Poland Has the Data. Now We Need to Make It Intelligent

Poland already has a substantial body of data describing its territory, economy and investment environment. For many industrial location problems, the main obstacle is no longer the absence of data. It is the fragmentation of sources, uneven data quality and the difficulty of combining information collected for different purposes. This is precisely where GeoAI can add value: by bringing machine learning, text analysis and other computational methods into geospatial decision support.

The European Commission’s 2025 Open Data Maturity Report places Poland among Europe’s leading countries in overall open-data maturity.[1] The report is not a ranking of national GIS resources, so it should not be used as one. Its result is nevertheless consistent with what a review of Polish public datasets suggests: there is already a broad base of spatial, statistical, economic and administrative information that can support investment decisions.

The rest of this article looks at how these resources can be combined and analysed to answer practical questions about industrial location.

What data are already available?

The table below summarises the main data families relevant to industrial real estate and site selection. The distinction between public and commercial sources matters. Professional analysis rarely relies on open data alone.

Group	Data	Data form	Public availability	Typical spatial detail	Main public sources	Commercial / directly acquired complement
Reference data	Cadastral parcels	vector	YES / PARTIAL	parcel	EGiB, Geoportal	full cadastral extracts, ownership information
	Buildings	vector / 3D	YES / PARTIAL	building	EGiB, BDOT10k, Geoportal	enriched property databases
	Administrative boundaries	vector	YES	administrative unit	PRG, GUGiK	—
	Address points	point	YES	address	PRG, EMUiA, GUGiK	HERE, TomTom, Google
	Terrain models	raster / point cloud	YES	c. 1 m and locally finer	NMT, NMPT, LiDAR, GUGiK	commissioned surveys and photogrammetry
	Orthophotos	raster	YES	commonly 5–10 cm, locally finer	GUGiK, Geoportal	commissioned aerial surveys
	Land cover and use	vector / raster	YES / PARTIAL	object / 10 m for Sentinel-2	BDOT10k, Sentinel	thematic commercial data
	Road and rail networks	network / vector	YES / PARTIAL	network segment	BDOT10k, OSM, infrastructure managers	HERE, TomTom, PTV
Planning and legal status	Local zoning plans	vector / document	YES / PARTIAL	parcel / planning zone	Urban Register, BIP, municipalities	aggregated planning databases
	General plans and spatial policy	vector / document	YES / PARTIAL	planning zone / municipality	Urban Register, BIP	planning databases
	Property ownership	register	RESTRICTED	property	EGiB and EKW within legal access rules	legal due diligence
	Cadastral information	vector / table	PARTIAL	parcel / building	EGiB, Geoportal	official extracts
	Permits and administrative decisions	table / document / location	YES / PARTIAL	project / parcel	RWDZ, BIP	development databases
	Environmental constraints	vector / document	YES	parcel / zone	GDOŚ, Wody Polskie, PIG-PIB, GIOŚ	specialist studies
	Protected areas	vector	YES	area / parcel	GDOŚ	—
	Easements and other property rights	register / document	RESTRICTED	property	land and mortgage registers where the register number is known	legal due diligence
	Development restrictions	derived data	PARTIAL	parcel	planning data, GDOŚ, Wody Polskie, PIG-PIB, heritage registers	development due diligence
Transport and logistics	Access to major roads	network / vector	YES	parcel–junction / road segment	GDDKiA, BDOT10k, OSM	HERE, TomTom, PTV
	Travel times	network / time series	PARTIAL	origin–destination pair	OSM, GDDKiA, local ITS	HERE, TomTom, fleet data
	Rail terminals and sidings	point / line	PARTIAL	facility	UTK, PKP PLK, OSM	logistics databases
	Ports	point / object	PARTIAL	port / terminal	port authorities, GUS, OSM	logistics databases
	Airports	point / object	YES	airport	ULC, BDOT10k, OSM	—
	Intermodal terminals	point	PARTIAL	terminal	UTK, OSM, operators	logistics databases
	Traffic and congestion	time series / network segments	PARTIAL	road segment	GDDKiA, local ITS	TomTom, HERE
	HGV restrictions	network / attribute	PARTIAL	road segment	road authorities, OSM	truck-routing systems
	Traffic counts	table / point / segment	YES	point / road segment	GDDKiA	—
