POI Data Explained: How Points of Interest Get Collected and Verified

A point of interest, or POI, is a single mapped location, such as a shop, ATM, or clinic, stored with coordinates and descriptive attributes so software can find, display, and route to it. That's the textbook version. Here's the version that matters in practice.

Two bakeries sit fifty metres apart on the same street in Lyon. One closed eight months ago. Every delivery app on your phone still routes riders to its door most Tuesday evenings, because nobody told the map it shut. The other bakery, still very much open, doesn't show up in search at all — the owner never claimed the listing, and no crawler ever walked past with a camera.

That gap between what a database says and what's actually standing on a street corner is the whole POI problem, compressed into one block. This piece walks through how POI data actually gets built, who's responsible for it across the EU, why it decays faster than most people assume, and what real verification looks like once you get past the dashboard.

What Actually Counts as a POI?

The Open Geospatial Consortium, the body that writes interoperability standards for mapping software, keeps its definition deliberately loose. In the broadest terms, a POI is a location about which information of general interest is available. That's it. No requirement for a street address, no minimum set of attributes.

In practice, that flexibility spans a huge range. A POI can be as simple as a set of coordinates and an identifier, or as complex as a three-dimensional building model with names in several languages, opening hours, and a full civic address. A hiking trailhead with no street number is a POI. So is a multinational retailer's flagship store with forty attributes attached to it.

This is also why the OGC bothered publishing a formal model at all. POI data has traditionally been exchanged in proprietary formats through different transport mechanisms, and the OGC's Points of Interest activity focuses on a standard model meant to make POI data interoperable across platforms. Before that kind of standardization, a "restaurant" in one vendor's schema might not map cleanly onto a "restaurant" in another's — which sounds trivial until you're trying to merge two datasets covering the same city.

Worth separating early: a POI is not the same thing as an address. Plenty of POIs sit mid-block with no formal address at all — a market stall, a bus shelter, a trailhead. And plenty of valid addresses, like a residential apartment, aren't POIs because nobody outside the household has a reason to care where they are.

Four Ways POI Data Gets Built

Ask five GIS teams where their POI data comes from and you'll usually get some blend of four sources, each with a different failure mode.

Authoritative government registries pull from business licenses, cadastral records, and permits. They're accurate at the moment of registration and then age exactly as fast as the bureaucracy behind them updates. Crowdsourced platforms like OpenStreetMap rely on volunteers editing a shared map. Coverage is famously uneven — dense in cities with active mapping communities, thin everywhere else. Commercial aggregators such as Google, HERE, or Overture Maps Foundation crawl the web, ingest business listings, and license data from multiple upstream providers. Overture's open dataset alone now lists nearly 60 million places worldwide , though it isn't easy to download for GIS users, since it ships in GeoParquet format rather than a simple spreadsheet. And physical field verification means someone, or some team, actually goes to the location and confirms what's there.

None of these is complete on its own. That's the part vendors don't lead with.

Method	Best for	Where it stops
Government registry	Legal identity, ownership, official address	Lags real-world change; doesn't confirm the business is still trading
Crowdsourced mapping (OSM)	Dense urban coverage, free access, fast community fixes	Coverage drops sharply outside active mapper communities
Commercial aggregation (Google, HERE, Overture)	Broad scale, rich attributes, frequent automated refresh	Confidence varies by category; low-traffic POIs get checked less often
Physical field verification	Ground truth on status, category, and condition	Cost and time scale with headcount, not with dataset size

Europe's Address Backbone: One Continent, Several Systems

If you're building or buying POI data for the EU, the first thing to understand is that there's no single European address authority. There's a framework that requires interoperability, and then a patchwork of national systems underneath it.

That framework is the INSPIRE Directive (2007/2/EC), which entered into force in May 2007 to build an infrastructure for spatial information in Europe supporting environmental and related policy, and addresses 34 spatial data themes through technical implementing rules . More than 7,000 European public institutions now contribute data under it. But it's a plumbing standard, not a data-collection mandate. The legislation does not require the collection of new spatial data — it just says that if you have it, you have to make it discoverable and interoperable.

Underneath that, each country runs its own address infrastructure, and the differences are real. Our related piece on Europe's street-level mapping gap covers how uneven this coverage gets once you leave city centres.

Country	Address authority / system	Best for	Where it stops
France	Base Adresse Nationale (BAN), piloted by DINUM and IGN	Open, centralized, API-accessible; recognized as the official reference address database of the state	Confirms an address exists, not whether a business still trades there
Germany	Federated cadastral offices (AdV / ALKIS), Land by Land	Cadastral-grade parcel and building precision	Sixteen separate regional systems means format and update-cycle inconsistency
Switzerland	Federal Register of Buildings and Dwellings (GWR), Federal Statistical Office	Tight linkage between buildings and official statistics	Cantonal permitting means updates lag on the ground
Belgium	Regional systems (CRAB, URBIS) feeding a federal BeST address register	Harmonization underway across regions	Historically split by region, so legacy inconsistencies persist
Netherlands	BAG, run through Kadaster	Strong cadastre-to-address integration	Struggles with multi-tenant sites like malls, where one address covers dozens of businesses

Outside the EU, the picture is different again. The UK runs address data through Ordnance Survey's national mapping infrastructure. The US has no equivalent central authority at all — POI and address data there gets assembled from a mix of county assessor records, city GIS departments, and the postal service, which is part of why US commercial POI vendors lean so heavily on crowdsourcing and crawling.

