A single pair of coordinates looks precise. Six decimal places suggest a location accurate to roughly ten centimetres. So why does a delivery driver, handed exactly those coordinates, so often end up standing on the wrong side of a building with a parcel they cannot drop?
The answer is that precision and accuracy are not the same thing. A rooftop geocode can be precise to the centimetre and still point at the wrong place: the middle of a roof rather than the door you walk through. For a detached house that gap is a few metres and nobody notices. For a hospital, an apartment block, an airport, or a distribution park, the gap between the rooftop pin and the real entrance can be the difference between an on-time drop and a failed delivery.
This is the quiet shift happening across location data right now. The industry is moving from "where is this building" to "where do I actually go". Here is what entrance-level geocoding is, why last-mile teams care, and what to look for in an API.
The Problem With the Rooftop Pin
Classic geocoding turns an address into one point. That point is usually the centroid of a building outline or a position interpolated along the street. Both are reasonable averages, and both ignore the single most operationally important fact about a building: where you get in.
Consider a few everyday cases:
- A warehouse with the office entrance at the front and the goods-in dock at the rear, reached by a separate service road.
- An apartment block where the addressed entrance faces a courtyard, not the street the centroid snaps to.
- A campus or shopping centre where the rooftop point lands on a roof in the middle of a 300-metre-wide complex.
In each case the rooftop pin is precise and wrong. A driver routed to it arrives near the building, then loses two to five minutes finding the actual door. Multiply that across a 120-stop route and the rooftop pin has quietly cost an hour. Worse, when the entrance is genuinely hard to reach, the stop gets marked as failed and the parcel goes back to the depot.
What Entrance-Level Geocoding Returns
Entrance-level geocoding answers a richer question. Instead of one point, a precise response can include:
- The display point: where to drop a marker on a map.
- One or more entrance points: the tagged doors of the building, sometimes split into main and service entrances.
- A navigation point: a position on the drivable road network where a vehicle should stop to reach that entrance.
- The building outline: the footprint polygon, so you can show the building shape and reason about its size.
The navigation point is the one that changes operations. Routing to the door is not enough if the door is in a pedestrian zone; you want the nearest point a van can actually stop. A good geocoder derives that road-side point from the entrance and the surrounding street geometry, so the driver is sent to a place they can park and walk in.
Why This Matters Beyond Delivery
Last-mile delivery is the obvious case, but entrance precision shows up wherever someone has to physically arrive:
- Field service and home healthcare: a technician or nurse with a tight schedule cannot afford to circle a building.
- Ride-hailing and mobility: pickups at large venues work far better when the app sends the car to a named entrance rather than the building middle.
- Accessibility: routing a wheelchair user to a step-free entrance is only possible if the data distinguishes entrances at all.
- Emergency and logistics planning: knowing the building outline and access points supports safer, faster on-site decisions.
In every one of these, the cost of the rooftop pin is paid in minutes and frustration, repeated thousands of times a day.
Where the Data Comes From
Entrance and outline data is not magic, and it is not locked inside one vendor. OpenStreetMap contributors have mapped building footprints and tagged individual entrances on a vast and growing share of the world's buildings, marking main doors, service doors, and access points. Because the map is open, a geocoding service built on it can return that structure directly: the outline, the entrances, and a navigation point derived from them.
Coverage is uneven, as you would expect from a map edited by people. It is densest in cities and around large public buildings, and it improves every day as the map is corrected and extended. The practical takeaway is to treat precision as a spectrum: a good API tells you how confident it is, so your routing logic can fall back gracefully when only a rooftop point exists.
What to Look For in a Geocoding API
If entrance-level precision matters to your product, evaluate a geocoding API on more than its hit rate:
- Structured geometry, not just lat and lon. Does the response carry an entrance or navigation point and a building outline, or only a single coordinate?
- A precision or match-type field. You need to know whether a result is rooftop, interpolated, or entrance-grade so your code can react.
- A confidence score. Graceful fallback depends on knowing when the precise answer is not available.
- Honest coverage. Ask where the data is strong and how it is maintained, rather than assuming uniform global precision.
- Data residency. If you geocode customer addresses, where that processing happens is a compliance question, not just a technical one.
That last point is easy to overlook. An address is personal data. Sending European customers' addresses to be geocoded outside the EU turns a routing decision into a data-transfer decision.
The MapAtlas Approach
The MapAtlas Geocoding API is built on OpenStreetMap, so it can return building outlines and entrance-derived navigation points where the open map provides them, alongside a clear precision signal you can route on. It runs on EU-hosted infrastructure with no transfer of personal data outside the EU, which means you can chase entrance-level accuracy for European addresses without trading away GDPR compliance to get it.
The rooftop pin had a good run. It was the right abstraction when maps were for looking at, not for arriving at. Now that a coordinate is an instruction to a driver, a rider, or a robot, the question has changed from "where is this building" to "where do I actually go". Geocoding is finally catching up to it.
Ready to route to the door instead of the roof? Explore the MapAtlas Geocoding API or see how address validation cleans the input first.
Frequently Asked Questions
What is building-entrance geocoding?
Building-entrance geocoding returns the coordinates of the door a person actually uses to enter a building, plus a navigation point on the road where a vehicle should stop, instead of a single pin on the building's rooftop or centroid. For a large complex such as a hospital, campus, or shopping centre, the entrance can sit tens or hundreds of metres from the rooftop point, which is the gap that causes failed deliveries and wrong drop-offs.
Why does the rooftop pin fail for last-mile delivery?
A rooftop or centroid pin marks the middle of a building outline, not the way in. For a detached house the difference is small, but for warehouses, apartment blocks, airports, and campuses the real entrance can be on the far side of the building or down a service road. A driver routed to the centroid arrives near the building but not at the door, loses minutes hunting for the entrance, and sometimes marks the stop as failed.
Where does entrance and building data come from?
OpenStreetMap contributors map building outlines and tag individual entrances, including main and service doors, on millions of buildings worldwide. A geocoding API built on OpenStreetMap can return the building outline, the tagged entrance points, and a road-side navigation point derived from them. Coverage is strongest in dense urban areas and improves continuously as the open map is edited.
How do I get entrance-level precision from a geocoding API?
Use a geocoding API that exposes a precision or match-type field and returns structured geometry, not just a single lat and lon. Check whether the response includes an entrance or navigation point, a building outline, and a confidence score, then route your vehicle to the navigation point rather than the display point. The MapAtlas Geocoding API is built on OpenStreetMap and is EU-hosted, so you get precise European address data without transferring personal data outside the EU.
