Address to Coordinates
Convert a street address to latitude and longitude. Worldwide forward geocoding via Mapbox, with confidence bands surfaced so you can pick the right match when results are ambiguous.
How to use this tool
Type the address
Enter as much detail as you have: street number, street name, city, region, and postcode if you have it. As you type past two characters, an autocomplete dropdown will show up to five live suggestions from Mapbox. You can pick one with the arrow keys + Enter, click it, or finish typing and submit the form.
Read the hero coordinates
The top of the result panel shows a high-impact "These are your address coordinates" card with the decimal-degree pair in monospace and a "Copy" button. Below the hero, the same point is broken down into all six common notations (DD, DMS, DDM, UTM, MGRS, Plus Code) with their own copy buttons.
Inspect alternative candidates
When Mapbox returns more than one nearby address — common for ambiguous queries like "350 5th Avenue, New York" — the alternatives are listed below the hero with their own confidence bands. Click any candidate card to make it the active selection; the hero, the map pin, and the copyable coordinates all update together.
What “address to coordinates” actually does
Forward geocoding converts a free-form postal address into a geographic coordinate pair. It is the inverse of reverse geocoding (coordinate → address) and a separate operation from point-of-interest search (place name → coordinate).
A forward geocoder never invents a coordinate. It parses the input, tokenizes it against an address hierarchy, searches a reference dataset, ranks the matches, and returns the best candidates with a confidence assessment for each.
Per ISO 19112:2019, an address is a hierarchical locator: country → administrative area → locality → thoroughfare → number → sub-building. The Mapbox response carries that hierarchy in its context object.
The reverse of that hierarchy is the precision ladder below — the level at which the geocoder can actually place your query.
Precision tiers — what the result actually points at
| Tier | Typical accuracy | Example query |
|---|---|---|
| Rooftop | ≤ 5 m | 350 5th Avenue, Apt 4B, New York, NY 10118 |
| Parcel | 5 – 30 m | 350 5th Avenue, New York, NY 10118 |
| Street centerline | 30 – 200 m | 5th Avenue, New York |
| Postal code | 0.5 – 5 km | 10118, USA |
| Locality | 1 – 50 km | New York City |
| Region | 50 km or more | New York State |
Rooftop precision means the dataset has the individual building footprint — the gold standard, available for verified delivery addresses in well-mapped urban areas.
Parcel precision is one tier down: the dataset knows the property but not the exact door. The returned coordinate points at the front of the parcel or its centroid.
Street centerlineprecision means Mapbox interpolated along the road segment using the house-number range. The result is on the right street, perhaps 30–200 m from the actual building.
The lower tiers fall back to postcode, locality, and region centroids when the query is missing detail.
How accurate is forward geocoding?
Mapbox reports a match_code.confidence field on every forward-geocoded feature: high, medium, or low.
The chart below shows the typical horizontal error for each band, on a log scale because the spread is two orders of magnitude.
| Band | What Mapbox is saying | Typical horizontal offset | Trust it for… |
|---|---|---|---|
| high | The query unambiguously matches one address in the dataset. | 5 – 15 m | Routing, billing, dispatch, automated workflows |
| medium | Query is partially specified or matches multiple records. | 20 – 150 m | Human review with alternatives presented — never auto-select |
| low | Dataset is sparse near the query or the query is very vague. | 100 m – 5 km | Discovery / context only — do not feed downstream systems |
What forward geocoding is not
Not a place-name lookup.The Mapbox v6 forward endpoint is designed for addresses, not points of interest. Typing “Empire State Building” returns Empire State Trail — a hiking path 200 km north of Manhattan.
The right query is the building's street address: 350 5th Avenue, New York, NY 10118.
Not address validation.Validation means “is this a deliverable address per USPS / Royal Mail / La Poste?” — a different operation governed by USPS Publication 28 and provided by carriers like USPS CASS or vendors (Stripe Radar, Lob, Smarty).
A geocoder will happily return coordinates for an undeliverable address; only a validator will flag that the unit number doesn't exist.
