $140 million (Adams, 2020).

to the advancement in the learning game, is the main source of revenue for the company, although only 3% of the users choose to subscribe (Adams, 2020). Another source of profit is derived by companies and organizations that send material to Duolingo which has it translated by users of the app (duolingo.com).

The reasons for its success appear to be various: the official site of the company explicitly states that the learning experience is tailored on the user, who approaches the language through learning by playing, namely through game-like features and rewards. Duolingo works on a translation-based system, whereby the learner acquires a skill through the constant translation of short sentences from the target language to the native one and vice versa. The user unlocks new vocabulary and grammar upon completing preceding activities; this pre-determined path is called ‘tree’ in the platform (Munday, 2017).

Over the years, studies have proven its effectiveness and this boosted the reputation of the company further (Vesselinov & Grego, 2012).

Arguably, though, the fact that it is completely free contributed to its massive success. The company’s mission, according to its CEO Luis von Ahn, is allowing people to upgrade in life thanks to the acquisition of a new skill: foreign language abilities. The app was made available for free so that everyone could have the possibility of learning a new language.

Duolingo has gained wide governmental support, especially in Latin America (Garcia Botero, Garcia Botero, & Questier, 2017), where the project Duolingo for schools has been providing high quality language education in difficult boroughs. These governments chose to partner with Duolingo also because of the cost advantage.

Amidst the covid-19 pandemic of 2020, Duolingo reported a 500% growth in users taking their digital English test because of testing centers closing. The company states that it is the first high-stakes test powered by machine learning. Taking the test currently costs $20, which is a price on average five times lower than any other English testing center. Duolingo poses a threat to such institutions (Adams, 2020).

Chart 1: Increase in new users of Duolingo during March 2020. Source: duolingo.com

Product management in the MALL industry

In a comprehensive definition, the role of product management is described as the long-term mapping and balancing needs of different stakeholders in order to maximize benefits (Ebert, 2007; Ebert & Brinkkemper, 2014; Karlsson, Dahlstedt, Dag, Regnell, & Persson, 2002)

Technology-based enterprises require their product managers to be particularly receptive and fast in adopting new strategies. IT technology has been developing at a uniquely swift rate, thus meaning that product responsible have to be aware of the latest developments and innovations and react with a constant adaptation of their products in order to survive and grow in this fast-moving environment (Roberts & Meyer, 1991). Roberts & Meyers (1991) stressed the importance of two factors:

1) the necessity to keep up with innovation by building a solid engineering base which constantly works on the improvement of the firm’s core technology;

2) the relevance of users in the development process. The product has to be adapted to the basic needs of its customers. Identifying groups of customers or stakeholders is therefore of key importance. The second point made by Roberts & Meyer constitutes a focal point in the activities of a product manager. The stakeholders involved in the e-learning industry are several: customers/users appear as the main stakeholders, however the influence of other has significant impact over product management. Customers will value a number of distinctive features in the product, such as usability, cost, technical requirements, flexibility, fun and, most importantly, learning outcomes (Bidin & Ziden, 2013).

It appears clear that these numerous requirements are ideally to be found in a successful product that a customer can enjoy, but on the path towards the development of a product, managers have to keep an eye open for the assessment of pedagogists, sociologists and IT-experts as well.

The product management is assumed to be in constant contact with the AI-department (present in all the MALL companies analyzed). The two sectors communicate feedbacks to each other and come up with the right strategy to implement to their product. The AI-department is responsible for the correct functioning of machine learning, creation of algorithms and the collection of the inputs. User feedbacks are presumably collected by other departments involved in user interaction, which then communicate with the product managers.

The AI-department constitutes, within product management, the most important sector in the achievement of an improved, user-friendlier software.

Flow chart 2: core elements of product management in the MALL industry

Machine learning in the MALL industry

With an increase in the use of language-learning IT tools, new horizons for the industry have been opened. Machine learning is thought to be a contributor not only to the success of individual firms, but to the advancement of the whole industry, as applied to the nuanced field of language learning. In fact, it entails the sub-skills of vocabulary building, grammar, morpho-syntactic knowledge and more (Settles, B. et al. 2020).

