Mapping an App Ecosystem: Exploring “Female Health” on Google Play

Team Members

Team members: Alette Schoon, Mathis Guerreiro, Morgane Billuart, Sne Scott Hansen, Tom Weber

Designer: Andrea Benedetti (DensityDesign)

Project facilitators: Anne Helmond, Esther Weltevrede, Michael Dieter, Jason Chao

Contents

Summary of key findings

Following the rise in the development and use of health apps for the collection, storage, processing, monitoring, and tracking of intimate data about the body, "femtech" – a distinct branch in the category of health apps targeted at women – has begun to flourish on the consumer market for apps accessible on platforms such as the Google Play store. Due to the rapid development of these types of apps, we still know little about the landscape of apps that go by the name femtech, as well as the permissions for data collection and tracking that these apps are imposing on their users. This study took an exploratory approach at mapping exactly this ecosystem of femtech apps by querying 50 “female health” apps from the Google Play store in the USA. The initial findings show two main categories of female health apps, including “medical” apps, and “health and fitness” apps that most prominently were based on words such as “workout”, “fitness”, “women”, “symptoms” and “ovulation”. Furthermore, the findings point towards a number of problematic tracking permissions that the apps are asking from their users, including users’ precise location and audio recording. We argue that these types of permissions seem far from relevant for the apps to capture female health and thus conflict with the way in which the apps typically market themselves towards consumers based on the above-mentioned categories and keywords found within the app sample. Additionally, the findings also show that all of the 50 apps collected are working together with third parties within the wider app ecosystem and for the most part Google and Facebook. In light of these findings, we suggest future studies look into issues concerning data privacy, not only in the use of health apps in general but also in more distinct types of health apps.

Introduction

Over the course of the past years, health apps and digital health have been continuously expanding on app stores, offering users the possibility to track their mood, symptoms, progress, and overall state of being. “Femtech”, a term introduced in 2016 by the founder of period tracking application Clue, Ida Tin, focuses on female health and experiences of women and menstruating bodies (Mishra and Suresh 598). The promises of this field are numerous, one of their ambitions being to challenge and fill the gaps of knowledge in regards to male bias in medicine and science (Hendl and Jansky 1). As for digital health, this collection of information and visualization can only happen through the collection of data and experiences of their users, leading to question the ways in which those apps are designed and their data stored.

Within the context of femtech, mobile applications are of particular interest and exhibit particular importance due to their unique mode of distribution – mobile devices are becoming increasingly affordable and accessible in large parts of the world, and similarly, the requirements for developing and publishing mobile applications are becoming increasingly easy to meet (Fox et al. 541). Given these circumstances, apps as part of the wider femtech space have a high potential to meet the promises of the field and to contribute to the filling of knowledge gaps, due to their accessibility, rapid circulation, and their affordances for the generation of user data. At the same time, however, contemporary app landscapes exhibit high degrees of platform dependencies, with the two main app stores as sites of distribution owned by large platform corporations Apple and Google, who fulfill strong gatekeeping functions with respect to what apps are allowed on their store, what apps are given visibility over others and modes of app monetization (Carolin Gerlitz et al.). Further, app development is heavily reliant on connections to third parties’ tools and infrastructures, such as social media platforms, analytics platforms, or ad networks, for feature development, data analysis, and monetization (C. Gerlitz et al.). Said dependencies have in the past raised concerns about the collection and storage of user data and who is granted access to it (Fox et al. 541).

Contributing to this emerging field, our research aims at exploring the current state of femtech as part of the wider app landscape. As such, we aim to answer the following research question: What do different mappings of female health app spaces reveal about the state of femtech as part of larger app economies? Through several mappings of an app space, we contribute to offering a starting point for a conversation about what categories apps related to female health operate in, what their target group is, linguistic and visual narratives, and their connections to third parties and their infrastructures.

