Cycle route networks are increasingly recognised as a critical component of national and regional mobility strategies. The 2025 UNECE Guide for Designating Cycle Route Networks [1] (UNECE, 2025) underscores the importance of coherent, hierarchically integrated cycling infrastructure (from international backbones such as EuroVelo down to local networks) capable of serving both everyday commuters, leisure and tourism users. Crucially, the [1] UNECE Guide (2025) makes explicit that realising this vision is inseparable from data: it calls for assessments to be data-driven, drawing on multiple sources to evaluate the appropriateness of existing roads and alternative corridors for cycling. Against this background, OpenStreetMap has the potential to be a significant source of cycling network data for exactly this purpose, offering global coverage, granular way-level attributes, and a relational structure capable of representing network hierarchies from international to local scale.

Yet OSM data quality is known to be spatially and thematically uneven. Previous literature has consistently demonstrated that OSM quality varies significantly across regions, feature classes, and mapping communities [2-5] (Hakley, 2010; Hochmair et al. 2015; Guth et al., 2021; Neis et al., 2013). In the context of cycling infrastructure, this unevenness implies that the capacity to assess network quality is itself spatially and thematically uneven, varying across regions and attribute types [6-7] (Vierø et al., 2024; Ferster et al., 2020). Consequently, any assessment of cycling network quality is inherently constrained by where data are more complete or sparse, and by how consistently key attributes are recorded.

Adopting this perspective requires that quality assessment be both spatially disaggregated, examining variation across the territory rather than relying on global aggregates, and thematically targeted, focusing on the attributes and network properties that are most relevant to the analytical task.This is particularly important in the context of cycle route planning, where requirements concerning safety, continuity, comfort, directness and infrastructure separation vary across user groups, route hierarchies and expected traffic volumes, making uniform or purely aggregate assessments insufficient for meaningful evaluation and decision-making [1] (UNECE, 2025).

However, doing so at scale is not straightforward. Much of the information relevant to cycling network quality (infrastructure type, surface, legal access conditions) is stored at the way level but only meaningful when understood through bicycle relations, which define the network hierarchies that planning and certification frameworks operate on. Conducting a country-level, relation-aware quality assessment has historically required substantial preprocessing that constrained both the scale and depth of such work [8-9] (Deri et al., 2015; Pruvost et al., 2017). This paper demonstrates that this constraint can be overcome.

The analysis uses the ohsome planet tool developed by HeiGIT, which converts the full OpenStreetMap planet into an analysis-ready Parquet dataset while preserving its complete relational structure [10] (HeiGIT, 2025). Combined with DuckDB for fast and scalable querying, this setup enables a fully reproducible analysis pipeline for the entire Dutch cycling network. The dataset covers all levels of the network hierarchy, including 22 international route relations that form the backbone of the system, as well as 118 national, 16,964 regional and 264 local route relations that support mobility. Overall, the analysis includes 183,147 ways representing approximately 39,455 km of roads and 17,368 route relations (Figure 1).

Central to the paper is a two-phase analytical framework that makes the relationship between OSM data quality and the assessment of cycling network quality explicit.

In the first phase, the focus is entirely on what the OSM data itself can and cannot support. Tag completeness is measured across four attributes central to cycling infrastructure evaluation: surface type, lighting, width, and maximum speed (Figure 2). The internal consistency of OSM tagging conventions is tested. Topological connectivity is assessed using NetworkX across each network hierarchy, examining the degree to which relations form coherent, traversable networks rather than fragmented collections of ways (Figure 3).

The second phase proceeds only where the first establishes that the data permits it. Where tags are sufficiently present and internally consistent, infrastructure attributes are evaluated against established external standards, specifically the [1] UNECE Guide (2025) and the EuroVelo European Certification Standard [11] (ECF, 2021). These provide concrete, operationalisable criteria: minimum width thresholds differentiated by route hierarchy, surface quality tiers appropriate to infrastructure type, and the relationship between posted speed limits and the level of physical separation required to make cycling conditions acceptable. Grounding the second phase in normative standards rather than data-derived benchmarks produces assessments that are externally meaningful rather than self-referential.

Together, the two phases allow the paper to make statements not only about patterns of OSM data quality across the Netherlands, but about the quality of the bicycle network itself, where the data supports such a conclusion. In doing so, the paper foregrounds the role of bicycle relations as the structural backbone that links way-level attributes to network-level meaning, and without which neither the completeness assessment nor the evaluation against external standards would be possible at a meaningful scale.