This unprecedented surge in the amount of data provided by active digital learners is expected to provide firms with better ways to serve their stakeholders. The user-tracking feature allowed by machine learning in language learning software has allowed researchers to investigate the pedagogical benefits of this kind of individual approach to learning. The outputs given by the software in response to the cumulative inputs provided by the user has yielded positive results in a pedagogical study by Collentine (2000).

Even though Collentine’s results refer to a technologically outdated environment, his results are relevant because they showed how machine learning had indeed an interesting potential in the improvement of such tools. In more recent years, machine learning has been identified as a founding premise for many current big players of the industry: the aim behind these platforms is help people learn a language while their inputs help the software improve (Garcia, 2013). Machine learning, in any case, provides the user with something that a classroom teacher cannot: an endless memory, a full record of the mistakes and a tailor-made program based on them (Heil et al., 2016).

In his paper, Garcia (2013) analyzed the early use of the machine learning software reCAPTCHA (CAPTCHA stands for Completely Automated Public Turing test to tell Computers and Humans Apart, an invention of Duolingo’s founder) made by Duolingo: the software types two words, one known to the computer and the other not. The computer assumes that if the user types in correctly the second word, then it did know the first one as well. The translation-based system adopted by most of the firms involved in MALL is constantly refined by the users’ inputs.

As mentioned above, learning a language is a complex skill which comprehends several sub-categories within the process. It is an individual process, and so is translation. For these reasons, the machine collects all the different inputs and responses to a single label and, among all the results, finds the perfect match, though allowing for variation. Duolingo, moreover, features the possibility to give direct feedback to a sentence in particular: this allows for further implementations by software developers who read their users’ comments (Garcia, 2013; Munday, 2017).

On another note, Duolingo has been able to acquire funding to improve machine translation as a whole: through the inputs given by the software users, computers try to refine their automatic translation features. The massive use of Duolingo, therefore, allows computers and humans to learn simultaneously from each other (Seave, 2016).

of points through the correct linking of a label to an image: this simple game, which is fun and fast for the users, helps the computer improve its image recognition capabilities.

Machine learning has been involved in the process of teaching oral skills, too. The literature mentions, in particular, the advanced speech recognition technology applied by Rosetta Stone (Garcia, 2013; Hönig, Batliner, & Nöth, 2012). The company’s proprietary speech recognition technology is constantly improved by user interaction: the software is reactive to a learner’s performance and provides them with personalized content. Inputs are memorized by the software which then adapts its offer to perfect the experience of every user. Recurring error patterns are recorded and transmitted to the server so that new methods can be implemented (Hönig et al., 2012).

It is interesting to notice how language learning apps and websites all appear to have a common goal: improving skills on both sides of the screen. Users interacting with the machine allow for a general improvement of different features. The data produced by user’s inputs then serve a purpose that goes beyond the borders of the firm.

Rosetta Stone’s speech recognition system has created a database of pronunciation errors made by pools of learners that could be studied in pedagogical research, as Hönig et al. (2012) suggest. Duolingo’s translation-based system has been heavily funded and put to use to improve machine translation software as a whole (Garcia, 2013). It is likely that Duolingo’s success and profitability is supported by external firms and stakeholders, such as Google, Central American governments and the European Union, which have already expressed interest in the firm’s activities (Garcia, 2013; Munday, 2017). The same goes for Rosetta Stone, which has been supported by popular firms and educational institutions alike (Hönig et al., 2012; rosettastone.com). It is striking how machine learning and human-machine interaction serve a higher social purpose when applied to the MALL industry. User inputs Machine learning Outputs Improved effectiveness Tailoring of the

method to the user

Secondary uses (research, machine

translation)

Flow chart 2: core elements of machine learning in the MALL industry

Theory

So far, the discussion around product management and machine learning has clarified that, in order to understand the creation of competitive advantage in this industry, it is important to analyze the impact of internal and external pressures on product managers. This has led to the necessity of building a theoretical background that encompasses both a resource-based view and a stakeholder-driven view.

Sustained competitive advantage

For starters, it is appropriate to clearly define what a competitive advantage is: according to Barney (1991) and Wu (2013), a firm has a competitive advantage when it is implementing a value creating strategy that is unique to the market, that no other firm is adhering to. This is likely to result in a higher financial performance and, consequently, a superior position in the market (Marques Morgado, 2014). A competitive advantage is said to be sustained if it resists to changes in the market, is independent to the time frame and it is such that no other firm is in the position to replicate it (Barney, 1991; Marques Morgado, 2014).

