2.1. The ‘Platform Economy’ and its extension into public welfare

The ‘Platform Economy’ is typically understood as a system of digital infrastructures that mediate interactions among distinct user groups through the governance of access, participation, and data flows by platform operators (Kenney et al., Reference Kenney, Bearson and Zysman2021). While originally associated with commercial innovations – such as Amazon, Uber, and Airbnb – platforms have since evolved into an organisational paradigm that is reshaping institutional roles and service delivery mechanisms (Chimenti et al., Reference Chimenti, Hagberg and Araujo2025). In the public domain, platform logic is increasingly being adopted across welfare sectors, with modular and interoperable systems now permeating healthcare, employment services, and social assistance (Xue et al., Reference Xue, Tian and Zhao2020; Zander et al., Reference Zander, Lu and Chimenti2025). This development constitutes what Margetts and Dunleavy (Reference Margetts and Dunleavy2013) describe as a second wave of digital-era governance in which governments focus less on direct provision and more on orchestrating service interfaces and coordinating multi-actor ecosystems.

2.2. Digital welfare, co-production, and e-government platforms

Digital welfare platforms are frequently embedded within broader e-government strategies aimed at enhancing integration, accessibility, and user-centred approaches. Early iterations, such as one-stop government portals, emphasised administrative simplification and centralised access (Wimmer and Tambouris, Reference Wimmer, Tambouris and Traunmüller2002). More recent developments encompass a wider range of digital functionalities, including open data initiatives and algorithmic eligibility assessment tools (Kawabata and Camargo, Reference Kawabata and Camargo2023). The co-production literature offers a critical interpretive framework: Osborne et al. (Reference Osborne, Radnor and Strokosch2016) argue that public value is not simply delivered by institutions but is co-created through dynamic interactions among users, providers, and systems. Digital platforms may facilitate such interaction through participatory design and feedback mechanisms, but they also risk marginalising users when efficiency imperatives overshadow inclusivity (Novello and Garavaglia, Reference Novello and Garavaglia2026).

2.3. Fragmentation, integration, and recomposition of welfare provision

The governance of welfare services has become increasingly fragmented, marked by the disaggregation of roles among public, private, and third-sector actors (Powell, Reference Powell2019). Digital platforms have emerged as potential solutions to this fragmentation, functioning as integrative infrastructures capable of reconfiguring previously disjointed systems (Bennett and Meers, Reference Bennett and Meers2025). Longo and Maino (Reference Longo and Maino2021) distinguish between multichannel platforms, which aim to re-establish coordination by offering multiple points of access, and marketplace platforms, which standardise service supply by aggregating providers. Both configurations position platforms as institutional intermediaries that can bridge fragmented welfare landscapes, although their integrative capacity is contingent upon governance arrangements and normative commitments to equity.

2.4. Platform infrastructures and institutional logics

Platforms are not merely technical artefacts but are embedded within socio-technical infrastructures that embody specific institutional logics. These logics shape the rules and values that govern eligibility, participation, and performance evaluation (Frapporti, Reference Frapporti, Mezzadra, Cuppini, Frapporti and Pirone2024; Zuiderwijk et al., Reference Zuiderwijk, Chen and Salem2021). For example, marketplace platforms may generate quasi-markets in which public, private, and non-profit actors negotiate access and pricing, embedding competing values such as efficiency, equity, and profitability (Dulong de Rosnay and Stalder, Reference Dulong de Rosnay and Stalder2020). Alternatively, commons-based platforms prioritise transparency and participatory governance. These divergent architectures shift control from administrative discretion to codified digital processes (Lund-Tønnesen, Reference Lund-Tønnesen2026), with significant implications for user experience – including the potential for exclusion, frustration, or disempowerment (Oude Geerdink et al., Reference Oude Geerdink, Sewdas, van Kempen, van Weeghel, Anema and Huysmans2023).

2.5. Research gaps and the need for exploratory global mapping of digital welfare platforms

Despite a growing interest in digital welfare platforms, existing empirical research remains limited in scope. Most studies concentrate on individual countries, specific sectors, or flagship cases, while comparative analyses are often theoretical or conceptual in nature. A comprehensive, global mapping of welfare platforms is thus lacking. The present study addresses this empirical gap through the construction and analysis of a novel dataset encompassing 285 digital welfare platforms, categorised across seven analytical dimensions. This approach offers an empirical foundation for theorising emergent configurations of welfare digitalisation and contributes to ongoing debates on platform governance and the reconfiguration of welfare states.