	Route cost and path	derived data	PARTIAL	origin–destination pair	road networks + public toll data	PTV, HERE, TomTom
	Supplier and customer accessibility	derived / network	PARTIAL	company–facility / route	business registers + transport networks	supply-chain databases
Utilities	Power networks and available capacity	network / table	PARTIAL	network segment / connection point	GESUT, PSE, utilities	direct utility enquiries
	Substations	point / object	PARTIAL	facility	PSE, GESUT, OSM	utilities
	Gas networks	network	PARTIAL	network segment	GESUT, GAZ-SYSTEM	utilities
	Water networks	network	PARTIAL	network segment	GESUT, municipal utilities	utilities
	Wastewater networks	network	PARTIAL	network segment	GESUT, municipal utilities	utilities
	Telecommunications	network / point / address	PARTIAL	address / network segment	internet.gov.pl, SIDUSIS, GESUT	operators
	District heating	network	PARTIAL	network segment	GESUT, URE, local operators	operators
	Renewable-energy potential	raster / time series	YES / PARTIAL	raster / location	IMGW, public meteorological data	Solargis and similar services
Environment and site conditions	Elevation and slope	raster / point cloud	YES	c. 1 m	NMT, LiDAR, GUGiK	site surveys
	Geology	vector / profiles / boreholes	YES	depends on source	PIG-PIB, CBDG	geotechnical investigations
	Groundwater	vector / point	YES / PARTIAL	hydrogeological unit / borehole	PIG-PIB, Wody Polskie	local investigations
	Flood risk	vector / raster	YES	parcel / zone	Wody Polskie	—
	Soil conditions	vector / map	PARTIAL	depends on source	IUNG, PIG-PIB	geotechnical investigations
	Soil contamination	point / area / document	PARTIAL	property / area	GDOŚ	environmental site assessments
	Noise	raster / contours	PARTIAL	local	GIOŚ, municipalities	measurements
	Air quality	point / raster / time series	YES	station / modelled raster	GIOŚ	—
	Natura 2000 and other protected areas	vector	YES	area	GDOŚ	—
	Climate and heat risk	raster / time series	YES / PARTIAL	raster / region	IMGW, public climate models	specialist risk databases
Population and labour	Population	table / grid	YES	statistical grid / municipality	GUS	—
	Labour supply	table	YES / PARTIAL	municipality / county	GUS	commercial labour analytics
	Unemployment	table	YES	county / region	GUS, labour offices	—
	Wages	table	YES / PARTIAL	county / region	GUS	salary surveys
	Skills and occupations	table	PARTIAL	region / county	GUS, education and labour sources	recruitment databases and employer surveys
	Commuting	flow matrix	YES, periodically	municipality–municipality	GUS, census	mobile-network data
	Demographic structure	table / grid	YES	grid / municipality	GUS	—
	Education	table / point	YES / PARTIAL	institution / territorial unit	GUS, education registers	—
	Migration	table	YES / PARTIAL	municipality / county / region	GUS	—
Industrial real-estate market	Existing industrial and logistics stock	point / object / table	PARTIAL	building / park	EGiB, BDOT10k, OSM, market reports	REDD, brokers
	Warehouse locations	point / object	PARTIAL	building	OSM, BDOT10k, company websites	REDD, brokers
	Building areas	vector / table	PARTIAL	building	EGiB, building geometry	owners, brokers
	Vacancy	table	NO at property level	building / unit	aggregate market reports	REDD, brokers
	Rents	table	PARTIAL	market / property	market reports	brokers, landlords
	Transaction prices	table / object	YES / expanding	transaction / property	RCN	Cenatorium, MSCI, brokers
	Land asking prices	table / point	PARTIAL	listing / parcel	listing portals	portal analytics, brokers
	Development pipeline	table / document / location	PARTIAL	project / parcel	RWDZ, company announcements	REDD, brokers
	Take-up and absorption	table	NO at property level	market / park	aggregate reports	brokers
	Developers and owners	table / relationship graph	PARTIAL	entity / property	KRS, CEIDG, company disclosures	commercial databases
	Occupiers and tenants	table / point	PARTIAL	company / property	company announcements and websites	leasing and company databases
Business geography	Company registers	table	YES	legal entity	CEIDG, REGON, KRS	D&B, EMIS, Coface and others
	Establishment locations	point / address	PARTIAL	establishment / address	REGON, CEIDG, company websites	commercial company databases
	Industry classifications	table	YES	company	PKD, REGON, CEIDG, KRS	—
	Company employment	table	PARTIAL	company / territorial unit	aggregated GUS data	company databases