Ten ways forward geocoding gets used in production
Anywhere a user types an address into a form, forward geocoding is probably running underneath.
The ten cases below cover the bulk of the traffic against commercial geocoders. Each is paired with a worked example using a real address you can paste into the tool above.
1. E-commerce checkout (shipping cost & tax)
Shopify, Amazon, Stripe Checkout, BigCommerce, and every modern shopping cart geocode the buyer's shipping address at checkout.
The coordinate drives shipping-cost computation (via carrier zone), tax-jurisdiction lookup, and the “estimated delivery” ETA. Many carts also display a pin on the confirmation page.
Worked example: Buyer types 350 5th Avenue, New York, NY 10118. The geocoder returns 40.748400, −73.985700(Empire State Building lobby, Manhattan) with high confidence. The cart reads UPS zone 2 from that coordinate, computes shipping cost, and matches the buyer to New York state and city sales tax rates.
2. Last-mile delivery routing
DoorDash, Instacart, Amazon Logistics, Uber Eats, and traditional carriers (UPS, FedEx, DHL) geocode the delivery address at order placement.
The coordinate feeds route-optimization algorithms; the driver app navigates to it; the GPS capture at drop-off becomes the proof-of-delivery (see the inverse flow under /tools/coordinates-to-address).
Worked example: Customer orders groceries to 1 Apple Park Way, Cupertino, CA 95014. The geocoder returns 37.334900, −122.009100 with high confidence. The dispatcher uses that coordinate to assign the order to the courier closest to the destination at the time of pickup, then routes via the captured coordinate rather than the address text (which is more ambiguous than the coords).
3. Real-estate listing onboarding (MLS feeds)
When a listing enters the MLS, the agent enters the property address. Zillow, Redfin, Realtor.com, and aggregators forward-geocode that address at ingest.
The coordinate drops a map pin, renders “nearby” comparables, and feeds school / walkability overlays. Address-level precision matters because adjacent parcels can carry very different listing prices.
Worked example: A new listing posts at 1600 Pennsylvania Avenue NW, Washington, DC 20500. The geocoder returns 38.897700, −77.036500(the White House) with high confidence. The aggregator uses the coordinate to compute distance to the nearest Metro stop and pulls comparable sales within a 0.5 km radius for the listing's “neighborhood” section.
4. Marketing radius and catchment targeting
Local-business advertisers on Meta, Google, and TikTok geocode their store address, then target ads to users whose home location sits within a configurable radius.
The same pattern powers “catchment area” analytics: retail chains geocode stores and customer addresses, then compute drive-time isochrones to see where each store's customers come from.
Worked example:A coffee chain's flagship branch sits at 1912 Pike Place, Seattle, WA 98101. The geocoder returns 47.609100, −122.342500. The marketing platform restricts ad delivery to phones whose home location falls within a 5-mile catchment of that coordinate.
5. Property insurance underwriting
Every property-insurance quote starts with geocoding the property address.
Underwriters overlay the coordinate against FEMA flood-zone polygons, wildfire risk maps, earthquake hazard tiles, and crime statistics by census tract. The resulting risk profile drives the premium.
Small geocoding errors matter here: a parcel 30 m on the wrong side of a Special Flood Hazard Area line can mean a quote that is thousands of dollars off.
Worked example: A homeowner requests a quote for 2025 Center St, Berkeley, CA 94704. The geocoder returns 37.870000, −122.270900. The underwriter checks the coordinate against the USGS Earthquake Hazard map (the Hayward Fault runs about 3 km east) and FEMA Zone X (no significant flood risk). The quote reflects high seismic exposure but low flood exposure.
6. Banking address verification & fraud detection
Issuers and acquirers (Visa, Stripe Radar, Square) geocode the billing address at checkout, then compare it to the device's IP-derived location and historical patterns.
A billing address in Tokyo with a card swiped on an IP that geo-resolves to Lagos triggers a fraud flag. The tolerable radius is configurable per merchant.
