Illuminating Mobility as a Service ecosystems using natural language processing

Keywords:

Abstract:

Purpose
The automotive industry is undergoing fundamental changes (Jacobides et al., 2016) due to increased disruptive competition from future mobility tech companies (Adner and Lieberman, 2021), the pressure to reach the decarbonization goals stated in the Paris Agreement (United Nations, 2015) and requests for more equal mobility options (Hickman and Banister, 2019). One solution to reach the sustainability transition currently debated about and enabled by technology is participating in Mobility as a service (MaaS) with the integration of different mobility services, including sharing options and electromobility, on one platform (Mulley et al., 2020). A MaaS ecosystem is a special form of transaction ecosystem, consisting of a multi-sided marketplace which allows the use of shared resources to create joint value of which each partner tries to capture a share (Polydoropoulou et al., 2020a). Research on MaaS has primarily looked at its social and environmental impact (Alyavina et al., 2020). The business models of companies contributing to MaaS offerings have received little attention so far (Polydoropoulou et al., 2020b). Consequently, the economic viability of MaaS has not yet been proven (Liljamo et al., 2020).
One prerequisite is the breaking up of the individual mobility service providers’ traditional organizational structures into functioning ecosystems (Hensher et al., 2020). Most of the many pilot projects have not reached the implementation stage (Karlsson et al., 2020). Therefore, the aim of this paper is to better understand the challenges of designing enduring MaaS ecosystems for private (sharing) companies by answering the first research question:
1. What is the status quo of MaaS ecosystems involving private companies?
In order to identify possible barriers to MaaS ecosystem development (Karlsson et al., 2020), a second research question is formulated:
2. How can the barriers to further implementation of MaaS ecosystems be identified and overcome?
Design
Natural language processing (NLP, e.g. Otter et al., 2021) is employed to escape the failure of traditional textual analyses because start-ups and small businesses involved in MaaS offerings are not subject to publicity obligations. NLP can not only be used to investigate the status quo of MaaS offerings, but also to find explanations for possible barriers which hinder the development of MaaS ecosystems (Karlsson et al., 2020). The study approach is based on Miric et al. (2023) and Choudhury et al. (2021) and consists of indicator definition, indicator iteration, and quality assurance.
First, each building block of a MaaS ecosystem, i.e. an overarching focal value proposition (Lusch and Vargo, 2006), operating model (Jacobides et al., 2018), a technical platform (Karlsson et al., 2020), value drivers (Leppänen et al., 2023), and governance (Dyer et al., 2018), was further broken down into variables (categories). Furthermore, information on customers and on external MaaS governance was collected (see Figure 1).
The second step concerned the technical implementation of the AI including iterations to increase data volume and indicator scope. Here, the method of topic modeling was applied to articles in newspapers and magazines as well as social media entries on 43 private firms each in majority privately owned, offers at least two transportation services and is a partner in at least one MaaS ecosystem with their worldwide activities in MaaS ecosystems in the years 2019 to 2022. The indicator iteration process took place in three steps: a) construction of training data: diagnosis of size, balance and representativeness, b) construction of numerical representation of text data and c) comparison of classification models.
The quality assurance succeeded as a check for data volume, accuracy and metadata across indicators in iterative process with indicator definition and set up.
Findings
For the years 2019 to 2022, 59,787 pieces of evidence from 8,546 individual texts were obtained. The results reached an accuracy of 76% to cover both width and depth. Of these, 52,055 pieces of evidence concerned the five building blocks of MaaS ecosystems. The communication in MaaS concentrated on a few MaaS players: 80% of all 59,787 evidences come mainly from five (11.6%) of the private partner companies studied and there are only few evidences on each of the remaining 38 companies (only 20% altogether). In addition, an above-average proportion of the evidences came from North America. This is undoubtedly partly because only English-language texts could be considered.
Regarding the indicators (building blocks of MaaS ecosystems), most were found for (1), the value proposition (approx. 80% of the mentions, Figure 1a). The other building blocks were mentioned less frequently. This also applied to the organisational framework of such ecosystems (the operating model (2), with only about 6% of the evidence) and the technical basis (the platform (3), with just 1% of the evidence), although both building blocks were mentioned as necessary conditions for the design of minimally viable MaaS ecosystems and have been well investigated in the scientific literature. Correspondingly, value drivers (4) and governance mechanisms (5) as sufficient conditions for joint value creation and individual value capture in MaaS ecosystems were also rarely mentioned only little in the analysed texts (9% and 4% of the evidence respectively). There are also major differences in the categories of the individual building blocks. Among the variables of (1) the value proposition, the basic benefit was mentioned most frequently (almost 69.3% of the mentions), additional benefits accounted for 30.7% of the mentions.
Over the years 2019 to 2022, changes were seen in the frequency of evidence of the building blocks of MaaS ecosystems and their categories. Figure 1b shows the normalised categories (2019 = 100% respectively 0) to compare the graphs. The figure reveals that in the last four years the structure of the results shown in Figure 1a has remained essentially the same.