In what follows, we will first outline our methodological approach for this study, explaining how we narrowed down our scope to generate an initial selection of apps for further inspection, as well as tools we used to execute said exploration. Subsequently, we will situate our research in contemporary academic debates within femtech and the larger field of app studies, followed by a presentation of our main findings. Lastly, we will synthesize our findings, offer a conclusion through a comparison of different mappings and their implications and make suggestions for future research.

Methodology and initial datasets

In order to answer the above research question, we employed an exploratory approach to best account for the multifaceted and multi-relational nature of apps as a research object, where demands of end-users, developers and platform owners meet. As such, mobile applications allow for the collection and analysis of data on a multitude of levels: “visible” and openly accessible data, such as close readings and content analysis of linguistic, textual, and visual data generated by developers themselves or by users through the app or the app store, “invisible” data, hard-coded in an application and accessible only through the backend, and data generated by app stores as virtual marketplaces and sites of distribution. Through distinct analysis and mappings of app data through these different lenses, we gain insights into the wider state of femtech as part of the app economy – more concretely, what “female health” means through the eyes of the app landscape and what elements of female health most apps respond to, how those are visually and linguistically represented to users, and how female health apps are connected, stand in relation to, and are infrastructurally dependent on each other in larger networks with other apps and third-party providers.

In order to generate an initial dataset, we decided to query the term “female health” on Google Play, an app distribution site operated by US-American technology conglomerate Google (Alphabet Inc.) and the official app store for mobile devices that run on the Android operating system. Further, we limited the query to the US market and English language results. Our query “female health” is appropriate for and produces search results representative for an exploration of mobile applications that are part of the wider femtech space or the category of apps that revolve around women’s health. Despite deciding on an exploratory approach for our study, we narrowed down the scope of our exploration in these ways due to resource and time constraints. We chose to focus solely on a selection of apps on Google Play, disregarding Apple’s App Store for devices that run on the iOS operating system, for the reason that apps on Google Play allow for a deeper level of data collection and analysis due to less strict store developer regulations. Further, we chose the US market and English language results to avoid having to rely on translation software for query design and data analysis that could potentially falsify results (all members of our research group are proficient in English). In order to generate our initial data set, we then selected the first 50 mobile apps that responded to the query “female health” under the scope mentioned above, ensuring that the results of our exploration are to a certain extent representative of the respective app space, but also allowing for qualitative inquiry of individual apps.

Moving forward, we made use of several tools for the gathering and analysis of data from the 50 apps that made up our initial selection. We used an app scraper provided by the Digital Methods Initiative, a tool that extracts data from processes and flows not designed for human analysis (“scraping”) and translates said data for further inspection, to gather full details of each app, such as title and description, its category in the store, as well as the number of total downloads, but also an app's connection to related apps and permissions it asks a user to agree to. Further, we used a tool named ‘AppInspect’ to extract data from the actual application files (APKs) and their source code, providing insights into what third-party trackers are integrated into an application, as well as the sensors in a mobile device that it is trying to access.

The tools described above allowed us to gather several sets of data focusing on different aspects of the 50 applications we selected for this study. Based on those data sets, we created different mappings of store-generated data, such as ranking apps by number of downloads, grouping them by their categories or in a network of recommended/related apps, and developer-generated data, such as keywords most prevalent in app descriptions, and networks between apps and third-party tools and services.

Theoretical framework

Contemporary research in the field of app studies recognizes problematic structures and dependencies within the larger state of contemporary app development, where mobile applications are developed for and on platforms governed by large technology conglomerates that dictate their practices and modes of monetization (Dieter et al. 22). Scholars reflect the critical view on mobile applications with regards to concerns around data privacy (Weltevrede and Jansen). In their study on period tracking apps, Fox et al. highlight the multilayered nature of data protection and data privacy in the context of apps, that emerges from complex relationships between app developers, end-users, store operators, and third-party providers: “responsibility for privacy and data protection is distributed among users, app developers, operating system, and platform providers, where responsibility rests in the hands of all and none at the same time” (551).