Worked example: A cardholder enters 10 Downing Street, London SW1A 2AA, United Kingdom. The geocoder returns 51.503400, −0.127500. The risk engine confirms the IP geolocation resolves to London (within a tolerable ~50 km radius) and waves the transaction through.
7. US Census tract assignment (federal & municipal stats)
The US Census Bureau Geocoder geocodes survey responses and administrative records, assigning each household to a census tract, block group, and block.
The downstream effect is broad: every “by neighborhood” statistic published by the Census, the CDC, HUD, and most US municipalities relies on that tract assignment.
The Census Geocoder publishes its methodology openly and is the de-facto standard for US address → tract conversion.
Worked example: A 2020 Census response carries the address 400 N St NW, Washington, DC 20001. The Census Geocoder returns 38.907700, −77.013600, which falls in Census Tract 47.02. Every aggregated statistic for that respondent is reported at the tract level for privacy reasons.
8. Healthcare and epidemiology (CDC, JCAH, ICD)
Public-health agencies geocode patient addresses to map disease incidence, identify case clusters, and target interventions.
Privacy laws constrain the precision at which results are published. The HIPAA Safe Harbor de-identification standard typically prohibits finer-grained geography than the first three ZIP digits (the “rule of small numbers”), but geocoding still happens internally for cluster detection.
Worked example: A state health department receives a positive disease report for a patient at 1600 Clifton Rd, Atlanta, GA 30329 (CDC headquarters). The coordinate 33.798000, −84.323000is stored privately; for the public dashboard the case is rolled up to ZIP 30329 with a 7-day rolling window.
9. CRM sales-territory assignment
Sales tools (Salesforce, HubSpot, Pipedrive) geocode lead addresses on entry so accounts can be routed to the territory-owning sales rep automatically.
Territory boundaries are usually polygons — state lines, ZIP-code regions, custom polygons drawn by an admin. The geocoder is the bridge from a typed address to the right polygon.
Worked example: A new lead enters Salesforce with 500 Brickell Key Dr, Miami, FL 33131. The geocoder returns 25.766400, −80.187700. The lead-router checks that coordinate against the “South Florida” territory polygon and assigns the lead to the Miami account executive.
10. Field-service installation routing
Solar installers, EV-charger contractors, smart-meter techs, and cable-company crews geocode every service order's address to schedule technicians, batch jobs by neighbourhood, and detect work orders that land outside the contractor's service area.
Worked example: A solar installer receives a new job at 3500 Deer Creek Rd, Palo Alto, CA 94304. The geocoder returns 37.413100, −122.165000. The dispatcher matches that coordinate against the licensed-service polygon (Bay Area), batches the job with three nearby installations already on the schedule, and sends the route to the truck.
The “wrong borough” problem
The most famous failure mode of forward geocoding is the wrong-borough match.
“350 5th Avenue, New York” has at least two real-world matches: 350 5th Avenue in Manhattan (the Empire State Building, 40.7484° N, −73.9857° W) and 350 5th Avenue in Brooklyn (40.6754° N, −73.9863° W). They sit about 8 km apart.
Without a borough name or postcode, Mapbox returns both with medium confidence. Any geocoder that auto-selected the “top” result would deliver to the wrong place roughly half the time.
The fix is editorial discipline at the application layer: always request limit=5, always read match_code.confidence, and never auto-select a low or medium result.
Coordinately's /methodology § 19.1 codifies this rule. The other major geocoders (Google, HERE, OpenCage, AWS Location Service) all expose comparable disambiguation signals.
