Tencent Maps Scraper — China POI, Contacts & Reviews

腾讯地图 / QQ Maps business places — names, phones, ratings, reviews, hours & coordinates across mainland China.

Scrape business places (POI) from Tencent Maps (map.qq.com) — one of China's major mapping platforms alongside Baidu Maps and Amap. Give it keywords and cities, get back a clean row per place: business name, address, phone number, star rating, review count, opening hours and map coordinates. No API key, no account, no Chinese phone number.

💡 China market entry in one run. Search 咖啡 in 北京, 上海, 广州 and wake up to thousands of real café leads with phones and ratings — ready for outreach, mapping or analysis.

⚡ At a glance

🗺️ Source	Tencent Maps / QQ Maps (腾讯地图) — public place listings
🏢 Records	One row per business place: identity, contact, location, reputation
🔍 Search	Keyword × city (Chinese recommended) with automatic pagination
📇 Contacts	Phone numbers where businesses publish them
⭐ Reputation	Star rating + review count per place
📍 Location	Address, district, city + map coordinates
🔑 Access	No API key · No account · No Chinese phone number
📤 Export	JSON, CSV, Excel via the Storage tab or API

Why Tencent Maps data?

Tencent Maps (腾讯地图, also known as QQ Maps) is one of the three map apps nearly every urban Chinese consumer carries. Its place database covers restaurants, hotels, cafés, gyms, clinics, salons, factories, offices — millions of businesses across every mainland city. For anyone selling into China, researching Chinese markets or building location products, this is ground-truth local-business data you cannot get from Google Maps, which barely covers the mainland.

Official map APIs for China require a China-verified developer account — Chinese phone number plus real-name verification — putting them out of reach for most overseas teams. This actor takes the visitor's path instead: it runs the same searches a logged-out visitor runs on the public site and returns structured results. You get the data with zero signup friction.

Why this actor?

• Zero credentials. No developer key, no login, no Chinese phone number. Paste keywords, press Start.
• Contact-first records. Phone numbers, addresses and business hours come on the same row as ratings — built for lead generation, not just dots on a map.
• Keyword × city matrix. Every keyword runs against every city automatically. Three keywords × five cities = fifteen searches in one run, deduplicated.
• Automatic pagination. Keeps turning result pages until your per-search cap is reached — no manual paging.
• Fresh residential routing. Searches run through residential IPs with automatic retry on a fresh session, so transient blocks don't kill your run.
• Clean structured output. One flat JSON row per place with stable field names — drops straight into Sheets, CRMs and databases.

What you get per place

🏢 Identity

Field	Description	Coverage
name	Business name (Chinese)	🟢 every record
poiId	Stable Tencent place id	🟢 every record
category	Place category, e.g. 娱乐休闲:咖啡厅	🟢 ~100%
keyword	Your search keyword (for attribution)	🟢 every record
city	Your search city (for attribution)	🟢 every record

📇 Contact & reputation

Field	Description	Coverage
phone	Contact phone number	🟡 when published (~95% cafés, ~100% hotels/restaurants)
rating	Star rating (0–5)	🟢 ~100%
reviewCount	Number of user reviews	🟡 when published
reviewSource	Review provider (often 大众点评网)	🟡 with reviewCount
averagePrice	Average spend per person (CNY)	🟡 food & drink mostly
openTime	Opening hours, e.g. 09:00-21:00	🟢 ~97%
businessStatus	Trading status, e.g. 正常商户	🟢 ~97%

📍 Location

Field	Description	Coverage
address	Full street address (Chinese)	🟢 ~100%
district	District / county name	🟢 ~95%
latitude / longitude	Map coordinates (GCJ-02, see FAQ)	🟢 ~100%
coordSystem	Always GCJ-02 when coordinates exist	🟢 with coords
detailUrl	Public Tencent Maps place URL	🟢 every record
scrapedAt	When the record was collected (ISO)	🟢 every record

💡 Coverage is measured from live test runs (20/20 phones-and-ratings check, October 2026), not estimated. Phone coverage genuinely varies by category — hotels and restaurants almost always publish one, small snack stalls often don't.