Femtech as an emerging space reflects many of these concerns. In their exploration of femtech through reproductive technology and applications, Mishra and Suresh bring attention to the need for exploring what categories of “health” apps in the femtech space relate to: “ femtech reassociates women in medicine with their biological reproductive functions, overlooking the wide variety of health conditions women may experience” (599). Further, the authors point to the problematic narratives surrounding body tracking apps, that are advertised by developers as a way to gain control over one’s own physical condition through systematic tracking and data, whereas in many cases, said data is being used for advertising and monetization purposes (Mishra and Suresh 603). Hendl and Jansky emphasize an apparent disconnect between the representation of apps through store marketing materials, the actual purpose of these apps, and how these fit into wider femtech imaginaries: “A rhetoric of apps that sells users a mediated, selective and inaccurate notion of knowledge cloaked as a reliable form of self-knowledge, critiqued heterosexist conceptualizations of women’s sexuality and problematized an oppressive and overburdening notion of feminine self-optimization promoted in the discourse of apps as well as pointed towards troubling evidence of privacy and consent violations involved in the commercialization of sensitive user data by some apps” (Hendl and Jansky 22).

Findings

Our findings can be categorized into six sections. First, we looked at the categories of the first fifty applications that appear when "female health" is queried on the Google Play Store. Then, we determined the most downloaded apps. The third part of our findings focuses on the main keywords that appear in the apps' descriptions. The fourth part emphasizes the recommendation network of those apps. The fifth section focuses on the third-party trackers and finally, the sixth focuses on the permissions required by those apps.

App Categories

Fig. 1 Apps by categories

When "female health" is queried on the US's Google Play Store, the categories of the first fifty apps can be collected by using the Google Play Scraper developed by the Digital Method Initiative (DMI). As the above diagram shows, 36 apps out of 50 are categorized in Health and Fitness which corresponds to 72% of the total sample size. The second major category in which those apps are classified is Medical, with 11 apps and 22% of the sample size. The remaining categories are marginal, with one app in each category.

Fig. 2 Apps by number of downloads

Among the first fifty apps that appear when "Female Health" is queried, Flo is by far the most downloaded app on the Google Play Store with more than 80 million downloads. Therefore, we can see that Flo, a menstruation calendar, is predominant in the query for "Female Health".

App Description Keywords

Fig. 3 Keywords in app descriptions

We analyzed the descriptions of the sample of apps in order to create a tag cloud. It visually represents the predominance of certain keywords among the aggregation of all the apps' descriptions. This predominance is represented by the size of the keyword in question. The above tag cloud shows that the main keywords are "health", "women", "workout", and "period". The first two keywords are close or identical to the original query. Although, "period" and "workout" hint at the type of service that fall into the broader scope of "Female Health" on the Google Play Store. Additional keywords can be said to fall into the same category as "period", such as "ovulation", "pregnancy", "cycle", or "fertility". The same can be said about the keyword "workout", with other entries such as "weight", "fitness", "exercise", or "coach". Thus, it is clear that two main domains stand out for the query under investigation. Menstruations on one hand, and physical self-development on the other. Also, it is worth mentioning that tracking technology is central to this tag cloud with the keywords "track" and "tracker" which accumulate 118 and 117 entries respectively.

Recommendation Network

Fig. 4 Network of apps and their recommended apps

When users land on an app's description page on the Google Play Store, other apps are suggested to them. This network graph shows two main and distinct groups. On the top left corner, apps that relate to the Fitness category are isolated. While on the bottom right corner, apps that relate to the Medical, Health, and Menstruation categories form a solid group of recommendations. Apps that pertain to the same relational group recommend each other on their description page. They can also recommend apps from another group but to a lesser extent.

Third-party Trackers

Fig. 5 Apps and their third-party trackers

This third-party tracker network shows trackers in purple and app IDs in yellow. The size of tracker nodes varies according to the number of relations they have with app IDs. In other words, the bigger a tracker node is, the more omnipresent it is among the app sample. As the network above shows, Google Firebase Analytics is the most present tracker among "Female Health" apps on the Google Play Store. Followed by Facebook Share, Facebook Login, Facebook Analytics, Google AdMob, and Google Crashlytics. Overall, it appears clearly that the two main tracker providers for the apps under study are related to Google and Facebook.