Why POI Data Goes Stale So Fast

Here's the uncomfortable number underneath all of this: roughly 20% of new businesses fail within their first year, according to U.S. Bureau of Labor Statistics data . That figure only counts closures — it doesn't include businesses that change phone numbers, shift hours, or simply move . Multiply that churn across every retail category, every year, and a "complete" POI dataset starts looking like a snapshot that's already out of date the moment it's finished.

Consider a national pharmacy chain opening forty new branches a year while quietly closing a dozen underperforming ones. Each opening takes weeks to appear correctly across major map platforms. Each closure can linger, uncorrected, for months — long enough for delivery drivers, patients, and analysts to keep treating a dead location as live. Location-data providers try to counter this by continuously updating building footprints, addresses, and classifications from authoritative geographic and municipal sources , but mitigating that decay requires ongoing verification, not a single cleanup pass .

The financial stakes aren't abstract, either. Analysts at Gartner estimate that poor data quality costs the average organization somewhere between $12.9 million and $15 million every year , and roughly 60% of companies don't even measure that financial impact — so the cost sits, unassigned, on nobody's budget line. Bad POI data isn't a cosmetic problem. It's a line item, whether or not anyone's tracking it. We've written about the same dynamic playing out inside physical retail spaces in Mapping Debt: The Bill Every Mall Eventually Pays — the bill for outdated location data always comes due, just later than expected.

How Verification Actually Happens in the Field

Ground truthing refers to physically confirming that what a database claims about a location matches reality, usually through a site visit, a photograph, or a direct measurement. Everything else — registry checks, satellite imagery, mobile signal analysis — is an attempt to approximate ground truth without paying for a human to go and look.

Each proxy has a blind spot. Satellite and street-level imagery catches new construction and demolished buildings but misses a shop that changed ownership and kept the same signage. Mobile signal and foot-traffic data can flag that a location has gone quiet, but not why — a closed store and a slow Tuesday look identical to a phone. Crowdsourced corrections are fast when they happen but wildly uneven in coverage, concentrated wherever an engaged local community already exists.

For instance, Teampl has run business-status verification programs photographing hundreds of points of interest per day for clients who needed to know, definitively, which locations were actually open. Photographic ground truth removes the ambiguity that imagery and signal data leave behind — a picture of a locked shutter and a "for lease" sign settles the question in a way that a stale foot-traffic graph never quite does. This is the same logic behind continuous, human-verified checks we cover in The Mystery Shop That Never Clocks Out: a status check that happens once a year tells you less than you think, because the interesting failures happen in the months between checks.

Comparing Verification Methods

No serious POI program relies on just one of these. But knowing where each one is strong, and where it quietly fails, determines how you combine them.

Method	Best for	Where it stops
Registry cross-check	Confirming legal existence, ownership, and official address	Registries update on bureaucratic time, not real-world time
Crowdsourced correction	Fast fixes in actively mapped areas, at zero direct cost	Uneven coverage; silence isn't the same as accuracy
Imagery / remote sensing	Detecting demolition, new construction, large physical change	Blind to ownership, category, and status changes without visible signage
Mobile signal / foot traffic	Flagging likely closures at scale, cheaply	Tells you activity dropped, not why, or what replaced it
Physical field verification	Definitive open/closed status, category accuracy, photographic evidence	Costs scale with headcount and POI count, so it suits hundreds or thousands of locations better than millions

That last limitation is real, and it's worth being honest about it. Field verification is the most reliable single method and the least scalable one. Nobody is sending a photographer to every POI on a continent every month. The organizations that get this right use field verification selectively — on the highest-churn categories, the highest-value locations, or the records other methods have already flagged as suspicious.

Building a Pipeline That Doesn't Rot

The mistake most teams make with POI data is treating it as a purchase instead of a subscription. A dataset bought once and never re-checked starts decaying the moment it lands, at a rate roughly tracking the underlying business churn already covered above. The fix isn't more data. It's a cadence.

A workable pattern looks something like this: authoritative registries and commercial feeds as the baseline layer, crowdsourced corrections as a free early-warning signal, and scheduled physical verification concentrated on the categories that churn fastest — restaurants and retail move far more often than clinics, schools, or public infrastructure. Field checks get triggered by anomalies: a location with no recent reviews, a foot-traffic graph that flatlined, a registry entry with no matching commercial listing.

Why does the combination matter more than any single source? Because each method's blind spot is another method's strength. A registry won't tell you a shop closed last month. A foot-traffic graph won't tell you it reopened as something else. Only a person standing in front of it, camera in hand, resolves both questions at once — which is precisely the gap physical field verification is built to close.

Start a Project

Teampl runs field data collection programs — egocentric video, in-store audits, mystery visits, GIS surveys — designed around a specific research question, not a generic panel. If this topic touches a program you're planning, tell us what you're trying to learn.

Frequently Asked Questions