Major Geocoding Services Compared

The /learn/what-is-geocoding pillar surveyed the major providers briefly; this article goes deeper with side-by-side comparisons and use-case guidance.

The overview table

| Provider | Data source | Pricing model | Self-host? | Specialty | | --------------------- | --------------------------------- | -------------------------- | ---------- | ----------------------------- | | Google Maps Platform | Proprietary + TIGER + others | Pay-per-query (~$5/1K) | No | Broadest commercial coverage | | Mapbox | OSM + commercial + own field | Free tier + paid | No (Atlas) | Strong global, fair pricing | | HERE | Proprietary + commercial | Free tier + paid | No | Automotive / European data | | OpenCage | Aggregates OSM + others | Generous free + paid | No | Open-source-friendly bridge | | Nominatim (OSM) | OpenStreetMap | Free (public limited) / self-host | Yes | Free, OSM-only | | LocationIQ | OSM + commercial | Free tier + paid | No | OSM-based commercial | | SmartyStreets | USPS / proprietary US | Per-query, US-focused | No | CASS-certified US | | Esri / ArcGIS World Geocoder | Proprietary + national | Pay-per-query / credit | No | Enterprise GIS integration | | Pelias | OSM + OpenAddresses + others | Self-host only | Yes | Open-source, modular pipeline | | TomTom | Proprietary + commercial | Free tier + paid | No | Automotive / European | | AWS Location Service | Esri + HERE under the hood | Per-query (AWS) | No | AWS-integrated |

Google Maps Platform

The default for many enterprise applications. Strengths:

• Broadest commercial coverage, particularly in developing-world regions where competitors lag.
• High rooftop precision for major metros worldwide.
• Mature autocomplete API with sub-100 ms latency.
• Address validation add-on for delivery-quality verification.

Weaknesses:

• Expensive: ~$5 per 1,000 geocoding queries (with $200/month free credit). High-volume applications can spend thousands per month.
• Display requirements: results are usually constrained to be displayed on a Google Map (Maps Embedded SDK or similar). Read the ToS carefully.
• Vendor lock-in: switching from Google requires re-engineering result handling because the location_type taxonomy is Google-specific.

Right for: large enterprises with budget; applications where maximum coverage is non-negotiable; teams that already use Google Cloud and want integration.

Mapbox

The Coordinately choice. Strengths:

• Generous free tier (100K geocoding requests / month).
• Competitive paid pricing (~$0.50–0.75 per 1,000 above free tier).
• Strong global coverage comparable to Google.
• Good developer experience — well-documented API, Python / JS / Mobile SDKs.
• Confidence bands (high / medium / low) exposed in the API.

Weaknesses:

• No-retention ToS (arch §19.2): cannot cache results. Forces per-request live API calls.
• limit=5 recommended (arch §19.1): top result not always best; always request multiple.
• POI search requires a separate API (Mapbox Search) — Geocoding API alone returns addresses, not landmarks.

Right for: web applications with moderate query volume; applications that need good global coverage without Google's pricing; applications that can live with the no-retention constraint. Coordinately uses Mapbox.

HERE

Strong European and automotive focus. Strengths:

• Excellent European coverage, including detailed multi-language address handling.
• Strong automotive data: routing, traffic, drive-time estimates.
• Generous free tier for non-commercial use.

Weaknesses:

• Less prominent in US market; coverage is good but Google / Mapbox dominate the developer mindshare.
• Older API design in some places; documentation scattered across HERE's legacy and new APIs.

Right for: European-focused applications; automotive / fleet-management work; applications that need multi-language European address handling.

OpenCage

A relatively newer entrant. Strengths:

• Aggregates multiple geocoding sources (Nominatim, OpenAddresses, etc.) into a single API.
• Generous free tier (2,500 requests / day for non-commercial).
• Open-source-friendly: explicit support for hobbyist / academic use.
• Transparent ToS: no caching restrictions; data attribution per OSM ODbL.

