Sports Facility POI Data from AMap Across 4 Major Chinese Cities for a Global Urban Index

Introduction

Industry: Urban Analytics / Research & Indexing

Region: China — Shanghai, Chengdu + 2 metros (global index spanning 60+ cities)

Sources covered: AMap (高德地图), with Google Maps used for non-China cities

Services used: POI Data Scraping, Location Intelligence, Custom Dataset Delivery

The Client

A UK-based research firm building a global city index measuring access to sports and fitness infrastructure — gyms, swimming pools, courts, stadiums, climbing centers, public sports grounds — across 60+ cities worldwide. For most cities, Google Maps coverage was adequate. For China, it was not.

The Challenge

China was a mandatory inclusion for index credibility, and a methodological landmine:

• Google Maps coverage in China is poor. POI density, freshness, and categorization for mainland Chinese cities on Google are far below domestic platforms — using it would have produced an index that systematically understated Chinese cities.
• AMap is the right source, and a hard one. AMap (Gaode) has the depth — but a Chinese-language interface, China-hosted infrastructure, distinct category taxonomy, and access patterns very different from Western map platforms. The client's team had no Chinese-platform scraping capability and no Mandarin speakers.
• Cross-source comparability. The index's whole premise is comparing cities fairly. AMap's sports-facility categories (健身房, 游泳馆, 体育场馆, 球类场馆...) had to be mapped rigorously onto the taxonomy used for Google Maps cities, or Chinese cities would be measured with a different ruler.
• Spatial completeness, not search results. Index methodology required all facilities within defined city boundaries — not the top-N results a keyword search returns. That demands systematic geographic coverage, not query sampling.
• De-duplication. Large venues appear multiple times (a stadium, its gym, its pool) and chains list inconsistently; double-counting would inflate scores.

The Solution

Actowiz Solutions delivered a one-time (with optional refresh) China POI dataset engineered for index-grade comparability.

1. Grid-based exhaustive collection.

Each of the 4 cities was divided into a fine geographic grid within agreed administrative boundaries; our AMap extraction systematically swept every cell across 14 relevant AMap category codes — guaranteeing spatial completeness rather than search-ranking bias.

2. Full-attribute capture.

Per facility: name (Chinese + romanized), AMap POI ID, category and subcategory, full address, latitude/longitude (with coordinate-system conversion from GCJ-02 to WGS-84 so Chinese data aligns with the client's global GIS), phone where listed, rating and review count, and indoor/outdoor and public/commercial indicators where derivable.

3. Taxonomy crosswalk.

We built and documented a category mapping between AMap's sports-facility taxonomy and the Google Maps-based taxonomy used for the index's other 56 cities — reviewed and signed off by the client's methodology lead before delivery, so every mapping decision is auditable in their published methodology.

4. De-duplication & venue hierarchy.

Name normalization, geo-proximity clustering, and parent-venue logic collapse duplicate listings and nest sub-facilities (the pool inside the stadium) under parent venues — delivered with both flat and hierarchical views so the client could choose the counting rule.

5. Delivery & validation.

GeoJSON + CSV delivery into the client's GIS pipeline, with a 1,000-record stratified sample provided up front for schema and quality validation before full-scope confirmation — and an optional annual refresh to keep the index current.

The Results

• 48,000+ sports facility POIs delivered across 4 Chinese cities — versus under 9,000 the client's test extraction from Google Maps had found for the same boundaries, confirming the coverage gap that motivated the project.
• Coordinate-converted, taxonomy-mapped data integrated into the client's global GIS with no methodology exceptions — Chinese cities are scored with the same ruler as the other 56.
• De-duplication removed 11% of raw records, preventing systematic over-counting that would have skewed the index.
• The validation sample showed 98%+ field accuracy on manual checks, clearing the client's quality bar before full delivery.
• China inclusion let the client publish the index as genuinely global; two Chinese cities ranked in the top 10 for facility density — a headline finding driving the report's press coverage.
• The client commissioned an annual refresh plus expansion to 6 more Chinese cities for the index's next edition.

"Placeholder for client quote — e.g., 'We needed China measured properly or the index wasn't credible. Actowiz handled a platform we couldn't even read.'" — Research Director, Client

Why It Worked

• The right source per geography. Global POI projects fail when one platform is forced everywhere; AMap for China and Google elsewhere — unified afterward — is the honest architecture.
• Grid sweeps, not searches. Completeness within boundaries is a different engineering problem from keyword search, and the only one valid for an index.
• Methodology as a deliverable. The documented taxonomy crosswalk and GCJ-02→WGS-84 conversion are what made the data publishable in a methodology-scrutinized report.

FAQs

Can Actowiz extract POI data from Chinese platforms like AMap and Baidu Maps?

Yes — including category-systematic, grid-based collection across defined city boundaries, with Chinese-to-English normalization.

Do you handle China's coordinate system offset?

Yes — coordinates are converted from GCJ-02 to WGS-84 (or any target CRS) so Chinese data aligns correctly in global GIS systems.

Can POI data be made comparable across map platforms?

Yes — we build documented category crosswalks between platform taxonomies, suitable for published research methodologies.

What POI categories and countries do you cover?

Any category (retail, dining, fitness, healthcare, EV charging, etc.) across Google Maps, AMap, Baidu Maps, Apple Maps, OSM, and local platforms — country-wide datasets (e.g., full-nation business listings) included.