	Financial information	table / document	PARTIAL	company	KRS, company reports	EMIS, D&B and others
	Suppliers	relationship graph / table	RESTRICTED	company relationship	public procurement, company disclosures	supply-chain databases
	Customers	relationship graph / table	RESTRICTED	company relationship	company disclosures	supply-chain databases
	Industrial clusters	derived data	PARTIAL	cluster / region	company registers + spatial data	commercial databases
Earth observation	Sentinel	raster / time series	YES	10–60 m depending on band	Copernicus Data Space	—
	Landsat	raster / time series	YES	commonly 30 m	USGS / Landsat	—
	High-resolution commercial satellite imagery	raster	NO	sub-metre possible	—	Maxar, Planet, Airbus, ICEYE
	Aerial imagery	raster	PARTIAL	centimetres	GUGiK orthophotos	commissioned flights
	LiDAR	point cloud	YES	point / object	GUGiK	commissioned surveys
	Satellite SAR	raster / time series	YES / PARTIAL	sensor dependent	Sentinel-1	ICEYE, Capella
Movement of people and goods	Mobile-network mobility data	flows / time series	NO at useful detail	aggregated zones	—	mobile operators
	GPS traces	trajectories	RESTRICTED	device / vehicle	limited research datasets	fleets, telematics
	Floating-car / observed vehicle movement data	time series / network	RESTRICTED	road segment	local ITS	TomTom, HERE
	Public-transport usage	table / time series	PARTIAL	line / stop	municipalities, GTFS, GUS	operators
	Footfall	point / time series	PARTIAL	point / zone	local counts	mobile operators, retail analytics
	Commuting	flow matrix	YES, periodically	municipality–municipality	GUS	telecom data
	Freight movement	table / flows	PARTIAL	corridor / region	GUS, roads, rail, ports	logistics and shipment data
Web and volunteered data	OpenStreetMap	vector / network	YES	object	OSM	—
	Business directories	table / point	PARTIAL	company / address	CEIDG, REGON	commercial directories
	Websites	text / document	PARTIAL	page / entity	public web	specialist databases
	Real-estate listings	text / table / point	visible, usually not open in bulk	listing	listing portals	portal analytics
	POIs	point / vector	YES / PARTIAL	object	OSM, Wikidata	HERE, TomTom, Foursquare, Google
	Crowdsourced data	various	licence dependent	object	OSM, Wikidata	—
Location-linked documents	Planning documents	text / vector	YES / PARTIAL	document / area	Urban Register, BIP	planning databases
	Environmental decisions	document / location	PARTIAL	project / parcel	Ekoportal, OOŚ database, RDOŚ, BIP	development databases
	Permits	document / location	YES / PARTIAL	project / parcel	RWDZ	development databases
	Municipal strategies	document	PARTIAL	municipality	BIP, municipal websites	—
	Infrastructure plans	document / map	PARTIAL	region / network	PSE, GDDKiA, PKP PLK, GAZ-SYSTEM	utilities
	Company announcements	text / document	PARTIAL	company / project	ESPI, EBI, company websites	business-information databases
	Investment and market reports	document	PARTIAL	market / region	PAIH, broker reports	paid reports and databases
	Tender documentation	document	YES	procedure / project	e-Zamówienia, BZP, TED	—
Company data	Plant locations	point / object	NO for internal data	plant	—	company records
	Supplier and customer database	table / relationship graph	NO	company / relationship	—	ERP, CRM, SRM
	Shipments and transport	table / trajectory	NO	shipment / route	—	TMS, WMS
	Production	time series / table	NO	line / machine / job	—	MES, SCADA
	Workforce and organisation	table	NO	employee / team / plant	—	HRIS
	Energy use	time series	NO	meter / device / building	—	EMS, BMS
	Operational KPIs	time series / table	NO	process / line / plant	—	company systems
	CAPEX	table	NO	project / asset	—	financial systems
	Maintenance	table / time series	NO	asset	—	CMMS, EAM
Building and operations	Building geometry and information model	BIM / 3D	NO for private assets	building element	public 3D models only as context	BIM model supplied by investor or occupier
	Room and equipment locations	BIM / point / object	NO	room / asset	—	BIM, as-built documentation
	Asset registers	table / BIM	NO	asset	—	EAM, CMMS, BIM
	Energy meters	time series	NO	meter / asset	—	BMS, EMS, IoT
	Environmental sensors	time series	NO	sensor / zone	—	BMS, IoT
	Production sensors	time series	NO	machine / line	—	MES, SCADA, IoT
	Movement inside the facility	trajectories / events	NO	person / vehicle / zone	—	RFID, UWB, internal systems