Weaknesses:

• Smaller coverage than commercial alternatives in some regions (because it's built on open-source data).
• Lower commercial-grade quality for high-volume enterprise use.

Right for: hobby projects, academic work, non-commercial applications, open-source initiatives. Not generally the right answer for high-volume commercial deployment.

Nominatim (OSM)

The open-source standard. Strengths:

• Free (both public service and self-host).
• Global coverage via OpenStreetMap.
• No commercial restrictions when self-hosted.
• Reference open-source implementation: many alternatives are derivatives.

Weaknesses:

• Public service is strictly rate-limited: 1 request / second, no commercial use, no bulk processing.
• Self-hosting requires substantial infrastructure: ~100 GB SSD and 32+ GB RAM for global coverage.
• OSM data quality varies by region: excellent in major metros, patchy in rural and developing-world areas.
• Slower API than commercial alternatives.

Right for: self-hosted production use where licensing is the dominant constraint; academic and research work; applications with strict data-privacy requirements (all queries stay in-house).

SmartyStreets

US-focused commercial. Strengths:

• CASS-certified for US addresses; USPS-compliant standardization and deliverability validation.
• High accuracy for US addresses.
• Bulk processing at competitive pricing.

Weaknesses:

• US-only: no international coverage.
• Pay-per-query pricing can be expensive at scale.

Right for: US-focused mailing-list cleanup, direct-mail campaigns, US delivery-address verification.

Esri / ArcGIS World Geocoder

Enterprise GIS integration. Strengths:

• Integration with ArcGIS for organisations already on Esri platforms.
• Strong commercial coverage.
• Enterprise-grade SLAs.

Weaknesses:

• Esri ecosystem lock-in: not commonly used outside enterprise GIS contexts.
• Credit-based pricing can be opaque.

Right for: enterprise GIS organisations; ArcGIS-integrated workflows.

Pelias

Open-source self-hostable. Strengths:

• Open-source (MIT license); fully self-hostable.
• Modular pipeline: pluggable parsing, lookup, ranking.
• Combines multiple data sources: OSM, OpenAddresses, WhosOnFirst, Geonames.

Weaknesses:

• Operational complexity: setting up a production Pelias deployment is non-trivial.
• Smaller community than Nominatim.

Right for: self-hosted production where Nominatim's OSM-only limitation is too restrictive and the infrastructure can support the more-complex deployment.

TomTom

Strengths:

• Strong European coverage.
• Automotive-focused: routing, traffic.

Weaknesses:

• Smaller US presence than Google / Mapbox.

Right for: European automotive applications.

AWS Location Service

Strengths:

• AWS-integrated: pay through AWS billing.
• Backend providers selectable (Esri or HERE).

Weaknesses:

• Thin wrapper over the underlying provider; you inherit their ToS.

Right for: AWS-only organisations that want geocoding through a single AWS bill.

A decision matrix

Quick lookup by use case:

| Use case | Recommended provider | | ------------------------------------- | ------------------------------ | | Personal / hobby project | OpenCage or Nominatim (public) | | Open-source application | Nominatim (self-host) or Pelias | | Small-to-medium commercial | Mapbox (free tier + paid) | | High-volume commercial | Google (cost vs accuracy) or Mapbox (cost) | | US mail-quality work | SmartyStreets | | European-focused | HERE or TomTom | | Enterprise GIS | Esri ArcGIS | | Privacy-sensitive (no third-party) | Nominatim self-host or Pelias | | Bulk batch processing | Provider-specific batch APIs | | Automotive / routing | HERE or TomTom | | AWS-integrated | AWS Location Service |

Benchmark approach

Comparing geocoders honestly requires controlled benchmarking. The typical methodology:

1. Curate a ground-truth dataset: a few hundred to a few thousand known-correct address-coordinate pairs, representing the target region(s) and address types (urban, suburban, rural).
2. Query each geocoder with the address; record the returned coordinate and tier.
3. Measure error: distance from ground-truth coordinate to returned coordinate, broken down by tier.
4. Measure coverage: percentage of addresses successfully matched at each tier.
5. Repeat regularly: data quality changes; benchmarks age fast.