3.1. Research design and rationale

This study adopts an exploratory, inductive, and descriptive research design to investigate the global landscape of digital welfare platforms. Such an approach is particularly appropriate given the relative paucity of empirical studies in the field and the absence of widely accepted definitions of these platforms. The heterogeneity of digital infrastructures operating within the welfare domain further underscores the need for an open-ended analytical framework (Alexopoulou, Reference Alexopoulou2025). As such, rather than aiming for statistical representativeness, the study seeks to capture the breadth of emerging platform configurations, offering an empirically grounded basis for conceptual clarification and future theory-building. The research design prioritises analytical openness and iterative categorisation, allowing analytical criteria to evolve as new cases emerge. This methodological approach enables the development of a data-driven typology reflective of actual platform diversity and intended to identify significant patterns within this rapidly evolving domain. The resultant dataset of 285 validated cases offers a novel empirical basis for theorising emerging configurations of digital welfare governance. While primarily descriptive, the analysis incorporates a comparative lens across countries, institutional settings, and welfare domains, thereby offering a systematic foundation for future theoretical and normative inquiry into welfare digitalisation. This design involves trade-offs: it captures recurring platform configurations, but does not account for welfare services that are not accessible through online search, including informal or locally embedded initiatives. As a consequence, the methodological approach favours platforms with a clear and stable online presence and may overlook services that are less visible or operate through fragmented or non-digital settings.

3.2. Data collection and cleaning

Given the lack of centralised registries or standard taxonomies for digital welfare platforms, the study employed a multi-stage open web-based search strategy conducted over a 12-month period (January 2024–January 2025). This method aligns with recent scholarship on digital public services, which has emphasised the necessity of open-source discovery in under-regulated or fragmented policy domains (OECD, 2016; Urs et al., Reference Urs, Špaček and Nõmmik2025). Search queries were developed in English and tailored to reflect major welfare domains, including healthcare (medical and mental health), elderly care, disability services, youth policy (education and training), family and housing assistance, and general local citizenship services. In cases where platform content was not available in English, online translation tools were used to support interpretation and screening. For each target group, two queries were run (combined with country terms such as ‘Australia’, ‘Brazil’, ‘Italy’, ‘South Africa’) via Google, open data catalogues, and official directories, as follows:

• • ‘digital platform for local citizens’ AND ‘digital welfare platform for local citizens’

• • ‘digital platform for disabled people’ AND ‘digital welfare platform for disabled people’

• • ‘digital platform for elderly people’ AND ‘digital welfare platform for elderly people’

• • ‘digital platform for young people’ AND ‘digital welfare platform for young people’

• • ‘digital platform for patients’ AND ‘digital welfare platform for patients’

• • ‘digital platform for mental health patients’ AND ‘digital welfare platform for mental health patients’

The search process followed an iterative and adaptive logic aimed at achieving broad coverage of the phenomenon. For each query, at least the first five pages of search results were screened. Beyond this threshold, the search was extended when results were still relevant to the study scope. The stopping criterion was based on a qualitative assessment of relevance: searches were discontinued when additional results no longer yielded platforms consistent with the analytical focus and instead returned content outside the domain of digital welfare provision. In more information-rich contexts, the search was intentionally prolonged to ensure that no analytically distinct cases were emerging, approximating qualitative saturation in terms of empirical coverage, rather than theoretical saturation. The overall strategy combined structured query design with iterative expansion. While individual search strings were not modified during each execution, insights from completed searches informed subsequent query design, including the use of alternative terms, synonyms, and recombination of keywords, which enabled the identification of additional platforms that might not have been captured through a single formulation.

Platforms were included only if they met all of the following criteria:

• • Focus on welfare-related services or collective enhancement of welfare ecosystems; AND

• • Operational digital interface (i.e., website/app); AND

• • Identifiable ownership (i.e., public, for-profit, not-for-profit, or combinations); AND

• • Sufficient public information to enable classification.

Platforms were excluded if they were:

• • Prototypes/inactive platforms; OR

• • Internal/admin tools without a user-facing interface; OR

• • No digital access channel; OR

• • Purely commercial services unrelated to social welfare.

Out of 440 initially identified platforms, 155 were excluded during the data validation and cleaning phase – particularly those with static or non-functional digital presences. The final dataset comprises 285 fully validated platforms. All entries were manually reviewed using a standardised data collection template encompassing ten variables distributed across seven analytical dimensions: geographical scope, functional purpose, target population, ownership model, delivery channel, pricing structure, and reporting logic. Three dimensions (i.e., ownership, delivery, pricing) were each disaggregated into two sub-dimensions to improve granularity and coding precision. Classification criteria for these dimensions were based on the analytical framework and initially developed through discussion with a second researcher. They were subsequently refined during the data collection process, as described in the coding procedure below.

