The Map Isn't Finished: Inside Europe's Street-Level Data Gap

Point-of-interest (POI) verification is the process, manual or automated, of confirming that a business, address, or map feature listed in a geodata layer actually exists, remains operational, and sits at the coordinates assigned to it. It sounds like a solved problem in 2026, given the volume of satellite imagery, crowdsourced contributions, and AI-generated map features now available. It is not solved, and a campaign currently running across Europe is a useful illustration of why.

Between August 10 and September 30, 2026, HERE Technologies and Mapillary are running "Complete The Map 2026," a joint crowdsourcing push asking contributors across European cities to capture fresh street-level imagery. It is a collaborative effort between HERE Maps and Mapillary aimed at increasing and enhancing street-level imagery, expanding the number of cities to inspire more of their joint communities to improve street imagery locally. The organizers are direct about the motivation: many areas across Europe still have outdated or missing coverage, gaps that affect routing accuracy, points-of-interest discovery, and overall map quality.

That admission, from two of the largest mapping platforms operating in Europe, is worth sitting with. If HERE and Mapillary still need volunteer photographers to patch coverage gaps in 2026, the assumption that global map data is simply "there" and current no longer holds. This post looks at what's driving the gap, what the current research says about closing it with AI, and where field verification still does work that no dataset can.

Why Is HERE Paying People to Photograph Their Own Cities?

Street-level imagery underwrites almost every downstream map product: routing engines, POI directories, delivery-navigation apps, and the machine-learning pipelines that extract building footprints and address points from raw photos. When that imagery goes stale, everything built on top of it inherits the error. The Complete The Map campaign is HERE and Mapillary's answer to exactly that decay, incentivizing contributors with gift cards and hardware prizes to re-shoot cities where coverage has aged out or never existed at street level in the first place.

The campaign's own framing is instructive. By contributing captures, participants directly improve both OpenStreetMap and HERE Maps, enhancing navigation, planning, and exploration for millions of users worldwide. That single sentence quietly confirms two things: that OpenStreetMap and a commercial navigation giant are drawing on the same volunteer imagery pool, and that neither considers its European coverage complete enough to stop asking for help. For a market this mature, that's a notable admission rather than a routine community-engagement push.

What Does "Geodata Freshness" Actually Mean?

Geodata freshness refers to how closely a mapped feature's recorded state, whether that's a business's operating status, a road's connectivity, or a building's footprint, matches its real-world condition at the moment someone queries it. Freshness and coverage are not the same metric: a map layer can be geographically complete and still be wrong about half its POI attributes because nobody re-checked them after the initial capture. Points of interest refers to places of interest frequently visited by human traffic throughout the day, including restaurants, supermarkets, transportation hubs, parks, cafes, and tourist attractions , and every one of those categories changes state constantly: businesses close, relocate, change hours, or get repurposed into something else entirely.

Industry analysis on POI accuracy makes the same point from a different angle. The best available source of location intelligence in 2026 will very likely not be from one source alone, particularly when it comes to mapping points of interest across dynamic geographies. Crowdsourced layers like OpenStreetMap fill gaps official registries miss, but some types of points of interest may be underrepresented, and inconsistencies can also arise at the regional level due to fragmented governance, while funding and policy decisions might affect the pace at which data are released or revised. The Europe OpenStreetMap extract itself demonstrates the scale-versus-freshness tension directly: it runs to a 32.6 GB file, and the version pulled for this piece was last modified 13 hours ago and contained all OSM data up to 2026-09-18T20:21:10Z . High-frequency updates at the file level say nothing about whether any individual POI record inside it has been checked against reality in the past year.

Why Do Public and Crowdsourced Datasets Fall Short for Address and POI Verification?

Consider a grocery delivery platform operating across the Netherlands and Belgium. Its routing engine pulls addresses and pickup-point coordinates from a blended feed of OSM extracts and licensed commercial data. A depot listed as active in the feed closed three months earlier when the operator consolidated warehouses, but nobody flagged the change upstream, so drivers keep getting routed to a shuttered address until enough failed deliveries force a manual audit. That failure mode is not a bug in the dataset's coverage; it's a gap in verification cadence, and no amount of additional crowdsourced imagery fixes it on its own, because imagery shows a building exists, not whether the business inside it still does.

