There are two types of geocoding: forward and reverse. Forward geocoding converts addresses, such as 1600 Pennsylvania Avenue, Washington, D.C. 20500, into geographic coordinates (i.e., latitude and longitude). Reverse geocoding converts a coordinate, such as (-77.03655, 38.89768), into the aforementioned address.

The following are examples of Forward Geocoding (converting an address into geographic coordinates) and Reverse Geocoding (converting geographic coordinates into an address):

Forward Geocoding: A delivery app needs to be able to place a customer's address on a map so the driver can deliver food there. So when a customer enters their address, they then send that address to our Geocoder API, which returns the coordinates of that address, along with other metadata.

‍Reverse Geocoding: A weather app customer needs to display a weather report to their end user based on their location. So when they retrieve the coordinates of the location from the device, the customer then sends them to our Geocoder API, which returns the address data. Then, they can use that address data to label the weather report with the correct city and neighborhood.

Temporary geocoding allows a forward or reverse geocoding search to be retrieved and displayed immediately. This is useful for short-lived mapping applications, like showing a user where their destination is on a map.

Mapbox also offers the unique ability for customers to permanently store address and coordinate data served via the Geocoding API. This means that customers use Permanent Geocoding to save costs and improve performance for scenarios in which the same data needs to be geocoded repeatedly.

For example, an in-browser BI company may need to repeatedly load a map of their end users’s warehouse locations. Without permanent geocoding, they may have to make 100’s of calls for the same addresses every time the user loads the page. However, with permanent geocoding, they can cache these addresses, improving performance of their app and saving on requests.

Mapbox allows both Temporary and Permanent geocoding results to be displayed on a map from any vendor – unlike Google there is no “lock in” that limits display of geocoding data to our maps.

Customers use forward geocoding to analyze and visualize address data. For example, a rideshare company would use forward geocoding to enable the user to search for his or her destination (as a human-readable address), then visualize that address on a map (using geographic coordinates).

Customers use reverse geocoding to determine the physical address of vehicles, packages, and people. For example, the same rideshare company noted above would use reverse geocoding to display the user’s current location on the map.

Additional use cases for forward and reverse geocoding include:

• Weather apps use reverse geocoding to determine what town a user is in, or a ride-sharing app may use it to display the closest address to a rider.
• Retail companies use reverse geocoding to map shop locations in their store locator.
• Business intelligence apps use forward and reverse geocoding to visualize and group data spatially, for tasks like real estate analysis.
• Ecommerce uses forward geocoding to simplify and speed up the checkout process, autofilling shipping addresses for delivery.
• Travel companies use forward geocoding for global search enabling end users to discover travel destinations – from countries to local neighborhoods.
• Content tagging is enabled by reverse geocoding, for photo apps that want to show end users where their shots were taken.

There are different options when using the Geocoding API.  You can choose (1) if you want to use results immediately or store for future use, and (2) if you want to work from a coordinate or from an address.

Alongside these options, the Geocoding API offers the following features:

• Improved in v6: Context Object - In addition to a geographic coordinate, get hierarchical data about an address, like neighborhood, postcode, region, and country.   Each place in the hierarchy has an attached ID, allowing subsequent lookup.
• Autocomplete - Implement typeahead functionality to suggest potential address to end users as they type.
• Proximity Bias - Prioritize geocoding results around a particular location, such as a device location or IP address.
• New in v6: Structured Input - Label address components in request to get better search performance.
• Improved in v6: Batch Geocoding - Geocode multiple points in a single request.   
• Intersection Search - Search for a point where two streets intersect, without a precise address
• New in v6: Smart Address Match - Confidence scores in the response object give a better sense of accuracy level and keyword matching.
• Improved in v6: Multiple Language Support - Geocoding searches are now supported in 40+ languages.

You can check our coverage map to understand where Geocoding is supported. We have particularly strong coverage across North America, Western Europe, Australia and Japan. However, please check back often, as we are continually expanding address coverage.

Geocoding API supports 40+ languages:Language coverage. In v6, we expanded our coverage to include: Estonian, Ukrainian, Italian, Russian, Malay, Vietnamese, and Hindi.

Mapbox’s Geocoding API has a number of advantages over other solutions:

• Conflation of 1000s of data sources: Our platform has created the leading geospatial data set by leveraging a wide range of partners and sources. We combine sources and rank them to provide the most accurate results across our ever-growing coverage area.
• Permanent Geocoding: We offer the ability to store results from our API permanently, delivering cost savings & performance potential, plus opening up new uses cases for reporting and geotagging.
• Corrections Pipeline: We have a large team dedicated to updating our data based on feedback and telemetry across 600M MAU’s, making over 700K new map updates daily. The pipeline nominates, compares, aggregates and canonicalises data to provide the most up-to-date and accurate results.
• Mapbox Ecosystem: Mapbox search integrates seamlessly with our Navigation and Maps products. The Mapbox Boundaries product also supplements geocoding, enabling aggregation by a wide variety of partitions.

Geocoding is billed per-request across 2 different SKUs, temporary geocoding & permanent geocoding.

• The Temporary Geocoding API is billed per 1000 requests (CPM) at $0.75 ~ $0.45, with further discounts available for the largest volumes.
• The Permanent Geocoding API is billed per 1000 requests (CPM) at $5 ~ $4, with further discounts available for the largest volumes.

See https://www.mapbox.com/pricing#search for more information.

New in v6 are features that benefit customers needing to efficiently run large volumes of geocoding:

• New in v6: Unit-level geocoding - geocode specific apartment units and business suites, including the ability to retrieve their unique coordinates if available.
• New in v6: Structured Input - Label address components in request to get better search performance.
• New in v6: Match Codes - Confidence scores in the response object give a better sense of accuracy level and keyword matching.
• Improved in v6: Batch Geocoding - Geocode multiple points in a single request.

In addition we have upgraded our coverage, API understandability, and usability:

• New in v6: Japan Addresses - Geocode addresses in Japan with industry-leading coverage, with 34 million new addresses.
• Improved in v6: Context Object - Get hierarchical data about an address, like neighborhood, postcode, region, and country.
• Improved in v6: Multiple Language Support - Geocoding searches supported in 40+ languages.

Structured Input is an optional parameter that enables you to pre-parse address data. Instead of querying a string like 1600%20Pennsylvania%20Avenue%20Washington,%20D.C.%2020500, you can query num=1600&street=Pennsylvania%20Avenue&city=Washington,%20D.C.&postcode=20500. Doing so increases query understanding and the likelihood of providing the correct result as fast as possible, with structured queries being processed up to 20% faster.

Please read the documentation for more information about Structured Input.

Smart Address Match consists of a Confidence Score and Match Codes.

The Confidence Score assesses how well a query and its individual components match a known result. To determine the Confidence Score, each component of the address is assessed and given a true or false match code. The match code results are then analyzed, providing an overall Confidence Score ranging across Exact, High, Medium, or Low Confidence.

The Confidence Score and Match Codes help fine-tune geocoding results to a particular use case. If you are working in delivery, for example, you can use match codes to ensure that you only use the most accurate results; if, on the other hand, you are analyzing and visualizing a dataset with millions of address records, it may be sufficient to aggregate data at the postcode level.