Output example

One dataset row per place (trimmed from a live 咖啡 × 北京 run):

{
  "name": "VOYAGE COFFEE(南锣鼓巷店)",
  "poiId": "17215659268360268776",
  "category": "娱乐休闲:咖啡厅",
  "city": "北京",
  "keyword": "咖啡",
  "address": "北京市东城区地安门东大街97号(南锣鼓巷地铁站A西北口步行190米)",
  "district": "东城区",
  "phone": "13120260798",
  "rating": 4.4,
  "reviewCount": 3,
  "reviewSource": "大众点评网",
  "averagePrice": 52,
  "openTime": "10:00-18:00",
  "businessStatus": "正常商户",
  "latitude": 39.933719,
  "longitude": 116.400201,
  "coordSystem": "GCJ-02",
  "detailUrl": "https://map.qq.com/poi/?sm=17215659268360268776",
  "scrapedAt": "2026-10-19T20:31:17.267Z"
}

Input parameters

Parameter	Type	Default	Description
searchTerms	string list	["咖啡"]	Keywords to search — names or categories, Chinese recommended. Each runs against every city.
cities	string list	["北京"]	Cities to search in — Chinese names (北京, 上海, 广州, 深圳…).
maxItemsPerSearch	integer	100	Max places per city × keyword search (1–400).
maxItems	integer	0	Max places across the whole run. 0 = no cap.
maxConcurrency	integer	2	Parallel searches (1–6). Each uses a browser — keep modest.
proxyConfiguration	proxy	Residential	Preconfigured. Tencent Maps blocks datacenter IPs.

Basic: cafés in Beijing

{
  "searchTerms": ["咖啡"],
  "cities": ["北京"],
  "maxItemsPerSearch": 60
}

Multi-keyword, multi-city lead pull

{
  "searchTerms": ["咖啡", "酒店", "火锅"],
  "cities": ["北京", "上海", "广州"],
  "maxItemsPerSearch": 100,
  "maxItems": 500
}

Nine searches, one run, deduplicated output.

How it works

1. You enter keywords and cities — e.g. 火锅 × 成都, 重庆, 北京.
2. The actor searches Tencent Maps for each pair, turning result pages until your cap, retrying on a fresh session if a search stalls.
3. You download structured rows — JSON, CSV or Excel — one place per row with contact, location and reputation fields.

Use cases

📇 China lead generation

The core use case. Pull every hotel in Shanghai, every gym in Shenzhen, every dental clinic in Beijing — with phone numbers — and feed them to your sales team or dialer. Filter by rating and reviewCount to prioritize established businesses, use district to carve up territories.

If you sell…	Search this	In these cities
☕ Coffee equipment	咖啡, 奶茶, 面包店	北京, 上海, 广州, 深圳
🏨 Hotel supplies	酒店, 民宿, 青旅	上海, 三亚, 成都, 西安
🍲 Food distribution	火锅, 烧烤, 日料	成都, 重庆, 北京
💪 Fitness gear	健身房, 瑜伽馆, 游泳馆	深圳, 上海, 杭州
💄 Beauty wholesale	美容院, 美发店, 美甲店	广州, 上海, 北京
🏥 Medical devices	诊所, 牙科, 体检中心	北京, 上海, 广州

📊 Market & competitor mapping

Count every bubble-tea shop per district to size a market. Track openings and closures month over month by re-running the same searches. Benchmark competitor footprints: where they cluster, how their ratings compare, what price bands they play in (averagePrice).

🗺️ Expansion & site selection

Retail and F&B chains use POI density + ratings to shortlist neighborhoods. Pull all 商场 (malls) and complementary categories around candidate sites, then score each area on footfall proxies (review counts) and competition density.

🚚 Logistics & delivery planning

Last-mile operators enrich stop lists with verified addresses, districts and coordinates. The district + address pair geocodes cleanly against Chinese logistics systems.

🎓 Research & data journalism

Urbanists study amenity distribution; journalists quantify chain vs independent mix per city. Stable poiId values make longitudinal tracking easy — re-run quarterly and diff.

🧩 CRM & app enrichment

Already have a China company list? Match names against Tencent POI rows to append phones, hours, ratings and coordinates. detailUrl gives analysts a one-click verification link.

Step-by-step usage

In the Console (no code)

1. Open the actor and click Try for free — the example searches 咖啡 in 北京.
2. Replace the example with your keywords and cities (Chinese works best).
3. Set Max places per search — 20 for a quick test, 100+ for coverage.
4. Click Start, then open the Storage tab when it finishes.
5. Download JSON (full fidelity), CSV or Excel, or copy the dataset URL.

Via API

curl -X POST "https://api.apify.com/v2/acts/haketa~tencent-maps-scraper/runs?token=YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"searchTerms": ["咖啡", "酒店"], "cities": ["北京", "上海"], "maxItemsPerSearch": 100}'

Fetch items from the run's default dataset when it succeeds. See the API tab on the actor page for client libraries (JavaScript, Python) and ready-made snippets.

Scheduling

Place data goes stale — businesses open, close and change phones. Most users schedule a weekly or monthly re-run per city and diff on poiId to catch openings, closures and rating movement. Use Apify Tasks (one task per city is a clean pattern) plus Schedules, and push results to Google Sheets, your CRM or a webhook from the run's Integrations tab — no glue code needed.