It is important mentioning that, also according to the dates of the literary findings concerning MALL, the industry is rather young and in current development. It is early and risky to infer sustained competitive advantage. This paper will focus more on the establishment of a competitive advantage, sustainability can be more safely assessed in future research.

Resource-based view

The analysis of the competitive possibilities of a firm, according to a resource-based view, should start internally, namely from the firm’s resources. Barney (1991) describes everything that a firm owns – tangible and intangible assets, capabilities, organizational culture – as resources. Notably, not all resources present in a firm are relevant when looking to implement a value-creating strategy that is unique to the enterprise, or, in other words, a competitive advantage.

For the purpose of this paper, the most appropriate resources to be taken into consideration are software and IT tools available to the firm.

When it comes to assessing the value of a firm’s resources, Barney (1995) presents a widely applicable framework: the VRIO framework. According to this theoretical instrument, a firm’s resource is said to be creating competitive advantage when it respects the following four attributes:

 it is Rare: it is obvious that a resource owned by several firms cannot be a source of

competitive advantage as described above, the firm’s resource-based strategy is not unique;  it is Imperfectly imitable, meaning that the firm’s resource-based strategy is not easily

imitable or applicable or both by competitors;

 Organization indicates that a firm will be able to create a competitive advantage if it is aware of its resources and capable of exploiting them.

The resource of a firm are managed by its personnel and organizational structure which are seen as complementary resources due to their inability to generate a competitive advantage by themselves (Barney, 1995). The VRIO framework fits well with the paper’s goal: machine learning and other internally developed technologies are sustainable, valuable resources only if supported by an attentive organization.

Marques Morgado (2014) stresses that any of these characteristics, if taken singularly, does not result in a competitive advantage; such condition is met only if all four apply contemporarily.

This research supposes that firms in the MALL industry should aim at achieving a superior user acquisition capability in order to obtain competitive advantage. The companies need therefore to be carefully evaluating their users’ wishes and feedbacks. For this reason, it is important to integrate the resource-based view with a stakeholder theory, which will provide a more complete view over the achievement of competitive advantage in the industry.

Stakeholder-based view

Donaldson and Preston (1995) summarize the most relevant pieces of stakeholder theory produced thus far and identify common assumptions that define the umbrella term ‘stakeholder theory’. Stakeholders are broadly defined as people or groups that express interest towards the corporate activity of a firm and thereby influence it. The interest of a specific group is of intrinsic value and needs to be regarded by the firm as unique, even though it might be connected to other interest groups. Examples of stakeholders are customers, suppliers, the government, activists.

Firm resources  Valuable?  Rare?  Inimitable?  Organization? Competitive advantage

Flow chart 4: the VRIO framework

As stated before, the main stakeholders in the industry are users, but they are not the only ones. The literature shows that researchers involved in the fields of psycholinguistics, language acquisition and education are particularly attentive to the developments of the MALL industry. It is not uncommon for firms in the industry to fund research projects aimed at attesting the efficiency and superiority of their methods.

Governments might also express interest in the activities of the firms taken into consideration, as was the case for Duolingo which received governmental support and praise in, for example, Ireland and Latin America (Garcia Botero et al., 2017; theindependent.ie).

The interests of the industry lie in these stakeholders because, supposedly, of gains in terms of reputation, which then reflects to the customers.

Firms react to stakeholders’ demand and build ethical relationships based on trust, loyalty, specific care that, in return, will result in higher user acquisition. The stakeholder is thus also instrumental: if managers are able to identify a stakeholder’s needs and react accordingly, both interests are fulfilled (Jones, Harrison, & Felps, 2018).

This is reflected by the managerial nature of the stakeholder theory: stakeholder management is a practice that requires managers to be fully aware of the environment. Organizations, in managing stakeholders, communicate, negotiate and motivate; the strategy of the enterprise holds together all these actions which give it a definition of what it stands for, with attention to ethics (Freeman, 2004). Jones et al. (2018) have also explored the possibility of generating a sustainable competitive advantage by paying closer attention to the instrumental stakeholder theory (IST). Given the solid theoretical groundings, the present paper will also elaborate on the instrumental role of stakeholders. Jones et al. (2018) provide another take on the matter: they have proven that a close relationship between firms and stakeholders, where both parties interact closely in order to obtain benefits, is value creating and likely to create a sustainable competitive advantage. The concept is shared by Marques Morgado (Marques Morgado, 2014): firms which engage in such an exchange can experience benefits which incrementally surpass the costs of the strategy implemented to create it. Because this kind of relationship is generally rare and difficult to imitate, the firm may generate a source of competitive advantage in it. This value is expected to be derived from improved coordination and knowledge sharing, among others.

