Industrial-grade B2B prospecting pipeline — multi-source scraping, dedup, enrichment. One command, one clean CSV, ready for Mailchimp / CRM / Lemlist.
Article on Julienweb: https://julienweb.fr/blog/lead-scraper-pro-prospection-b2b-automatisee/10046/
| 🔎 Sources | PagesJaunes · Pappers · Google Maps · Planity · Cylex · Instagram |
| 🧹 Merge | Dedup by Name+Address, email-based consolidation, junk URL filtering |
| ✨ Enrich | Email + phone + website via Google → site scraping → Facebook (confirmed links only) → contact@domain fallback |
| 🛡️ Robust | Resume-safe (.state.json), manual captcha, isolated Chrome profile |
| 🎯 Use case | Local B2B outreach · craftsmen · freelancers · SMBs |
Stack: Node.js · Playwright · Isolated Chrome profile
Copy ./config/v5_bobigny.js to ./config/v6_xxx.js
Update ./config/index.js
npm install && npx playwright install chromium
npm run scan # collect from all sources
npm run enrich # merge + enrich
# → data/vX/results_final_enriched.csv ✅
✅ Final result, from config to Google Sheet:
Youtube demo video:
# Install dependencies
npm install
# Install Playwright browsers
npx playwright install chromiumThe enricher (npm run enrich) runs a visible Chrome window with a persistent profile so you can solve captchas and stay logged in to Google and Facebook across runs.
One-time setup:
- On first run, Chrome opens with the persistent profile defined in
config/index.js→chromeProfileDir - Log in to Google — solve the first captcha if prompted, then sign in with your Google account
- Log in to Facebook — the enricher will pause on the FB login page; log in manually, then the script resumes automatically
- The profile is saved permanently — you only do this once
⚠️ Keep the profile folder. Deleting it means re-doing all logins. Never commit it — it contains your session cookies.
Why the profile is stored outside the project:
After heavy use, the Chrome profile grows to ~16 000 files (cache, IndexedDB, session data). Storing it inside a Google Drive–synced folder causes Drive to re-sync thousands of tiny files after every scraping session, which is slow and pointless. The profile is therefore kept at a local path outside of Google Drive:
C:\Users\molok\Local Sites\chrome_scraper_profile
This path is configured in config/index.js via chromeProfileDir. To change it, edit that single line — all scripts pick it up automatically.
index.js — CLI entry point: node index.js <scan|merge|enrich>
config/ — version, city, and category settings
index.js — active config pointer (edit this to switch campaign)
v3_example.js — annotated reference config
v5_bobigny.js — latest campaign config (copy to create a new one)
sources/ — scraper modules, one file per source
pagesjaunes.js — PagesJaunes.fr (Playwright headless, captures Facebook link)
pappers.js — Pappers.fr (Playwright headless)
googlemaps.js — Google Maps Places API (HTTP)
planity.js — Planity.com (Algolia HTTP API)
instagram.js — Instagram (Chrome persistent profile + stealth)
cylex.js — Cylex.fr [disabled — bot detection]
pipeline/ — post-scan processing
merge.js — dedup + normalize raw CSVs → results_final.csv
enricher.js — email/phone/website enrichment → results_final_enriched.csv
data/ — campaign output, one subfolder per version (v1/, v2/, …)
tests/ — standalone test scripts (Google, Facebook, Instagram…)
images/ — screenshots used in this README
┌─────────────────────────────────────────────────────────┐
│ 1. SCAN npm run scan │
│ │
│ planity.js ──► planity_results.csv │
│ pappers.js ──► pappers_results.csv │
│ pagesjaunes ──► pagesjaunes_results.csv │
│ googlemaps ──► googlemaps_results.csv │
│ instagram.js ──► instagram_results.csv │
│ cylex.js [disabled — bot detection] │
│ │ │
│ 2. ENRICH npm run enrich │
│ │ │
│ ┌── merge (automatic first pass) ──────────────┐ │
│ │ dedup by Name+Address key │ │
│ │ consolidate by shared email │ │
│ │ normalize categories │ │
│ │ ▼ │ │
│ │ results_final.csv │ │
│ └──────────────────────────────────────────────┘ │
│ │ │
│ For each company (one by one): │
│ ① Google search │
│ ② Visit top organic links │
│ ③ Facebook (confirmed link only) │
│ │ │
│ Post-process (after full loop): │
│ ④ inject contact@domain.tld (no email + site) │
│ ▼ │
│ results_final_enriched.csv ✅ │
└─────────────────────────────────────────────────────────┘
npm run mergeis available as a standalone command if you need to regenerateresults_final.csvwithout running the full enrichment.