The strongest part of Poland’s public-data base concerns the location and characteristics of physical objects: parcels, buildings, roads, terrain, environmental constraints and administrative divisions. As analysis moves closer to an actual business decision and plant operation, commercial and proprietary information becomes more important. Utility availability, lease terms, supply-chain relationships, freight flows and operational data often have to be obtained from companies, infrastructure operators, municipalities, landlords or specialist providers.

This is not a uniquely Polish weakness. Detailed information on actual utility capacity, rents, supply chains, shipment flows or plant operations is commercially sensitive or proprietary in most markets.

From maps to decision models

The same datasets can support very different levels of analysis. The simplest question is: what is located where? The next step is to combine layers and compare alternatives against explicit criteria. More advanced methods can identify recurring patterns, estimate values or classes, analyse relationships in networks and extract information automatically from documents.

In their systematic review of GeoAI, Siqin Wang and colleagues distinguish raster imagery, vector data, spatiotemporal flows, official statistics and text among the main data types used in the field. These data are used for tasks such as classification, prediction, simulation, automated feature extraction and geoparsing.[2]

This progression builds on established spatial decision-support methods. Jacek Malczewski described GIS-based multicriteria decision analysis as a way of combining geographical information with the evaluation of alternatives in order to produce information for decision-making.[3] GeoAI extends that toolbox with methods capable of learning patterns from large and heterogeneous datasets.

In their review of 149 industrial site-selection studies, Dongbo Wang and colleagues show that research is increasingly combining multicriteria decision-making, GIS and artificial intelligence.[4] One direction is the integration of imagery, text, vector data and sensor observations within a single analytical workflow. Another is the use of knowledge graphs to represent relationships among companies, sites and infrastructure, and graph neural networks to learn from those relationships. A further strand of research focuses on explainability: not only producing a ranking or prediction, but also showing why a model preferred one location over another.

One decision spans several spatial scales

Industrial location decisions should be assessed across several scales, from the regional context down to the individual property and its surroundings.

At the regional level, the main issues are labour availability, access to customers and suppliers, transport connections and local market conditions. At the parcel level, the analysis shifts to development rights, planning constraints, road access, utilities, terrain, environmental restrictions and room for future expansion. If an existing building is under consideration, the analysis must also determine whether its geometry, clear height, structural grid, building services and internal layout can accommodate the intended production process.

Some questions necessarily cut across these scales. “Where should we locate the new plant?” requires a regional comparison first and a property-level assessment second. An even broader decision is “Should we build a new facility, acquire an existing one or expand the current plant?”. That requires a simultaneous comparison of market and logistics conditions, parcel development potential and utility access, and the adaptability of existing buildings.

Location analysis, understood in this way, goes beyond producing a suitability map. Its purpose is to compare real investment alternatives.

Polish applications already exist

Marcin Hernes and colleagues provide a useful Polish example of combining real-estate information with spatial variables and machine learning. Their study used 148 transactions involving investment land in Wrocław. Transaction records were supplemented with information on zoning, development intensity, infrastructure, road access, public transport, distances, flood hazard and surrounding nuisances.[5]

The authors compared multiple regression, random forest and an artificial neural network. In this particular sample, the neural network achieved a mean absolute percentage error (MAPE) of 13%, compared with 22.2% for multiple regression and 26.3% for random forest.[6] This should not be interpreted as evidence that neural networks will always outperform other methods. The more important point is that combining market and spatial variables produced a useful analytical model, while the limited number of transactions remained a major constraint.[7]

The question for Poland is therefore no longer whether such analysis is technically possible. It is how to move from isolated experiments to repeatable workflows that combine public, commercial and company data.