(Wu, 2013). Roberts and Meyer (1991) also embrace the union of the two theories in a framework they developed for IT and software firms: a successful product is the result of an accurate planning, driven by user necessities.

User necessities Firm

Firm resources

Flow chart 5: the stakeholders, the resources and the firm

RESEARCH DESIGN

The methodology adopted in this paper draws inspiration from the guidelines proposed by Saunders (Saunders, Lewis, & Thornhill, 2009) in the research ‘onion’.

The philosophical background of this work is interpretivism. According to this approach social actors are responsible for the manifestation of a phenomena. The contemporary person wishing to engage in language learning has needs that the MALL industry was born to satisfy. Each actor has different views, and each view contributes with varying nuances. My goal was to capture the varying needs and nuances and describe how MALL firms react upon them.

Conducting an empirical research has not been possible; therefore, the paper follows an inductive approach. The screening and analysis of the non-business literature has led to hypothesis- and assumption-building in the business field. The literature was retrieved through archival research. Online databases, such as EBSCO, Google Scholar, Business Source Premier, Orbis and the online services of the library of the University of Groningen have provided all the papers that were consulted. The development of technology and the adaptation of existing software to improvements and feedback require a longitudinal approach, whereby the data collected is representative of a longer period of time rather than a limited one.

The paper is based on a literature review, a so-called mono-method. The stakeholder- and resource-based theories are common topics of the business and management literature. On the other hand, given the gaps in business research, literature belonging to the fields of computational linguistics, psycholinguistics, psychology, IT and artificial intelligence has helped shape the MALL industry and its characteristics. The aim was to extract a business-oriented picture by combining business- and non-business-related literature.

ANALYSIS

Discussion

After outlining what factors and theories led up to the research hypotheses, this section will be dedicated to finding answers to the questions.

Hypothesis 1: In the MALL-industry, a superior user acquisition capability leads to higher revenues and to competitive advantage.

Rosetta Stone’s annual reports and financial information – over a 10-year scope - provide this research with an essential insight on the MALL industry. A relevant, specific concern appears for the first time in 2010’s annual report: the company noticed a sudden increase in language courses offered in the form of mobile applications. While at first the threat was only recognized, following 2010, the annual reports became more specific in the description of such new applications, right as the promising growth rate of the past years was stopped and reversed. Rosetta Stone recognized that the threat of new entry in the industry has become very high as new companies with extremely competitive features appeared. These new firms, according to the annual reports, can produce methods of good quality at much lower costs and offer them at competitive prices or even for free. These firms are boosted by capital and their advantageous prices combined with proven effectiveness of the methods are capable of acquiring more users than Rosetta Stone (Rosetta Stone, Inc. Years 2009-2019). The ability to acquire more users seems to be the biggest threat posed to Rosetta Stone by the new entries. For Rosetta Stone, having more users translates directly to profit, considering their subscription-based plan. These customers are seemingly subtracted by the extremely competitive new entries. In 2013, Rosetta Stone lists in the annual report some of its main competitors, and Duolingo appears to be one of them.

Duolingo’s revenue largely depends on in-app advertisements and voluntary subscription plans that can eliminate the ads and provide the user with additional, marginal features. Not having the paid subscription does not hinder from using the app. Being able to attract a bigger, more diverse pool of users means for Duolingo that more companies will want to place their ads on the platform.

There is likely a correlation between Rosetta Stone’s dropping sales around 2013 and Duolingo’s rise in the same years. As Rosetta Stone’s growth rate kept recording a negative trend, Duolingo observed the opposite: a fast rise, given that estimates are correct. Duolingo and other new entries are likely to have subtracted users and profits from Rosetta Stone, which in 2019 reported revenues that were lower of more than $40 million if compared to 2010.

