Data Scraping: The Direct Answer
Data scraping involves automatically extracting structured information from digital sources—web pages, public APIs, professional directories, HTML files—to transform it into actionable data for a CRM, prospecting database, or analysis tool. In 2026, it is one of the most in-demand skills in B2B: it fuels lead qualification, competitive intelligence, and large-scale data enrichment.
The starting point is not the technique, it's the problem you're trying to solve: outdated CRM data, a prospect list to build, intent signals to capture. Scraping is a means, not an end.
Why B2B Data Quality Has Become a Strategic Issue
Commercial databases degrade quickly. According to Dun & Bradstreet and ZoomInfo (2024), between 18% and 32% of B2B data becomes obsolete each year depending on the field type—and this erosion is accelerating with post-Covid professional mobility.
The consequences are direct: according to Salesforce (State of Sales 2024), 76% of B2B sales professionals identify poor data quality as their main obstacle to productivity. Gartner estimates that poor data quality costs companies an average of $12.9 million per year. Scraping, coupled with enrichment, is the operational answer to this problem.
The Three Data Scraping Methods in 2026
The choice of method depends on the target source and the expected volume. There are three main families.
1. Static HTML Scraping
The simplest case: the web page directly returns its HTML content without JavaScript. You send an HTTP request, retrieve the HTML, and parse it with a library like BeautifulSoup (Python) or Cheerio (Node.js) to extract structured data.
Typically used for: professional directories (Kompass pages, societe.com, Yellow Pages), search engine results, e-commerce sites without client-side rendering. Limitations: does not work on JavaScript-rendered pages; sites can block by user-agent, IP, or rate-limiting.
2. JavaScript Rendering Scraping (Headless Browser)
The majority of modern sites generate their content client-side via React, Vue, or Angular. Raw HTML is empty without JavaScript execution. The solution: a headless browser (Playwright or Puppeteer) that simulates a real browser, executes the JS, and exposes the final DOM.
Slower and more resource-intensive than static scraping, but essential for LinkedIn, SPAs, and marketplaces. Playwright has become the standard in 2025-2026 thanks to its native handling of network requests and authentication.
3. Extraction via API
When a public or semi-public API exists, it is always the preferred route: structured, stable, and legally clear data. Relevant B2B APIs include LinkedIn API (restricted), Pappers API (French company registry), INSEE Sirene, Hunter.io, and intent platforms like Bombora or G2 Intent Data.
The practical rule: always check if an API exists before scraping. Scraping is the technical last resort, not the starting point.
B2B Scraping Tool Stack 2026
The ecosystem has stabilized around a few tools that are widely accepted by B2B growth teams:
- Playwright (Microsoft) — the reference for JS-rendered scraping in 2026. Multi-browser support (Chromium, Firefox, WebKit), clean async API, advanced network interception management.
- Scrapy — Python framework for large-scale scraping (crawling thousands of pages). Robust, extensible, integrable into Airflow or Prefect pipelines.
- Apify — managed cloud scraping platform, with a marketplace of pre-built actors (LinkedIn, Amazon, Google Maps). Suitable for teams without DevOps resources.
- Firecrawl — converts any website into structured markdown. Excellent integration with LLMs for semantic extraction.
- Phantombuster — no-code/low-code automation for LinkedIn and other social platforms. Used by growth teams for low-volume prospecting actions.
- Dropcontact — GDPR-native B2B email enrichment, built on the SIRET registry and French professional data. Does not resell purchased databases.
- Proxycurl — LinkedIn profile enrichment API without violating terms of use. Covers names, titles, experience, professional contact details.
Scraping and GDPR: What the Legal Framework Says in 2026
GDPR does not prohibit the scraping of professional data—it regulates its use. The key distinction is as follows:
- Professional personal data (first name, last name, professional email, direct phone) : their collection and processing must be based on a legal basis. Legitimate interest (Article 6.1.f of the GDPR) is the most used basis in B2B prospecting, provided that the interest of the data subject does not override that of the data controller.
- Company data (SIRET, turnover, headcount, sector) : not subject to GDPR, this data can be freely processed. It forms the basis of B2B qualification.
- Data extracted from LinkedIn : the CJEU and the CNIL have progressively clarified that publicly displayed data does not benefit from absolute protection, but LinkedIn's terms of use contractually prohibit automated scraping. The legal route involves third-party enrichment providers (Proxycurl, Cognism) who assume the risk.
The practical rule: always inform scraped contacts about the use of their data (mention in the first email, link to the privacy policy), offer a simple opt-out, and do not cross-reference sensitive data (health, political opinions). The ANSSI and the CNIL regularly publish updated guides on these issues.