💡 Tips for best results

• Write keywords in Chinese. Tencent Maps indexes Chinese names and categories. 火锅 dramatically outperforms hotpot; 酒店 beats hotel. Machine-translate your category list once and reuse it.
• Use Chinese city names. 北京, 上海, 广州, 深圳, 成都, 杭州, 武汉, 西安 — pinyin sometimes works, characters always work.
• Probe before you scale. Run maxItemsPerSearch: 20 on a new category first. Check phone fill and result counts, then raise the cap.
• Split mega-pulls by city. One task per city parallelizes better, isolates failures and makes scheduling obvious.
• Mind the caps. maxItemsPerSearch is per search; maxItems caps the whole run. Set both deliberately on large matrices.
• Ratings are your quality filter. rating >= 4.5 with reviewCount >= 50 selects proven businesses; low-review rows are often brand-new openings (great for first-mover outreach).
• Coordinates are GCJ-02. They plot perfectly on Chinese maps (Tencent/Amap/Baidu). For Google Maps or GIS tools, convert GCJ-02 → WGS84 first (see FAQ).

Honest limitations

• Phones depend on the business. Hotels and restaurants publish phones ~100% of the time; small stalls and street vendors often don't. The actor returns null rather than guessing.
• Browser-based = heavier. Real map searches need a real browser plus residential routing, so runs cost more compute than pure-HTTP scrapers. Keep maxConcurrency modest (2–3).
• Coverage follows Tencent. If Tencent Maps lists few places for a niche keyword in a small city, results will be thin. Broaden the keyword or add nearby cities.
• Front-ends change. Map sites redesign periodically. The actor retries automatically against transient issues, but a major redesign may need an actor update — check the Changelog below for maintenance history.
• Chinese-language output. Names, addresses and categories come in Chinese as Tencent publishes them. Translate downstream if you need English.

FAQ

Do I need a Tencent account or API key? No. No key, no account, no Chinese phone number. The actor only reads public search listings.

Which cities are covered? Every mainland city Tencent Maps covers — all tier-1 through tier-4 cities. Search by Chinese city name.

Why do you recommend Chinese keywords? Because Tencent's index is Chinese-first. English queries often return a fraction of the matches, or none. Translate category terms once (coffee → 咖啡) and reuse them forever.

What coordinate system are the points in? GCJ-02 (Mars Coordinates), China's standard offset system — correct on Tencent, Amap and Baidu maps. For Google Maps, PostGIS or Leaflet with OSM tiles, convert to WGS84 with any open gcj02-to-wgs84 library (one function call, meter-level accuracy).

How many places can I get per search? Up to your maxItemsPerSearch cap (max 400), bounded by what Tencent returns for that query. Popular category × big-city searches yield the most; niche queries yield fewer.

How long does a run take? Roughly a minute per search including pagination and retries, divided by concurrency. The 20-place example finishes in about a minute; a 9-search matrix at concurrency 3 takes a few minutes.

Why do some rows lack phone numbers? Because the business didn't publish one on Tencent Maps. Coverage is category-dependent (see the coverage column above) — the actor never fabricates contact data.

Is poiId stable? Can I track places over time? Yes — poiId is Tencent's stable place identifier. Re-run the same searches periodically and join on poiId to detect new, removed or changed places.

Can I get reviews text, photos or menus? The actor returns review counts, ratings and review source per place — not full review text or photo galleries. Counts plus ratings cover ranking, filtering and monitoring use cases.

Does it deduplicate across searches? Yes. Overlapping keywords (e.g. 咖啡 and 咖啡厅) often return the same shops; the actor dedupes on poiId so each place appears once per run.

What happens if one search gets blocked? The actor retries that search on a fresh residential session (up to 4 attempts), then continues with the remaining searches. One stubborn query never kills the whole run.

Can I run this on a schedule? Yes — that is the recommended pattern for monitoring. Create a Task with your input, attach a Schedule (weekly/monthly), and connect the Integrations tab to Sheets, Slack, webhooks or your CRM.

How fresh is the data? Every run fetches live listings at run time. Nothing is cached between runs.

Why residential proxy and not datacenter? Tencent Maps serves captchas and empty results to datacenter IPs. Residential routing is preconfigured and required for reliable results.

Changelog

1.0 — Initial release (October 2026)

• Keyword × city search matrix against Tencent Maps place listings.
• Per-place contact (phone), reputation (rating, review count + source), hours, prices and GCJ-02 coordinates.
• Automatic result pagination plus fresh-session retry per search.
• Residential proxy preconfigured; datacenter IPs blocked by the source.
• JSON/CSV/Excel export via dataset; stable poiId for longitudinal tracking.

---

Every business place in mainland China — with its contacts, ratings and coordinates.