UCGIS (University Consortium for Geographic Information Science) publishes academic benchmarks. Independent comparisons exist but are often vendor-funded; treat specific numerical claims with skepticism.

For internal benchmarking, the OpenAddresses project publishes large open-source address-coordinate datasets useful as a baseline.

A worked cost comparison

For an application processing 500,000 geocoding queries per month (typical small-to-medium logistics app):

Google Maps Platform:
  $200/month free credit covers ~40,000 queries.
  500K queries: 460K × $0.005 = $2,300/month above free.

Mapbox:
  100K queries free per month.
  500K queries: 400K × $0.00075 = $300/month above free.

OpenCage:
  2,500/day = ~75K/month free.
  500K queries: 425K × ~$0.0008 = ~$340/month above free.

Self-hosted Nominatim:
  $0 per query, but server (8 vCPU, 32 GB RAM, 500 GB SSD):
  ~$300/month AWS or equivalent.

SmartyStreets (US-only):
  500K queries × $0.0035 = $1,750/month.

For this workload, Mapbox is the cheapest commercial option; self-hosted Nominatim is competitive on cost but requires operations overhead. Pricing structures change — always verify against current provider docs.

Common misconceptions

“Google is always best.” Google has the broadest commercial coverage and is the safe default for enterprise applications with budget. But Mapbox, HERE, and OpenCage are competitive in many regions and substantially cheaper. For specific niches (US CASS, automotive routing, European multi-language), other providers may beat Google.

“OSM-based geocoders are always worse than commercial.” Depends on region. In well-mapped OSM regions (Western Europe, parts of US, parts of Australia), OSM coverage is comparable to commercial. In other regions (parts of US suburbs, much of developing world), commercial proprietary data is better. Multi-provider applications often consolidate results from both.

“Pricing is fixed.” Most providers offer custom enterprise pricing at high volumes. The published per-query rates apply to small-medium volumes; for millions of queries per month, the negotiated rate may be much lower.

“A provider's coverage doesn't change.” Coverage updates continuously as providers ingest new data. Today's patchy region may be excellent in two years; today's leader may slip if a competitor gets newer data. Periodically re-benchmark for critical applications.

“Free public Nominatim is sufficient for production.” It explicitly is not. The OSM Foundation's usage policy limits to 1 request / second with no commercial use. Production deployment requires either self-hosting Nominatim or using a paid service. Many startups have been surprised by Nominatim service-level limits after launch.

“Switching providers is easy.” It's non-trivial. Each provider has its own accuracy taxonomy, ToS quirks, rate limits, and response format. Application code that uses Mapbox's relevance score won't work with Google's location_type without re-implementation. Multi-provider abstractions exist (some commercial address-validation services) but add their own complexity.

“Geocoders can be benchmarked once.” The data changes; pricing changes; ToS change. Benchmarks older than 12–24 months are increasingly stale. The Coordinately tools' current Mapbox choice (per arch §19) was made in May 2026; that decision is reviewed periodically.

“Free always means free of any constraint.” Most “free” geocoders have rate limits, usage attribution requirements, or commercial-use restrictions. The public OpenStreetMap Foundation Nominatim is the most constrained (1 request / second per user, no commercial). OpenCage is the most permissive (clear free tier with no attribution required for paid plans, attribution required for free use). Always read the specific service's usage policy.

“Picking the geocoder is a one-time decision.” For long-lived applications, plan for migration. Build abstractions that let you swap providers when pricing changes, when coverage in your region shifts, or when ToS become inconvenient. Hard-coding provider-specific result formats into application code makes future migration expensive.