Each scraper runs headless (or stealth) and writes its own CSV into data/[VERSION]/.
| Script | Source | Method |
|---|---|---|
planity.js |
Planity.com | Algolia HTTP API — categories from config.categoriesPlanity |
pappers.js |
Pappers.fr | Playwright headless |
pagesjaunes.js |
PagesJaunes.fr | Playwright headless — also captures Facebook profile link if present |
googlemaps.js |
Google Maps API | HTTP |
instagram.js |
Chrome persistent profile (stealth) — manual login → Google site:instagram.com → fallback Bing — profile URL validation → meta tag + DOM link extraction |
|
cylex.js |
Cylex.fr | Playwright headless — Currently disabled (bot detection) |
Runs Chrome in visible + stealth mode (same persistent profile as the enricher) to avoid Instagram bot detection.
Flow:
- Manual login — Chrome opens Instagram; script pauses until you press Enter after logging in
- Manual Google check — pauses on Google for cookie acceptance / captcha
- For each category in
config.categories:- Google search
<category> <city> site:instagram.com(20 results) - If 0 valid profiles → Bing fallback with Bing URL base64-decode
- Profile URL validation: only
instagram.com/<username>/accepted —/p/,/reel/,/stories/,/explore/etc. are rejected - Up to 8 profiles per search
- Google search
- Profile scrape — reads
<meta name="description">(contains bio + follower counts) and scans DOM links:- Email — extracted from the meta description text
- Website — decoded from
l.instagram.com/l.php?u=redirect links in the DOM (the "link in bio" field) - Bio and display name are not kept in the output
- Location filter — keeps only profiles where the meta description or search title contains the target city/zip
Output columns: Handle · Email · Site · Catégorie · URL
⚠️ Requires a live Instagram session in the Chrome profile (chromeProfileDirinconfig/index.js). If the session expires, re-run and log in again when prompted.
Reads all source CSVs and builds a single clean results_final.csv.
- Normalizes columns:
Nom,Adresse,Téléphone,Site Web,Email,Facebook,Catégorie,Source - Cleans phones — strips labels, keeps valid French formats (
0X XX XX XX XX,+33 X…) - Filters invalid websites — rejects Google Maps, social media, directories
- Primary dedup by normalized
Name+Addresskey — fills missing fields on collision - Secondary dedup by shared email — merges rows that share the same address
- Normalizes categories via a keyword map (60+ patterns → clean labels)
Runs Chrome in visible mode with a persistent profile. Requires a one-time manual login to Facebook and Google at startup.
For each company, in order:
Query: "Company Name" "City" email contact
- Extracts emails and phones from the page text
- Collects organic links (H3-based, all Google domains filtered out)
- Parses emails and phones from each page
- Looks for a Facebook profile link on the page
(ignores
/sharer,/share,/plugins— only real profile links)
Facebook is visited only when the profile URL is confirmed from a reliable source:
| Source | Trust |
|---|---|
| Link scraped from PagesJaunes listing | ✅ High |
| Link found on the company's own website | ✅ High |
Google search "name" site:facebook.com |
❌ Not used — too many false positives |
When a trusted Facebook link is found:
- Extract profile base URL (handles
/username,/pages/Name/ID,profile.php?id=, deep sub-pages like/photos/…) - Visit
profileBase/about - If empty → try
profileBase/about_contact_and_basic_info(personal profile) - If still empty → try
profileBase/directory_contact_info(business page)
Between each company: randomized pause of 15–40 seconds to avoid rate-limiting.
Resume: progress saved to filename.state.json — re-run npm run enrich to continue from where it stopped.
After the full loop completes, a final pass rewrites the enriched CSV:
- For every row that has a
Site Webbut an emptyEmail - Extracts the root domain (
www.stripped) - Injects
contact@<domain>— filtered against the same email blacklist - Logged as:
✉️ 12 email(s) contact@ ajouté(s) dans results_final_enriched.csv
This fills the "has a site, no scraped email" gap without extra HTTP requests.
