keywordbarrage/notino-scraper

Python Code:

import sqlite3
from datasets import load_dataset
import pandas as pd

# Method 1: Load via Hugging Face datasets library
# This repository contains an open 1,000-row verified benchmark sample.
dataset = load_dataset("daas-intelligence/enterprise-sku-depletion-telemetry-notino")
print("Hub Sample Record:", dataset["train"][0])

# Method 2: Load zero-latency local pre-indexed SQLite deliverable (for full enterprise datasets)
# For full enterprise datasets, a local SQLite file is provided.
# Example: conn = sqlite3.connect("notino_sku_feed.sqlite")
# Example: df = pd.read_sql_query("SELECT sku_id, price_eur, stock_level, depletion_velocity FROM notino_products ORDER BY timestamp DESC LIMIT 10", conn)
# print(df)

Dataset Card:

pretty_name: "Enterprise High-Frequency SKU Depletion & WAF Telemetry Feed: Notino" annotations_creators: - machine-generated language: - en - tr license: apache-2.0 task_categories: - tabular-classification - time-series-forecasting tags: - notino-ecommerce - time-series-sku-depletion - pre-indexed-sqlite - parquet-benchmark - retail-intelligence - cloudflare-waf-bypass size_categories: - 10K<n<100K configs: - config_name: default data_files: - split: train path: "data/train-.parquet" - split: validation path: "data/val-.parquet"

Dataset: Enterprise High-Frequency SKU Depletion & WAF Telemetry Feed: Notino

This repository contains an open 1,000-row verified benchmark sample of the Notino.de SKU depletion and WAF telemetry feed, structured with pre-indexed SQLite and Snappy Parquet partitions for quantitative market intelligence research and data engineering evaluation.

Dataset Summary & Research Motivation

Quantitative analysts, hedge funds, and ML engineering teams rely on high-frequency market intelligence to detect critical SKU velocity, pricing elasticity, and inventory movements. However, traditional web scraping pipelines are inherently brittle. They frequently fail due to sophisticated Web Application Firewalls (WAFs), advanced anti-bot measures like Cloudflare Bot Management and Enterprise Turnstile, and dynamic TLS/JA4 fingerprinting. These defense mechanisms lead to HTTP 403 errors, IP bans, and ultimately, broken data pipelines.

This dataset addresses this critical enterprise challenge by providing a robust, protocol-level data extraction solution that bypasses these edge challenge loops with 99.9% verified pipeline availability. It showcases how to reliably ingest structured data from heavily fortified e-commerce platforms like Notino.de, a pan-European beauty and fragrance retailer. The data includes granular product attributes, real-time pricing, stock levels, and critical telemetry on WAF detections and bypass techniques, enabling quantitative researchers to build resilient models and data architects to evaluate robust market intelligence feeds.

Zero-Code Enterprise Deliverables Matrix

This dataset demonstrates the four enterprise-grade formats provided for seamless integration into diverse data architectures:

1. Pre-indexed SQLite (.sqlite):

   • Description: A self-contained, single-file database for instant local querying. Indices are pre-built on critical fields such as sku_id, price_eur, stock_level, and timestamp for high-performance analytical queries without any setup overhead.
   • Benefit: Ideal for rapid prototyping, local development, and direct ingestion into analytical tools that support SQLite.

2. Snappy Parquet:

   • Description: Columnar data storage format optimized for analytical workloads. Data is compressed using Snappy for efficient storage and fast query performance.
   • Benefit: Designed for direct ingestion into cloud data warehouses and query engines like Snowflake, Google BigQuery, and AWS Athena, minimizing ETL complexity and maximizing query efficiency.

3. Zero-Dependency Offline DataGrid (.html):

   • Description: A self-contained, interactive HTML file that presents the dataset in a sortable, filterable table. Requires no internet connection or external dependencies.
   • Benefit: Enables quantitative stakeholders, business analysts, and non-technical users to visually inspect and explore the data without requiring SQL, API keys, or specialized software.

4. CDC JSONL Stream:

   • Description: Change Data Capture (CDC) formatted JSON Lines (JSONL) for efficient incremental updates. Each line represents a record, suitable for streaming ingestion.
   • Benefit: Perfect for Enterprise RAG (Retrieval-Augmented Generation) pipelines, vector database updates, and real-time data synchronization, ensuring that models and systems always operate on the freshest data.

Dataset Schema & Data Dictionary

The dataset captures comprehensive product and telemetry information from Notino.de.