Rosetta Stone claimed, in several annual reports, not to be able to compete with companies like Duolingo.

It appears clear that Duolingo and other new entries have been able to acquire resources and use them in a more innovative way than Rosetta Stone, thereby guaranteeing them a top position in the market. Moreover, these new companies have been able to identify the needs of the contemporary user and satisfy them. The VRIO framework and the stakeholder theory are fully satisfied, and it is fair to assume that a superior user acquisition leads to higher profits and, consequently, to a competitive advantage.

Hypothesis 1a: Superior effectiveness of the method/product leads to the acquisition of more users.

The literature review suggests that a method which is proven to be more effective as compared to the competition is likely to attract more users and create a stronger market position and higher profits. In a study conducted by Vesselinov (2009), Rosetta Stone users have been enthusiastic in using the product, stating that they noticed a considerable improvement in their foreign language skills, enjoyed the course and were very likely to recommend it. The effectiveness of the course, the enjoyment and easy-to-use features managed to involve users at such a point that they were willing to promote the product to others and, therefore, provide the company with more users.

According to Garcia (2013), the translation-based learning strategy, especially as adopted by Duolingo, has been deemed effective by users and researchers alike (Vesselinov & Grego, 2012). In general, one major advantage of learning from a machine, which can be extended to all the firms partaking in the industry, is that the learner only has to interact with the software and focus on single words or sentences, not having to worry about the traditional setting of a communicative situation. One of Duolingo’s main advantages over its competitors has been the ability to promote their courses as more effective than others, even more than traditional classroom courses at universities, thanks to a study performed by Vesselinov and Grego (2012).

What makes it innovative is that the user can “discuss” the mistake with the machine, implement it with suggestions, receive the reasons why it is not correct and, most importantly, debate the mistake directly with other users of the platform. This innovative way to process and communicate feedback contributes to the advancement of both user’s and machine’s knowledge (Vesselinov & Grego, 2012). Duolingo’s capability featured in these apps of allowing the user to focus on the meaning of a word, inserted in different contexts, has also been widely praised by researcher and deemed as an important feature that adds to the effectiveness of such tools (Garcia Botero et al., 2017).

Considering Duolingo’s superior market position, it is fair to assume that the proven superior effectiveness of a method will attract more users to the company and will produce higher returns. The assumption respects the VRIO framework in that the resource, i.e. the courses, produces value if it is able to attract more users, is rare, as every method is unique to the company, and is hardly imitable, especially if it relies on the internal knowledge built by the human resources within the company.

Hypothesis 1b: Gamification features in a product attract more users.

The presence of gamification features has been found to have a significant influence over the firm’s success. According to research, the learner feels motivated by the achievement of rewards: this method acts as a measure to show the user if they are learning correctly, and allows them to feel challenged when they do not succeed in conquering a reward (Garcia, 2013). Gamification offers the user a totally personalized learning experience and makes it in general more interesting (Garcia Botero et al., 2017; Nushi & Eqbali, 2018).

Hypothesis 2: Machine learning, as developed by the AI-department within product management, contributes in the establishment of competitive advantage by improving method/product effectiveness and gamification features.

Given how new the concept of machine learning and the industry are, it is indeed difficult trying to establish whether the implementation of such technologies in these newly born firms affects business. Anyway, the literature does offer some considerations on how machine learning helps improve the effectiveness of the software. It is therefore hypothesized that, with the support of machine learning, the quality and effectiveness of the courses improve. A superior product in terms of quality is supposed to attract more users and lead the company to a competitive advantage.

One major advantage of MALL products is that machine learning allows the user to have a course that is fully tailored to their needs. By implementing machine learning in such courses, the software is able to adapt automatically to the student’s capacities, leading eventually to a learning success (Milutinović et al., 2013). The machine improves its own performances by customizing the experience to the user and, at the same time, acquires data that might be relevant to others, as seen in the exclusive speech recognition technology of Rosetta Stone, which identified common mistakes and adapted to multiple speech patterns (Garcia, 2013).

The role of the AI-department becomes relevant when the company is required to finetune algorithm feedback systems and, if applicable, reply to the user’s comments.