3.3. Coding protocol and variable design

Coding was conducted by a single researcher using a standardised protocol applied consistently across all cases. Each platform was coded across all variables before proceeding to the next case, ensuring completeness and comparability at the level of individual observations. While the analytical dimensions and corresponding variables were defined ex ante, the internal categorisation within each dimension was not fixed a priori. Instead, category definitions were refined through an iterative and inductive process aimed at capturing the empirical diversity of the cases. This process involved multiple cycles of coding and recoding. When the analysis of new cases led to the identification of additional categories or to the refinement of existing ones, previously coded platforms were systematically revisited and reclassified in light of the updated scheme. This recursive procedure ensured that category definitions were progressively stabilised and consistently applied across the dataset. Particular attention was paid to maintaining internal coherence, both by aligning similar cases within the same categories and by preserving meaningful distinctions between cases with subtle but analytically relevant differences. Ambiguous or borderline cases – where classification across one or more variables was not immediately clear – were identified throughout the coding process and subjected to a detailed analysis. These cases were independently reviewed by a second researcher and subsequently discussed between the researchers, leading to a consolidated final classification. Beyond these targeted assessments, the second researcher was involved in reviewing the overall structure and consistency of the final dataset following the data cleaning phase, providing an additional layer of qualitative validation. Given the exploratory and inductive nature of the study, emphasis was placed on ensuring internal consistency, transparency of the coding process, and iterative refinement of category definitions. Robustness was thus pursued through the combination of a structured coding protocol, iterative recoding, and researcher triangulation. The resulting classification should therefore be interpreted as an empirically grounded description of the identified cases, designed to offer a coherent representation rather than statistical generalizability. To enable systematic analysis, the study developed a coding protocol structured around seven core analytical dimensions. Each dimension captures a critical aspect of platform architecture and governance, while categorical distinctions within each dimension ensure comparability across cases. The first analytical dimension is geographic scope, distinguishing where platforms operate, with ‘international’ assigned to cross-national access. VPN tools were used to resolve ambiguous cases. The second dimension is functional purpose, reflecting the main role of the platform in mediating interactions:

• • Service delivery: enables direct access to welfare services, such as booking care or making payments, often integrating providers.

• • Information repository: curates structured information about welfare services, supporting navigation without direct service provision.

• • Search engine: offers dynamic tools (i.e., filters, maps, keywords) to connect users with external providers.

• • Value co-creation: engages citizens in collaborative processes (e.g., reporting barriers or co-designing initiatives).

• • Community-building: fosters peer support, social cohesion, and mutual recognition among users with shared needs or conditions.

The third dimension is target population, identifying the intended user group:

• • General population: broad access to services or information.

• • Disabled people: tailored support ensuring accessibility and inclusion.

• • Elderly people: platforms addressing autonomy, care, and social participation.

• • Young people: initiatives for education, training, and youth services.

• • Medical patients: tools supporting continuity of care and treatment management.

• • Mental health patients: services providing timely, confidential psychological support.

• • Mixed or multiple targets: modular infrastructures serving diverse groups.

The fourth dimension is ownership model, distinguishing governance and service roles:

• • Public sector: ownership or provision by government authorities.

• • For-profit sector: platforms managed by private firms or start-ups.

• • Not-for-profit sector: actors such as NGOs or foundations.

• • Multi-stakeholder governance: partnerships combining different sectors.

The fifth dimension is delivery channel, referring to how users access and consume services:

• • Digital only: fully online access and service use.

• • Hybrid: combination of digital and physical channels.

• • Physical only: access limited to physical interaction, despite the presence of a supporting digital layer.

The sixth dimension is pricing structure, distinguishing financial models:

• • Free of charge: no fees for access or use, often publicly funded.

• • Freemium: basic services free, with paid premium options.

• • Fee-paying: full payment required for access or consumption.

The seventh dimension is reporting logic, assessing accountability practices:

• • No reporting: absence of data or evaluative content.

• • Basic display of results: limited metrics or success stories.

• • Performance evaluation: structured output/outcome reporting.

• • Social impact reporting: evidence of broader societal contributions.

3.4. Analytical principles and theoretical grounding

All variables were treated as categorical (i.e., nominal), with each platform assigned a single, mutually exclusive value per dimension. The classification framework followed the MECE principle (Mutually Exclusive, Collectively Exhaustive) to prevent overlap and ensure analytical coherence. While not intended to establish a definitive classification of the field, the typology captures the empirical diversity of platform configurations through an exploratory approach. The seven analytical dimensions were derived from theoretical and methodological considerations. Dimensions such as functional purpose, ownership structure, and delivery channel are grounded in the literature on e-government and digital-era governance (Margetts and Dunleavy, Reference Margetts and Dunleavy2013; Wimmer and Tambouris, Reference Wimmer, Tambouris and Traunmüller2002). Ownership and service provision distinctions reflect the logic of the mixed economy of welfare (Powell, Reference Powell2019), while the target population variable operationalises concerns around user fragmentation and public value co-creation (Janowski et al., Reference Janowski, Estevez and Baguma2018; Osborne et al., Reference Osborne, Radnor and Nasi2013). The reporting dimension draws on emerging debates on transparency and accountability in digital public services (Leone de Castris, Reference Leone de Castris2025). The coding scheme was refined throughout data collection to align with the study’s exploratory and inductive orientation.