Data Scraping vs. B2B Enrichment: Two Complementary Steps
Scraping collects raw data; enrichment completes and validates it. These two operations form a pipeline that should not be confused.
| Operation | What it produces | Typical tools | Limitation |
|---|---|---|---|
| Scraping | Raw data (name, URL, position, email if visible) | Playwright, Scrapy, Apify | Variable completeness, freshness not guaranteed |
| Enrichment | Validated professional email, direct phone, firmographic data | Dropcontact, Cognism, Hunter.io | Cost per record, coverage rate ≤ 85% |
| Validation | Active email (DNS, SMTP check), reachable number | NeverBounce, ZeroBounce, Bouncer | Does not guarantee that the contact is still in position |
An effective pipeline chains these three steps before any commercial action. Sending a sequence to an unenriched list means accepting a bounce rate above 5%, which degrades the sending domain's reputation—a problem difficult to correct afterwards.
To go further into practical implementation, our guide on multichannel prospecting details how to integrate this enriched data into a coherent outbound sequence.
The 5 Mistakes That Sabotage a B2B Scraping Project
- Scraping without a defined ICP target. Extracting thousands of lines without filtering by sector, size, or position produces an unusable database. Scraping must start from a precise ICP (Ideal Customer Profile): sector → size → title → geography. The finer the filter, the higher the qualification rate.
- Ignoring rate-limiting and getting banned. Sending 500 requests per minute to a site instantly activates anti-bot defenses (Cloudflare, Akamai). The basic rule: respect the
robots.txtfile, add random delays between requests (1–5 seconds), and use IP rotation via a residential proxy service if volume requires it. - Not versioning or dating extractions. Scraped data without a timestamp becomes unusable: you no longer know if the email was valid 3 weeks or 3 months ago. Each extraction must include a collection date and a validation date.
- Duplicating contacts without deduplication. Two successive scrapes from the same source produce duplicates. Without deduplication on the (email + company) pair or SIRET, the CRM gets polluted and sequences are sent twice. A scraping pipeline without a deduplication step is incomplete.
- Treating scraping as a one-shot. B2B data degrades at a rate of 2–3% per month. A scrape performed in January is 20–25% obsolete by December. You need to plan periodic rescans (quarterly at a minimum) on target accounts and priority segments.
FAQ — B2B Data Scraping
Is data scraping legal in France?
Scraping is not illegal in itself. Its legality depends on what is scraped (personal data vs. company data), how (respect for robots.txt, platform terms of use), and for what purpose (legitimate B2B prospecting vs. reselling databases). For professional personal data, the legal basis of legitimate interest (GDPR, art. 6.1.f) is generally invoked in B2B, provided that contacts are informed and an opt-out is offered. The CNIL specifies these conditions in its recommendations on commercial prospecting.
What is the difference between scraping and crawling?
Crawling refers to the automatic traversal of links to discover pages (as Googlebot does). Scraping refers to the extraction of structured data from these pages. In practice, you crawl to find URLs, then scrape each page to extract data. The two operations are often coupled in the same pipeline.
Can LinkedIn be legally scraped?
LinkedIn's terms of use explicitly prohibit automated scraping. The legal route involves the LinkedIn API (very restrictive, reserved for partners) or third-party enrichment providers like Proxycurl or Cognism, who aggregate LinkedIn data without direct scraping. The hiQ Labs v. LinkedIn ruling (US Federal Court, 2022) established that scraping public data does not violate the US CFAA, but this does not lift LinkedIn's contractual restrictions.
How much does a B2B scraping project cost?
A one-shot scrape on a simple source (directory, company pages) can be done with open-source tools (Scrapy, Playwright) for a cost primarily in development time—around 1 to 3 days depending on complexity. Large-scale scraping infrastructures (residential proxies, cloud, monitoring) represent fixed costs of €200 to €2,000/month depending on volume. Enrichment is added at a rate of €0.03 to €0.15 per enriched contact depending on the provider and the level of data required.
What accuracy can be expected from a scraped and then enriched database?
For well-defined targets (French SaaS SMEs, 10–200 employees, VP/Head of position), a scraping + enrichment + validation pipeline achieves 75–85% coverage for professional emails and 60–70% for direct phone numbers, according to Dropcontact and Cognism data (2024). This accuracy strongly depends on the freshness of the scraped source and the mobility rate of the target sector. Sectors with high turnover (early-stage startups, consulting) show faster depreciation.
How to integrate scraped data into a CRM?
Modern pipelines use a normalization intermediary before CRM insertion: either an ETL tool (Airbyte, Stitch) for large volumes, or a native API (HubSpot, Salesforce, Pipedrive all offer well-documented REST APIs) for one-time imports. The systematic point of friction is deduplication: HubSpot and Salesforce deduplicate by email by default, but not by phone number or SIRET. Plan a deduplication step upstream if these fields are the primary identifiers of your database.
