Parsing Yandex Maps with ZennoPoster - Case Study #1

Today I am going to walk you through a real-world Yandex Maps parsing project built with ZennoPoster — a browser automation tool that can handle just about anything you throw at it. From clicking buttons to complex multi-step workflows with Excel data substitution. This is Case Study #1, where a client needed to collect thousands of business listings from a specific region, and I will show you exactly how we solved it without proxy farms, manual labor, and within a reasonable timeframe.

+

Straight to it. The client came with a specific request: dump all organizations matching a keyword across the entire city of Moscow. At first glance — how hard can it be? Open maps, type the query, get results. But when you need not 10 or 50 but several thousand records with detailed fields — that is a completely different ballgame. Let me show you how we tackled it.

The client runs a B2B service that needed a database of Moscow-based organizations with full contact details. Export format: .CSV or .SQL. Use case: cold outreach and CRM population. No half-measures — we needed every field that could possibly be extracted from a Yandex Maps business listing.

Fields Requested for Extraction

Here is the complete list of data points we had to pull from each listing:

• Business name
• Full address with postal code
• Yandex Maps link
• Phone numbers — primary and additional
• Business hours — days and times
• Photos — image URLs
• Social networks — VK, Instagram, Facebook, website
• Metro stations — nearest
• Reviews — text, rating, count
• Description — company blurb
• Geo coordinates — latitude and longitude
• Categories — business rubrics

The market offers dozens of Yandex Maps parsers — desktop apps, cloud services. Prices range from $5 to $150 per extraction run. But there are pitfalls that only surface during real-world use:

We had two paths: classic browser-based parsing (emulating user actions) or working through the Yandex Maps internal API. Let me compare both approaches with real numbers:

We chose the API method because the client prioritized volume and speed. The browser method works well for small batches — 50 to 100 organizations when you are not in a rush. But when you are dealing with thousands, API is the only viable path.

\u{201c}

Yandex Maps returns far more data through its internal API than what is displayed to the user. The JSON response contains phone numbers, coordinates, business hours, social media links — everything you need for a complete database. You just need to parse it correctly.

Yandex Maps uses several internal endpoints to serve its frontend. One is a search endpoint that returns businesses within a given radius. Another is the business card endpoint that provides detailed information. We used both.

Step 1: Search Query

The first endpoint accepts a keyword and center point coordinates. The response is a JSON array of business listings, each containing id, name, coordinates, address and a few other basic fields. The limit is about 5000 results — beyond that, Yandex truncates the output. In practice, for Moscow with the keyword "apartment renovation" we got around 3200 businesses.

Step 2: Detailed Business Card

With each business id in hand, we send a second request for detailed information. The card endpoint returns:

• Phone numbers (primary and additional)
• Website and social media links
• Business hours
• Rating and review text
• Photos
• Categories and description

ZennoPoster supports C# code execution through actions and global snippets. Here are the key fragments from our template.

Basic HTTP Request via Zenno

We used the standard HttpClient but with mandatory browser-mimicking headers:

Parsing JSON and Writing to Table

After getting the response, we parse the JSON and populate the ZennoPoster table for export:

Anti-Ban Protection: Throttling and Random Delays

The most critical part of API parsing — staying undetected. We used a cascading delay system:

When collecting data from Yandex Maps, duplicates are inevitable. The same organization can appear under multiple keywords, have several branches with identical names, or simply appear twice in search results. We implemented a three-level cleanup system:

Level 1: ID-Based Deduplication

The most reliable method — remove records with identical organization IDs. Every Yandex Maps listing has a unique numeric id. After parsing, we sort the table by this field and remove rows where the id matches the previous one.

Level 2: Phone Number Comparison

Sometimes one organization is registered under different IDs (head office and a branch sharing the same phone number). We normalize phone numbers (strip spaces, brackets, hyphens) and compare them. Duplicates get removed.

Level 3: Coordinate Grid

If two legal entities are within 10 meters of each other and share the same name — it is a duplicate. This method effectively cleans up cases where the same location is registered both as a sole proprietorship and an LLC.

ZennoPoster can save tables directly to CSV and SQL. However, we wrote our own C# handler because the built-in export did not sanitize special characters and would break SQL insert statements.

CSV Export

When saving to CSV, it is critical to escape quotes and line breaks within fields. We used the standard approach: wrap each field in double quotes and double any internal quotes.

SQL Export

For SQL, we built INSERT statements in batches of 100 rows — the optimal balance between insertion speed and transaction size:

After export, the client receives two files: .CSV for quick viewing in Excel and .SQL for direct import into MySQL or PostgreSQL. Both files are ready to use without any additional processing.

Over the course of this project, we stepped on every possible rake. Here is what I learned so you do not repeat our mistakes.

Mistake 1: Greedy Speed

The first version of the template made requests with 50–100 ms delays. Result: IP got blocked in 3 minutes. Fix: increased the base delay to 200–700 ms with randomization. Speed dropped by 40%, but collection became stable.

Mistake 2: Unescaped Special Characters

The second CSV export version broke on names like LLC "Build-Project" — internal quotes shattered the CSV structure. Added C# escape handling — problem solved.

Mistake 3: Ignoring the Geo-Zone

Yandex Maps returns organizations within a radius from the specified point, not across the entire city. To cover all of Moscow, we had to split it into a 5×5 km grid and query each square separately.

Mistake 4: Repeated Card API Requests

If an organization was already processed in a previous run — there is no need to hit the card API again. We added id-based caching to a local JSON file. On subsequent runs, we check the cache and skip already-collected cards.

Here is a summary of what we achieved parsing Yandex Maps for Moscow with the keyword "apartment renovation":

• Total organizations collected: 3,247
• Parsing time: 1 hour 47 minutes
• CSV file size: 1.8 MB
• SQL file size: 2.4 MB
• Fields per record: 18
• Unique records after deduplication: 2,891
• Records with phone numbers: 2,412 (83.4%)
• Records with reviews: 1,987 (68.7%)

Frequently Asked Questions About Yandex Maps Parsing

Parsing Yandex Maps with ZennoPoster is a viable and cost-effective approach when you need regular updates to a business database. Key takeaways from our case study:

• API is 30–50x faster than browser parsing and returns more data
• Reasonable delays (200–700 ms) allow proxy-free operation
• Deduplication is mandatory at three levels: id, phone, coordinates
• ID-based caching saves hours on repeated runs
• A geo-grid (splitting the city into squares) is needed for full coverage

\u{201c}

Parsing is not just "grab and dump." It is a well-thought-out architecture of API requests, error handling, deduplication, and export. A ZennoPoster template is written once and then runs for years with minimal adjustments when the Yandex API changes.

If you need a similar extraction — reach out via the contact form. Specify the region, keywords, and the field list — I will tell you whether it is feasible and how long it will take.