By allowing users to discuss and interact, Duolingo has created an online learning community, rather than just a mobile app. Learners, upon completing an exercise, get the possibility to comment, give feedback and discuss alternatives, which are screened both by the Duolingo team and other users of the platform (Teske, 2017; Vesselinov & Grego, 2012). This feature is currently present in Duolingo only and might constitute the proof that a correct management of the inputs processed by machine learning can constitute a source of competitive advantage.

Machine learning is important also in matter of gamification: as an individual interacts with the game, the machine gathers information in order to improve its adaptability. The machine can understand the preferences of a user and adapt them to future content (Lopez & Tucker, 2018), although these models are currently based on groups of people; AI professionals are yet to develop a model that creates individual responses.

Barney (1991) claimed that IT devices will hardly provide the firm with a competitive advantage, as they are commercially available to every firm. In 1995, though, by mentioning the importance of organization within a company, he provided a take that appears very relevant to the present paper. Every firm currently involved in the MALL industry makes use of machine learning to improve the adaptability of its products and make them more effective and enjoyable. Few firms, however, managed to reach a market leading position. Supposedly, firms like Duolingo and Rosetta Stone have been able to create algorithms that other firms have no access to, yet.

In conclusion, literary findings point at a connection between a firm’s success and a unique application of machine learning features.

CONCEPTUAL MODEL

.

Hypothesis 1: In the MALL-industry, a superior user acquisition capability leads to higher revenues and to a competitive advantage.

Hypothesis 1a: Superior effectiveness of the method/product leads to the acquisition of more users. Hypothesis 1b: Gamification features in a product attract more users.

Hypothesis 2: Machine learning, as developed by the AI-department within product management, contributes in the establishment of competitive advantage by improving method/product effectiveness and gamification features.

CONCLUSION

The final aim of this paper was to establish whether machine learning has an impact on the establishment of competitive advantage in the MALL industry. Given the lack of business research about the industry, the paper also investigated what the sources of competitive advantage might be by looking at two popular market leaders, namely Duolingo and Rosetta Stone.

It follows that the choice for the hypotheses was dictated by two main reasons: the need to investigate the surface of this so-far untouched industry and the general lack of business resources.

The first hypothesis aimed at providing an answer for the initial sub-question: what leads companies in the MALL industry to competitive advantage? In my opinion, it was necessary to first provide an answer to this question before establishing whether machine learning plays a role in the commercial success of an enterprise.

Rosetta Stone’s annual reports have been fundamental in clarifying what constitutes a source of competitive advantage in the industry: user acquisition. The more users/customers, the higher the revenues. Some companies acquire their profits directly from their customers through subscription plans, others earn indirectly through in-app advertisements and other ventures. Both options require the collection of users.

When screening through the literature, it appeared clear that two motifs were of interest for researchers and journalists alike: the effectiveness of digital learning methods and the importance of gamification.

These two characteristics seemed to be of the utmost importance for stakeholders, therefore it is assumed that companies tried to adapt to the contemporary user as much as possible by providing a product that is both effective and enjoyable. A company’s product that is renowned for being a good learning tool and fun to use will gather more users, leading towards competitive advantage.

With the second hypothesis I addressed whether machine learning gives a contribution to firms in achieving competitive advantage by improving the effectiveness of a product and its gamification features.

The outputs generated by machine learning originate from algorithms developed by a firm’s AI-department, which is expected to be an independent fraction within product management.

What appeared clear from the qualitative research is that a product’s success is boosted both by a scientifically proven effectiveness and enjoyment. Duolingo’s extremely vast pool of users and popularity is explained by the several factors described above combined. The innovative use of machine learning, accurately supervised by the AI-department, allows users to proceed in a course that feels tailor-made. The learner does not perceive studying a new language as a hard task, because gamification lightens the process and makes it enjoyable. Surely, the cost advantage and ease of accessibility did play a role in attracting as many users as possible.

A successful product in this industry needs to be able to correspond to the contemporary learner’s needs. Business and product managers in the MALL industry should focus on a product that allows users to learn while not realizing it. A highly gamified product that promises rewards as the learning proceeds looks like an optimal way to attract learners.

The combination of stakeholder theory and resource-based view appear to have been the perfect fit for this research. Firms in the MALL industry are highly dependent on feedback from users in order to adapt their resources accordingly.