This is the same freshness-versus-verification problem Teampl has run into directly. In a nationwide field program covering more than 1,500 EV charging stations across the Netherlands, Teampl recorded station name, operator, address, number of charging points, connector types, charging capacity, and pricing as a structured dataset, with on-site photographic evidence attached to every location, spanning urban, suburban, and rural areas rather than sampling major cities only. That kind of ground-truthed infrastructure record is exactly what a routing or navigation feed needs to stay accurate between crowdsourcing waves, and it is the same discipline behind continuous business-status checks described in The Mystery Shop That Never Clocks Out. The decay problem also isn't unique to outdoor POIs; the same debt accumulates indoors, as covered in Mapping Debt: The Bill Every Mall Eventually Pays. Field verification is not a stylistic choice for these programs; it is a structural requirement for keeping any large geodata layer usable in production.

Can AI Close the Gap Automatically?

Automated quality assessment is the obvious next step, and 2026 research shows real progress alongside real limits. A June 2026 paper introduces Topo4Vec, a GeoAI framework built specifically to catch topological errors in vector geodata, the kind of overlapping polygons and broken street connections that plague large-scale map layers. Geospatial vector data quality is a foundational research topic in GIS, yet classic rule-based quality assessment algorithms often struggle with diverse urban morphologies and massive data volumes. The framework, developed partly under a German Research Foundation (DFG) project explicitly named "AI4OpenGeodataQuality: A Toolkit and Framework for Continuous Large-Scale Quality Measurement of Open Geodata with Efficient AI and Multimodal Remote Sensing," tested its method across Los Angeles, Munich, and Singapore.

The results are the interesting part. Topo4Vec achieved a peak accuracy of 0.99 for detecting overlapping building footprints and 0.60 for overshoots and undershoots in street networks. A near-perfect score on building overlaps next to a coin-flip-adjacent score on street connectivity errors tells you exactly where automated GIS quality control is mature and where it still isn't. Street topology errors, the kind that directly break routing and delivery navigation, remain the hardest category for a model to catch reliably. Automated vector quality checks are a genuine advance, but they are not yet a substitute for a human or a field team confirming that a road segment, a pickup point, or a business listing matches what's actually on the ground.

What's Driving Europe's GIS Market Growth?

None of this is happening against a shrinking market. The Europe GIS market was valued at USD 3.0 billion in 2025 and is projected to reach USD 5.0 billion by 2034, exhibiting a CAGR of 5.53% during 2026-2034 , according to IMARC Group. A separate forecast from Market Data Forecast puts the 2025 figure slightly higher, at USD 3.11 billion, projected to reach USD 5.34 billion by 2034 from USD 3.30 billion in 2026, at a CAGR of 6.20% . The modest divergence between two independent forecasts is itself a signal: this market's segmentation and measurement methodology are still settling, which tracks with an industry where data provenance and quality standards are also still being formalized.

Germany anchors the region. Germany currently dominates the Europe GIS market in 2025 , and growth broadly is driven by increasing demand for spatial data analytics in transportation and environmental management, alongside accelerating adoption of cloud-based GIS platforms and real-time data integration . Municipal investment adds another layer: growing smart city programs designed to streamline urban planning and resource management are pushing municipalities and governments toward more sophisticated spatial analytics . More software licenses and more cloud platforms mean more places to store and query geodata. They do not automatically mean the records inside those platforms are current, correct, or field-checked.

The Field Layer Is Still Human

Software growth, crowdsourced imagery campaigns, and GeoAI quality frameworks are all solving real, distinct problems: capacity, coverage, and structural error detection, respectively. None of them solves attribute-level truth on their own. A building footprint model can score 0.99 on overlap detection and still have no idea whether the bakery inside that footprint closed last month. That is the layer field verification exists to cover, and it's a complement to the automated stack, not a replacement for it.

Public datasets get you a map. They do not get you a map you can trust for routing a delivery van or listing a charging station's live connector count. That gap is not a footnote. It is the actual product.

Your next fix probably is not another dataset subscription. It's a field team with a clipboard, a camera, and a QA process behind them. Want the operational detail behind how programs like this actually get run — recruiting, quality control, chain of custody? That's what we do day to day.

Frequently Asked Questions