Choosing the right tool from the geocoding family
| Operation | Input | Output | Right tool on Coordinately |
|---|---|---|---|
| Forward geocoding (this tool) | Address string | Lat / lon + confidence | This page |
| Reverse geocoding | Lat / lon | Nearest known address | /tools/coordinates-to-address |
| ZIP / postal-code lookup | Postal code | Centroid lat / lon | /tools/zip-to-coordinates |
| Coordinate format conversion | Lat / lon in any of 6 notations | Lat / lon in all 6 notations | /tools/coordinate-converter |
| POI / place-name search | Place name ("Eiffel Tower") | Lat / lon of the place | Not yet — requires a separate API |
| Address validation (deliverability) | Address string | Boolean + standardised form | Not on Coordinately — use USPS CASS, Lob, or Smarty |
Why a result is sometimes wrong
Six recurring failure modes:
- Ambiguous query.“350 5th Avenue, New York” matches Manhattan and Brooklyn. Adding a borough, postcode, or country resolves to a single high-confidence match.
- Stale dataset. New construction can take weeks to months to enter Mapbox / OSM. Queries for a brand-new building may resolve to the previous parcel.
- Cross-country ambiguity.“Smolnik 62” matches villages in Slovakia, Slovenia, Poland, and Hungary. The autocomplete dropdown is the disambiguation tool — adding the country to the input always resolves it.
- Sub-building precision.“350 5th Avenue, Apt 4B” usually returns the building lobby coordinate. Unit numbers are rarely in geocoder datasets; for unit-level dispatch you need address validation, not geocoding.
- Vanity-versus-USPS spelling.“1 Apple Park Way” works; “Apple Headquarters” does not. Geocoders parse the postal form, not the marketing form.
- Non-Latin scripts. Mapbox v6 handles non-Latin addresses (Cyrillic, Arabic, CJK) reasonably well, but accuracy is highest when the address is supplied in the script and language used by the local postal authority. For Tokyo, prefer the Japanese form on local datasets; transliterations work but with reduced confidence.
Privacy and data-flow notes
Coordinately runs forward geocoding server-side. Your address (and every autocomplete keystroke after a 250 ms debounce) is sent to Mapbox; Mapbox returns ranked candidates; the page renders them and discards the response.
Per Mapbox's terms of service the response cannot be cached or retained, so Coordinately stores nothing. The full data-flow description is at /privacy-policy.
For bulk forward-geocoding of a large address list, the right pattern is to hit Mapbox directly with your own access token rather than scraping this tool — faster, cleaner, and the rate limits are aligned to bulk use.
Related tools
- Coordinates to address— The inverse — reverse geocoding for a lat/lon pair
- ZIP / postal code to coordinates— Postal-code centroid lookup, for less specific queries
- Coordinate format converter— Render the resolved point in all six common notations
- Distance between two coordinates— Geodesic distance once you have two geocoded addresses
- Elevation by coordinates— Height above sea level for the resolved location
Related articles
- What is geocoding?— Forward vs reverse, datasets, accuracy levels
- Forward versus reverse geocoding— When each direction is the right operation
- Geocoding accuracy levels— Rooftop, parcel, street, ZIP — what each tier really means
- Mapbox vs Google vs HERE geocoding— How the major providers compare on accuracy and coverage
- Coordinate formats explained— The six notations the result panel displays
Frequently asked questions
Why are there multiple candidates for one address?
Address geocoding is inherently ambiguous. The string "350 5th Avenue, New York" matches two real addresses — one in Manhattan and one in Brooklyn, about 8 km apart. Mapbox returns both (and any other matches) with confidence bands so the user, not the algorithm, picks the intended one. Adding context (a ZIP code, a borough name) usually resolves the ambiguity to a single high-confidence match.
What does the confidence band mean?
Mapbox classifies each forward-geocoded result as "high", "medium", or "low" confidence. High means Mapbox is confident the candidate is the right address for the query. Medium and low indicate Mapbox is less sure — usually because the query is incomplete (no city, no region) or because multiple real addresses match. When the top result is medium or low, this tool shows all candidates up front rather than auto-selecting; auto-selecting on low confidence is the canonical failure mode of address geocoders.
Why doesn't "Empire State Building" return the Empire State Building?
The Mapbox Geocoding API is an address geocoder, not a points-of-interest search. Place names — "Empire State Building", "Eiffel Tower", "Central Park" — are not addresses and may match unrelated entries (the literal query "Empire State Building" returns Empire State Trail, a hiking path 200 km north of the actual building). For POI lookups you need a place-search API, which Coordinately may add as a separate tool later. For now, look up the street address ("350 5th Avenue, New York, NY 10118") instead.