IT resources will hardly provide a firm with a competitive advantage by themselves, as Barney already established (1991). However, the ability of the organization, i.e. product management, is expected to be able to combine the internal resources of the firm with the inputs coming from outside of it. Product management Feedback from stakeholders Technological and human resources Competitive advantage Stakeholder theory VRIO framework Flow chart 6: linking theories

Limitations

The present paper is characterized by typical limitations. The first and most apparent one being the lack of quantitative research: it is indeed impossible to prove the robustness of qualitative findings without statistical support.

All the literary findings are supported by academic papers deriving from fields different from business. Every assumption about the business and management side of the industry relates, therefore, to research in fields distant from it, although not completely incompatible. Applying business theories to findings in other fields allows for hypothesizing only.

Given the space and time constraint, only two firms have been analyzed thoroughly, despite the presence of other relevant players in the industry such as Babbel, Lingodeer and Memrise.

Duolingo is a private company: many financials have been estimated following communications issued by the company itself, meaning the information is biased.

A possible advantage given by the price point of the products has been mentioned throughout the paper, but not supported by literary findings.

Choosing product management as an activity to investigate allowed for a big scope of research: it was possible to generalize the role of product managers when it comes to development of a course. Arguably, the generalization was maybe too wide and more departments within the scope of product management should have been mentioned.

Future research

Other than providing an answer for the research questions, a major goal of this research was to shine a light over this specific branch of the software/ mobile apps industry. The hope is that the current paper will rise questions, curiosity and will of engagement in business researchers.

Clearly, this research can be made more effective and relevant by an additional quantitative check: linear regressions between revenue growth and acquisition/loss of users and between acquisition/loss of users and promoted effectiveness of the methods would be fitting.

Moreover, future research should look into the work of other market-leading firms to further specify what characterizes competitive advantage in the MALL industry.

REFERENCES

Adams, S. 2019. Game of Tongues: How Duolingo Built A $700 Million Business With Its Addictive Language-Learning App. Forbes Magazine.

Adams, S. 2020. The Language App That’s Helping Bored People Get Through The Lockdown Susan. Forbes Magazine.

https://uk.babbel.com/

Barney, J. 1991. Firm Reources ad Sustained Competitive Advantege. Journal of Management. Barney, J. 1995. Looking inside for competitive advantage. Academy of Management Executive,

9(4): 49–61.

Bidin, S., & Ziden, A. A. 2013. Adoption and Application of Mobile Learning in the Education Industry. Procedia - Social and Behavioral Sciences.

Böhm, S., & Constantine, G. P. 2016. Impact of contextuality on mobile learning acceptance: An empirical study based on a language learning app. Interactive Technology and Smart

Education, 13(2): 107–122.

www.bvdinfo.com (orbis database)

Byron, P. 2020. What Is Artificial Intelligence? Journal of Accountancy, 32(3).

Collentine, J. 2000. Insights into the construction of grammatical knowledge provided by user-behavior tracking technologies. Language Learning and Technology, 3(2): 44–57.

Donaldson, T., & Preston, L. E. 1995. The stakeholder theory of the corporation. The Academy of

Management Review, (1): 65–91.

duolingo.com

Ebert, C. 2007. The impacts of software product management. Journal of Systems and Software, 80(6): 850–861.

Ebert, C., & Brinkkemper, S. 2014. Software product management - An industry evaluation. Journal

of Systems and Software, 95: 10–18.

Freeman, R. E. 2004. Zeitschrift für Wirtschafts- und Unternehmensethik. Zeitschrift Für

Wirtschafts- Und Unternehmensethik, 5(3): 228–254.

Review, 25(2): 42.

Garcia, I. 2013. Learning a Language for Free While Translating the Web. Does Duolingo Work?

International Journal of English Linguistics, 3(1): 19–25.

Godwin-Jones, R. 2017. Emerging technologies: Smartphones and Language Learning. Language

Learning & Technology, 21(2): 3–17.

Heil, C. R., Wu, J. S., Lee, J. ., & Schmidt, T. 2016. A review of mobile language learning applications: trends, challenges and opportunities. The EuroCALL Review, 24(2, September 2016). https://doi.org/10.5267/j.msl.2018.7.002.

Hönig, F., Batliner, A., & Nöth, E. 2012. Automatic Assessment of Non-Native Prosody — Annotation, Modelling and Evaluation. Proceedings of the International Symposium on

Automatic Detection of Errors in Pronunciation Training (IS ADEPT).