Figure 1: Frequency of evidences of the five indicators (building blocks) of MaaS ecosystems (own design)
Our research mainly shows internal barriers within the ecosystem. Above all, a holistic view of a minimum viable ecosystem is missing. Another internal barrier seems to be the lack of understanding the problems of implementing the operating model and the digital exchange platform. An external barrier to the implementation of MaaS solutions seems to be the lack of a roadmap for government regulation with a clear timeline of future support measures.
Practical and theoretical implications
Four approaches to overcome barriers can be deduced. Static approaches to overcome internal barriers are needed to broaden the activities in the core of MaaS –operating model, platform, value driver and internal governance, such as a more intensive investigation of the distribution of assets in the operating model and the integration of billing and controlling mechanisms via the platform.
Dynamic approaches to overcome internal barriers need to focus more on the process of minimum viable ecosystem design.
Static approaches to overcome external barriers are based on the insight that internal governance mechanisms are not sufficient, nor are self-regulations (Gawer and Srnicek, 2021).
Dynamic approaches to overcome external barriers need a regulatory roadmap to be developed for the next five to seven years.

References
Adner, R., Lieberman, M., 2021. Disruption Through Complements. Strategy Science 6 (1), 91–109. https://doi.org/10.1287/stsc.2021.0125.
Alyavina, E., Nikitas, A., Tchouamou Njoya, E., 2020. Mobility as a service and sustainable travel behaviour: A thematic analysis study. Transportation Research Part F: Traffic Psychology and Behaviour 73 (2), 362–381. https://doi.org/10.1016/j.trf.2020.07.004.
Choudhury, P., Allen, R.T., Endres, M.G., 2021. Machine learning for pattern discovery in management research. Strat Mgmt J 42 (1), 30–57.
Dattée, B., Alexy, O., Autio, E., 2018. Maneuvering in Poor Visibility: How Firms Play the Ecosystem Game when Uncertainty is High. AMJ 61 (2), 466–498. https://doi.org/10.5465/amj.2015.0869.
Dyer, J.H., Singh, H., Hesterly, W.S., 2018. The relational view revisited: A dynamic perspective on value creation and value capture. Strat Mgmt J 39 (12), 3140–3162. https://doi.org/10.1002/smj.2785.
Gawer, A., Srnicek, N., 2021. Online platforms: Economic and societal effects, 144 pp.
Hensher, D.A., Mulley, C., Ho, C., Wong, Y., Smith, G., Nelson, J.D., 2020. Understanding mobility as a service (maas): Past, present and future, 1st ed. Elsevier, San Diego, volumes cm.
Hickman, R., Banister, D., 2019. Transport and the environment. In: Stanley, J., Hensher, D.A. (Eds.) A Research Agenda for Transport Policy. Edward Elgar Publishing, Cheltenham, pp. 25–33.
Jacobides, M.G., Cennamo, C., Gawer, A., 2018. Towards a theory of ecosystems. Strat Mgmt J 39 (8), 2255–2276.
Jacobides, M.G., MacDuffie, J.P., Tae, C.J., 2016. Agency, structure, and the dominance of OEMs: Change and stability in the automotive sector. Strat. Mgmt. J. 37 (9), 1942–1967. https://doi.org/10.1002/smj.2426.
Karlsson, I., Mukhtar-Landgren, D., Smith, G., Koglin, T., Kronsell, A., Lund, E., Sarasini, S., Sochor, J., 2020. Development and implementation of Mobility-as-a-Service – A qualitative study of barriers and enabling factors. Transportation Research Part A: Policy and Practice 131, 283–295. https://doi.org/10.1016/j.tra.2019.09.028.
Leppänen, P., George, G., Alexy, O., 2023. When Do Novel Business Models Lead to High Performance? A Configurational Approach to Value Drivers, Competitive Strategy, and Firm Environment. AMJ 66 (1), 164–194. https://doi.org/10.5465/amj.2020.0969.
Liljamo, T., Liimatainen, H., Pöllänen, M., Utriainen, R., 2020. People’s current mobility costs and willingness to pay for Mobility as a Service offerings. Transportation Research Part A: Policy and Practice 136, 99–119. https://doi.org/10.1016/j.tra.2020.03.034.
Lusch, R.F., Vargo, S.L., 2006. Service-dominant logic: reactions, reflections and refinements. Marketing theory 6 (3).
Miric, M., Jia, N., Huang, K.G., 2023. Using supervised machine learning for large‐scale classification in management research: The case for identifying artificial intelligence patents. Strat Mgmt J 44 (2), 491–519. https://doi.org/10.1002/smj.3441.
Mulley, C., Ho, C., Balbontin, C., Hensher, D., Stevens, L., Nelson, J.D., Wright, S., 2020. Mobility as a service in community transport in Australia: Can it provide a sustainable future? Transportation Research Part A: Policy and Practice 131 (10), 107–122. https://doi.org/10.1016/j.tra.2019.04.001.
Otter, D.W., Medina, J.R., Kalita, J.K., 2021. A Survey of the Usages of Deep Learning for Natural Language Processing. IEEE Trans. Neural Netw. Learning Syst. 32 (2), 604–624. https://doi.org/10.1109/TNNLS.2020.2979670.
Polydoropoulou, A., Pagoni, I., Tsirimpa, A., Roumboutsos, A., Kamargianni, M., Tsouros, I., 2020a. Prototype Business Models for Mobility-as-a-Service. Transportation Research Part A 131, 149–162.
Polydoropoulou, A., Tsouros, I., Pagoni, I., Tsirimpa, A., 2020b. Exploring Individual Preferences and Willingness to Pay for Mobility as a Service. Transportation Research Record 2674 (11), 152–164. https://doi.org/10.1177/0361198120938054.
United Nations, 2015. Paris Agreement. https://unfccc.int/files/essential_background/convention/application/pdf.... Accessed 7 July 2020, 27 pp.