Adding a ZIP code improved the result. Why?
Postcodes disambiguate. "350 5th Avenue" matches two real addresses without further context; "350 5th Avenue, New York, NY 10118" narrows the postcode area to a single block in Manhattan, and Mapbox returns the Empire State Building lobby with high confidence. As a rule of thumb, providing street + city + region + postcode raises confidence to "high" for almost any deliverable address.
Can I forward-geocode in a specific country?
The /api/geocode/suggest endpoint accepts an optional `country=` parameter — comma-separated ISO 3166-1 alpha-2 codes — that narrows results to those countries. The UI doesn't expose a country picker because the autocomplete dropdown itself is the disambiguation tool: typing the country name into the address (e.g. "smolnik 62, slovakia") nudges Mapbox toward the right match without an extra control.
Does Coordinately store the addresses I look up?
No. Per Mapbox's terms of service and our methodology, queries and responses are server-rendered to your browser and not cached, logged, or persisted on our side. The autocomplete dropdown hits a server-side proxy that sets Cache-Control: no-store. See /privacy-policy for the full data-flow description and /methodology for the architectural rationale.
Why is the result different from what I get on Google Maps?
Different geocoding providers use different reference datasets. Mapbox draws from OpenStreetMap, the OpenAddresses project, and proprietary sources; Google uses its own Places dataset. The two sometimes disagree on the precise coordinates for a given address, especially in dense urban buildings (where the "address" can correspond to a lobby door, a rooftop, or a parcel centroid), in new developments, or in rural areas with sparse address data.
What about address validation — does this confirm an address is real?
No — geocoding and validation are different operations. A forward geocoder will happily return coordinates for an address that does not actually receive mail (an apartment that doesn't exist, a street that was renamed). For deliverability validation against the USPS database (or Royal Mail, La Poste, etc.) you need a validation API: USPS CASS, Lob, Smarty, Stripe Radar. Coordinately may add a validation tool in the future; for now, treat geocoding output as a coordinate, not as proof the address is deliverable.
Sources
- Mapbox Geocoding API v6 — Mapbox Geocoding API v6 reference documentation · https://docs.mapbox.com/api/search/geocoding-v6/ · Accessed .
- Mapbox Terms of Service — Mapbox Terms of Service — response retention restrictions for geocoding · https://www.mapbox.com/legal/tos · Accessed .
- OpenStreetMap — About OpenStreetMap — upstream contributor dataset used by Mapbox · https://www.openstreetmap.org/about · Accessed .
- OpenAddresses — OpenAddresses — open dataset of address points contributed by governments · https://openaddresses.io · Accessed .
- ISO 19112:2019 — ISO 19112:2019 — Geographic information — Referencing by identifiers · https://www.iso.org/standard/76496.html · Accessed .
- USPS Publication 28 — USPS Publication 28 — Postal Addressing Standards (canonical US address form) · https://about.usps.com/publications/pub28/welcome.htm · Accessed .
- US Census Bureau Geocoder — US Census Bureau Geocoder — federal forward-geocoding service and methodology · https://geocoding.geo.census.gov/geocoder/ · Accessed .
- US Census TIGER/Line — US Census Bureau TIGER/Line — authoritative US street centerlines used as ground truth · https://www.census.gov/geographies/mapping-files/time-series/geo/tiger-line-file.html · Accessed .
- HHS HIPAA Safe Harbor — HHS — HIPAA Privacy Rule, Safe Harbor de-identification (geographic constraints) · https://www.hhs.gov/hipaa/for-professionals/privacy/special-topics/de-identification/index.html · Accessed .
- FEMA National Flood Hazard Layer — FEMA National Flood Hazard Layer — flood-zone polygons referenced by underwriters after geocoding · https://www.fema.gov/flood-maps/national-flood-hazard-layer · Accessed .