Huynh, D., Zuo, L., & IIda, H. 2016. Analyzing Gamification of “Duolingo” with Focus on Its Course Structure. EyeMath: Identifying Mathematics Problem Solving Processes in a RTS Video

Game, vol. 10056: 268–277.

https://www.independent.ie/entertainment/banter/gizmo/1m-people-sign-up-to-learn-irish-on-duolingo-in-just-one-year-31524516.html

Jones, T. M., Harrison, J. S., & Felps, W. 2018. How Applying Instrumental Stakeholder Theory Can Provide Sustainable Competive Advantage: Business Source. Academy of Management

Review, 43(3): 371–391.

Karlsson, L., Dahlstedt, Å. G., Dag, J. N. O., Regnell, B., & Persson, A. 2002. Challenges in Market-Driven Requirements Engineering - an Industrial Interview Study. Information and Software

Technology, 49(6): 588–604.

Kukulska-Hulme, A., & Shield, L. 2008. An overview of mobile assisted language learning: From content delivery to supported collaboration and interaction. ReCALL, 20(3): 271–289.

Lopez, C., & Tucker, C. 2018. Towards Personalized Adaptive Gamification: A Machine Learning Model for Predicting Performance. IEEE Transactions on Games, PP(c): 1–1.

Lu, M. 2008. Effectiveness of vocabulary learning via mobile phone. Journal of Computer Assisted

Learning, 24(6): 515–525.

Marques Morgado, P. 2014. Sustained Competitiveness. Leids-Haags Allegorieen Lexicon, Leiden

McCarty, S., Sato, T., & Obari, H. 2017. Implementing Mobile Language Learning.

Milutinović, M., Labus, A., Stojiljković, V., Bogdanović, Z., & Despotović-Zrakić, M. 2013. Designing a mobile language learning system based on lightweight learning objects. Multimedia

Tools and Applications, 74(3): 903–935.

Munday, P. 2017. Duolingo. Gamified learning through translation. Journal of Spanish Language

Teaching, 4(2): 194–198.

Nushi, M., & Eqbali, M. H. 2018. Babbel : A Mobile Language Learning App Babbel : A Mobile

Language Learning App Review by Musa Nushi and Mohamad Hosein Eqbali, (April).

https://research.duolingo.com/

Roberts, E. B., & Meyer, M. H. 1991. Product strategy and corporate success. IEEE Engineering

Management Review, 19(1): 4–18.

https://www.rosettastone.com/

Rosetta Stone, Inc. 2009. Annual report. Rosetta Stone, Inc. 2010. Annual report. Rosetta Stone, Inc. 2011. Annual report. Rosetta Stone, Inc. 2012. Annual report. Rosetta Stone, Inc. 2013. Annual report. Rosetta Stone, Inc. 2014. Annual report. Rosetta Stone, Inc. 2015. Annual report. Rosetta Stone, Inc. 2016. Annual report. Rosetta Stone, Inc. 2017. Annual report. Rosetta Stone, Inc. 2018. Annual report. Rosetta Stone, Inc. 2019. Annual report.

Saloky, T., & Šeminský, J. 2019. Artificial Intelligence and Machine Learning. Studies in Health

Technology and Informatics, 261: 135.

Saunders, M., Lewis, P., & Thornhill, A. 2009. Research methods for business students - fifth edition.

Pearson Education. https://doi.org/10.1080/09523367.2012.743996.

https://www.forbes.com/sites/avaseave/2016/09/23/in-the-battle-of-online-language-learning-programs-who-is-winning/#793f1efc578b.

Settles, B and Lafair, G.T.; Hogiwara, M. 2020. Machine Learning Driven Language Assessment. Transactions of the Association for Computational Linguistics, 2020.

Teske, K. 2017. Duolingo, 34: 393–401.

Vesselinov, R. 2009. Measuring the Attitude and Motivation of Rosetta Stone ® Users. http://www.vesselinov.com/RV_Language_Motivation.pdf.

Vesselinov, R., & Grego, J. 2012. Duolingo Effectiveness Study. City University of New York, USA, (December 2012): 1–25.

Wu, M. 2013. Towards a Stakeholder Perspective on Competitive Advantage. International Journal