Not only the West and China

Quantitative and spatial methods for facility location are not developing only in the United States, Western Europe and China. Serena Fugaro and Antonino Sgalambro reviewed 288 papers on multi-objective facility-location problems published between 2011 and 2025. Among the 143 studies that used case studies, Iran accounted for 23.0% and China for 22.7%, followed by Turkey and North America.[8]

The industrial site-selection literature also includes specific Iranian examples, such as wind-farm siting in Semnan Province and industrial site-selection studies in Isfahan.[9]

These figures should not be read as evidence that Iran has more operational deployments than Poland. Fugaro and Sgalambro note that many case studies are used to validate methods on realistic data rather than to document tools that have been implemented in day-to-day practice.[8] The comparison is nevertheless revealing: serious work on computational location analysis is being carried out across a much wider range of countries than the usual focus on the largest technology centres might suggest.

The data are here. The next task is to use them better.

Poland does not need to wait for a new generation of spatial-data infrastructure before developing GeoAI applications. Much of the necessary foundation already exists. The priority is to connect sources reliably, control data quality and build analytical workflows that translate those resources into answers to real investment questions.

GeoAI is one part of a broader decision-support toolbox. Depending on the problem, GIS, multicriteria analysis, operations research, statistics, network analysis, text processing and conventional modelling will remain equally important. The objective is not to apply artificial intelligence for its own sake. It is to make better decisions by combining the right data with the right methods.

The literature shows that this approach is developing not only in the United States, Western Europe and China, but also in countries such as Iran. The question is therefore no longer whether GeoAI can be applied to industrial site selection. The question is why Poland — with such a substantial public-data foundation already in place — should not become one of the countries where these methods are developed and used in practice.

---

References

1. European Commission / data.europa.eu, 2025 Open Data Maturity Report. ↩
2. Wang, S. et al., “Mapping the landscape and roadmap of geospatial artificial intelligence (GeoAI) in quantitative human geography: An extensive systematic review”, International Journal of Applied Earth Observation and Geoinformation, 128 (2024), 103734, p. 5. ↩
3. Malczewski, J., “GIS-based multicriteria decision analysis: a survey of the literature”, International Journal of Geographical Information Science, 20(7) (2006), 703–726, pp. 703–704. ↩
4. Wang, D., Zhu, Y., Mao, X., Wang, J., Ji, X., “Industrial Site Selection: Methodologies, Advances and Challenges”, Applied Sciences, 15 (2025), 11379, pp. 22–26. ↩
5. Hernes, M., Tutak, P., Nadolny, M., Mazurek, A., “Real estate valuation using machine learning”, Procedia Computer Science, 246 (2024), 4592–4599, p. 4596. ↩
6. Hernes, M. et al., ibid., pp. 4596–4597. ↩
7. Hernes, M. et al., ibid., p. 4598. ↩
8. Fugaro, S., Sgalambro, A., “Multi-Objective Facility Location Problems: Systematic Literature Review and research agenda”, Computers & Industrial Engineering, 216 (2026), 111994, p. 7 and the discussion of case-study use later in the paper. ↩
9. Yousefi, H., Motlagh, S.G., Montazeri, M., “Multi-Criteria Decision-Making System for Wind Farm Site-Selection Using Geographic Information System (GIS): Case Study of Semnan Province, Iran”, Sustainability, 14 (2022), 7640; Reisi, M., Afzali, A., Aye, L., “Applications of analytical hierarchy process (AHP) and analytical network process (ANP) for industrial site selections in Isfahan, Iran”, Environmental Earth Sciences, 77 (2018), 537. ↩

Abbreviations used in the table

BIM — Building Information Modeling: a digital representation of building geometry and associated technical information.

BMS — Building Management System: a system used to monitor and control building services such as heating, ventilation, cooling, lighting and energy consumption.

IoT — Internet of Things: connected devices and sensors that automatically collect and transmit data.

MES — Manufacturing Execution System: software used to monitor and manage production execution at plant level.

SCADA — Supervisory Control and Data Acquisition: systems used to monitor, control and collect data from industrial processes.