Permissions

Fig. 6 Apps and their requested permissions

The above matrix plot shows what permission requests an app requires from its users. The abscissa lists permissions and the ordinate lists app IDs. As we can see, all apps require permissions. Some permissions can be considered more invasive than others. For instance, some apps require access to precise location data, downloading files without authorization, recording audio, or reading call logs. Permissions that are considered particularly invasive are colored in purple and red. Although, the invasive character of permission is largely subjective and may vary according to users.

Discussion and conclusion

Throughout this explorative research, we have mapped different categories and genres based on data we extracted from apps on Google Play. Our findings indicate a distinct discrepancy between the way apps are categorized in stores through the publishing infrastructure as provided by store operators, and their actual distribution in terms of their popularity and size of the user base, their focus or “purpose” as depicted and advertised linguistically through store marketing material, as well as their connection to third-party apps in the wider ecosystem. We have shown that Google Play as a site of distribution only allows for categorization of apps relating to our query “female health” under the main categories “Health and Fitness” and “Medical”, whereas the largest body of these apps (in terms of the number of downloads and linguistic representation) in fact relate to female reproductive health, menstruation and periods.

Further, we dissected the different uses and definitions of those apps to conclude that although a large part of female health apps is centered around fitness, the most downloaded ones were related to menstrual health, which led us to raise questions regarding the linguistic and visual uses of female health. In the further exploration of these apps and their utility, we have underlined concerns about data security, privacy, and surveillance based on the mapping of relationships with third parties and trackers. Lastly, our findings mirror concerns over contemporary states of app development and distribution, that jeopardize femtech agendas due to strong dependencies on platform governance and data flows to third parties.

We identify potential for future research on data storage and flows between apps and third parties but also qualitative explorations of popular apps and cross-cultural comparative mappings. We are aware of the limitations of our study in terms of its scope. A larger data set would have allowed for more representative results and deeper qualitative inquiry into single apps and their user base would have added additional layers of robustness.

References

• Dieter, Michael, et al. “Pandemic Platform Governance: Mapping the Global Ecosystem of COVID-19 Response Apps.” Internet Policy Review, vol. 10, no. 3, 2021. https://doi.org/10.14763/2021.3.1568.
• Fox, Sarah, et al. Vivewell: Speculating Near-Future Menstrual Tracking through Current Data Practices. ACM, 2019, pp. 541–52. https://doi.org/10.1145/3322276.3323695.
• Gerlitz, C., et al. “Apps and Infrastructures – a Research Agenda.” Computational Culture – A Journal of Software Studies, vol. 7, 2019. http://computationalculture.net/apps-and-infrastructures-a-research-agenda/
• Gerlitz, Carolin, et al. “Regramming the Platform: Infrastructural Relations between Apps and Social Media.” Computational Culture – A Journal of Software Studies, vol. 7, 2019. http://computationalculture.net/regramming-the-platform/
• Hendl, Tereza, and Bianca Jansky. “Tales of Self-Empowerment through Digital Health Technologies: A Closer Look at ‘Femtech.’” Review of Social Economy, https://doi.org/10.1080/00346764.2021.2018027.
• Mishra, Paro, and Yogita Suresh. “Datafied Body Projects in India: Femtech and the Rise of Reproductive Surveillance in the Digital Era.” Asian Journal of Women’s Studies, vol. 27, no. 4, Routledge, 2021, pp. 597–606. https://doi.org/10.1080/12259276.2021.2002010.
• Weltevrede, E., and F. Jansen. “Infrastructures of Intimate Data: Mapping the Inbound and Outbound Data Flows of Dating Apps.” Computational Culture – A Journal of Software Studies, vol. 7, 2019. http://computationalculture.net/infrastructures-of-intimate-data-mapping-the-inbound-and-outbound-data-flows-of-dating-apps/