results_final_enriched.csv:
| Column | Description |
|---|---|
Nom |
Business name |
Adresse |
Full address |
Téléphone |
Phone number(s) |
Site Web |
Website URL |
Email |
Found email(s), comma-separated |
Facebook |
Facebook profile URL (from source or site) |
Catégorie |
Normalized business category |
Source |
Origin(s): PagesJaunes+Planity etc. |
Emails are rejected if they match any of:
- Blacklisted domains (
google.com,facebook.com,sentry.io,wixpress.com…) - System prefixes (
noreply,no-reply,support@…) - Placeholder patterns (
nom@domaine.fr,email@example.com,jean.dupont@email.fr…) - Image file extensions (
.png,.jpg…) - Length > 80 characters
Create a config file in config/ and point config/index.js to it:
// config/v4_mytown.js
module.exports = {
version: 'v4',
location: {
name: 'Pantin 93',
keywords: ['pantin', '93500'],
zip: '93500',
coords: { lat: 48.8952, lng: 2.4008 },
radius: 2000
},
categories: ['coiffeur', 'boulangerie', 'kinésithérapeute'], // PagesJaunes, Google Maps, Instagram
pappersQueries: ['coiffeur', 'boulangerie'], // Pappers only
googleTypes: ['hair_care', 'bakery', 'physiotherapist'], // Google Maps API types
categoriesPlanity: ['kinesitherapeute', 'osteopathe'] // Planity only (slugs sans accent)
};// config/index.js
module.exports = require('./v4_mytown');# Test Google search + link extraction
node tests/test_google.js "Bryan Amram" "Pantin"
# Test Facebook profile discovery + contact extraction
node tests/test_facebook.js "FIT BARBER CUT" "Pantin"Both test scripts accept optional name and location CLI arguments and mirror the exact logic used in enricher.js.
| Problem | Solution |
|---|---|
| Google captcha blocks | Visible mode — solve manually, script resumes automatically |
| Facebook asks for login | Log in once, profile saved in chromeProfileDir (see config/index.js) |
| Process interrupted | Re-run npm run enrich — resumes from .state.json |
| Heavy rate-limiting | Increase the 15–40 s delay range in enricher.js |
| Empty results | Check config/index.js points to the correct version file |
| Instagram finds nothing | Google site:instagram.com ran first, Bing fallback tried — check if profiles exist |
npm run enrichis the primary command — it merges then enrichesnpm run mergeis optional — only if you need to regenerateresults_final.csvwithout enriching- Keep only
results_final_enriched.csvas the final deliverable - Keep
.state.jsonfiles while a run is in progress (resume cache) - One Chrome profile (path in
config/index.js→chromeProfileDir) is shared across all runs — keep it
- Respect each source's Terms of Service and
robots.txt - Collected data is B2B only (professional contacts), aligned with the GDPR "legitimate interest" legal basis
- Always include an unsubscribe link in any email campaign sent to this database
- Tool intended for clean professional outreach — not for spam or mass unsolicited messaging
This tool is provided for educational and professional automation purposes only.
By using, copying, forking or adapting this code, you agree to the following:
-
Personal responsibility — You are solely responsible for how you use this tool and for any data you collect. The author cannot be held liable for any misuse, legal violations, or damages resulting from your use of this software.
-
Compliance — It is your responsibility to ensure your usage complies with applicable laws in your jurisdiction, including but not limited to: GDPR (EU Regulation 2016/679), the French Loi Informatique et Libertés, each scraped platform's Terms of Service, and the CAN-SPAM / ePrivacy Directive for email outreach.
-
No warranty — This software is provided "as is", without warranty of any kind, express or implied. The author makes no guarantees regarding its accuracy, reliability, or fitness for any particular purpose.
-
Reasonable use — You are expected to use this tool responsibly and ethically: reasonable request rates, no harassment, no bulk unsolicited messaging, no resale of collected data without consent.
-
No endorsement — Publication of this code does not constitute an endorsement of scraping any particular website, nor an invitation to violate any platform's policies.
TL;DR — Don't spam, don't abuse, don't blame me. Use it wisely. 🙏
Julien Guézennec — Freelance web developer & AI consultant since 1998 🌐 JulienWeb.fr · 📍 Pantin (93), France · France Num certified Activator
Independent web studio specialized in WordPress development, local SEO/GEO, e-commerce, Google/Facebook Ads and digital training for craftsmen, freelancers and SMBs across Seine-Saint-Denis and the Greater Paris area.
💡 This project showcases a hands-on approach to business automation: turning hours of manual prospecting into a reproducible pipeline. Need a custom tool (scraping, automation, AI integration)? 👉 julienweb.fr/contact
📬 AI newsletter: La Gueznet IA — weekly insights on AI applied to web and local business.
Built with ☕ and Node.js in Pantin (93) · © 2026 Julien Guézennec — JulienWeb.fr