Column Name	Type	Nullable	Description
sku_id	STRING	FALSE	Unique identifier for the product (Notino Product Code / Item ID).
ean_barcode	STRING	TRUE	Universal European EAN-13 Barcode.
timestamp	DATETIME	FALSE	UTC timestamp of when the data record was extracted.
product_title	STRING	FALSE	Full title of the product.
brand	STRING	FALSE	Brand of the product (e.g., Dior, Chanel, Armani).
fragrance_collection	STRING	TRUE	Specific fragrance collection or designer line.
beauty_taxonomy	STRING	FALSE	High-level beauty category (e.g., Eau de Parfum, Gesichtspflege).
fragrance_classification	STRING	TRUE	Detailed fragrance classification (e.g., Eau de Toilette, Nische).
price_eur	FLOAT	FALSE	Current consumer retail price in EUR (€ inkl. MwSt).
list_price_uvp_eur	FLOAT	TRUE	Original Manufacturer's Suggested Retail Price (UVP) in EUR, if available.
promotional_discount_delta	FLOAT	TRUE	Price difference between list_price_uvp_eur and price_eur, indicating a discount.
bottle_volume_ml	INT	TRUE	Volume of the product in milliliters (e.g., 30, 50, 100).
unit_price_eur_per_100ml	FLOAT	TRUE	Comparative unit price in EUR per 100ml.
fragrance_top_notes	STRING	TRUE	Top notes of the fragrance pyramid (Kopfnote).
fragrance_heart_notes	STRING	TRUE	Heart notes of the fragrance pyramid (Herznote).
fragrance_base_notes	STRING	TRUE	Base notes of the fragrance pyramid (Basisnote).
fragrance_family	STRING	TRUE	General fragrance family (Duftfamilie).
value_added_services	ARRAY	TRUE	List of available personalization options (e.g., Laser Gravur-Service, Luxury Gift Box).
stock_availability	BOOLEAN	FALSE	Boolean indicating if the product is currently in stock.
warehouse_dispatch_delta_days	INT	TRUE	Estimated dispatch time from warehouse in days.
express_delivery_options	ARRAY	TRUE	Available express delivery methods (e.g., DHL, Hermes, Store Pickup).
customer_star_rating	FLOAT	TRUE	Average customer star rating out of 5.
total_review_count	INT	TRUE	Total number of customer reviews.
target_domain	STRING	FALSE	The target e-commerce domain from which data was extracted (e.g., notino.de).
detected_waf	STRING	FALSE	Primary WAF/anti-bot system detected (e.g., Cloudflare Bot Management).
tls_ja4_signature	STRING	TRUE	The TLS JA4 fingerprint signature used for the request, indicating successful bypass.
delivery_status	STRING	FALSE	Status of data delivery (e.g., "SUCCESS", "WAF_BYPASS_DETECTED").

Sample JSON Record

{
  "sku_id": "NOTINO-12345678",
  "ean_barcode": "8590666012345",
  "timestamp": "2026-09-24T10:30:00Z",
  "product_title": "Dior Sauvage Eau de Parfum 100ml",
  "brand": "Dior",
  "fragrance_collection": "Sauvage",
  "beauty_taxonomy": "Eau de Parfum",
  "fragrance_classification": "Herrenparfum",
  "price_eur": 99.99,
  "list_price_uvp_eur": 120.00,
  "promotional_discount_delta": 20.01,
  "bottle_volume_ml": 100,
  "unit_price_eur_per_100ml": 99.99,
  "fragrance_top_notes": "Bergamotte, Sichuanpfeffer",
  "fragrance_heart_notes": "Ambroxan",
  "fragrance_base_notes": "Vanille, Patschuli",
  "fragrance_family": "Würzig-Holzig",
  "value_added_services": ["Laser Gravur-Service", "Luxury Gift Box"],
  "stock_availability": true,
  "warehouse_dispatch_delta_days": 1,
  "express_delivery_options": ["DHL", "Hermes"],
  "customer_star_rating": 4.8,
  "total_review_count": 1567,
  "target_domain": "notino.de",
  "detected_waf": "Cloudflare Bot Management",
  "tls_ja4_signature": "t13d18080a10c1f01035b869680387532357",
  "delivery_status": "SUCCESS"
}

Python Usage Code

This section demonstrates how to load the benchmark sample using the Hugging Face datasets library and how a full enterprise delivery would be accessed via a pre-indexed SQLite database.

import sqlite3
from datasets import load_dataset
import pandas as pd

# Method 1: Load via Hugging Face datasets library
# This repository contains an open 1,000-row verified benchmark sample.
dataset = load_dataset("daas-intelligence/enterprise-sku-depletion-telemetry-notino")
print("Hub Sample Record:", dataset["train"][0])

# Method 2: Load zero-latency local pre-indexed SQLite deliverable (for full enterprise datasets)
# For full enterprise datasets, a local SQLite file is provided.
# Example: conn = sqlite3.connect("notino_sku_feed.sqlite")
# Example: df = pd.read_sql_query("SELECT sku_id, price_eur, stock_level, depletion_velocity FROM notino_products ORDER BY timestamp DESC LIMIT 10", conn)
# print(df)

Reproducibility & Data Quality Metrics

Our enterprise-grade ingestion pipelines are engineered for maximum reliability and data integrity, ensuring that quantitative research and machine learning models are built on a foundation of trust. Key metrics include:

• 99.9% Delivery Availability: Guaranteed uptime for continuous data streams, ensuring uninterrupted access to critical market intelligence.
• Deterministic Schema Parity: Strict schema enforcement and validation at the extraction layer prevent data drift, ensuring consistent data structures across all deliveries.
• Sub-120ms Protocol Execution Benchmarks: Optimized transport-level protocol execution to achieve minimal latency in data retrieval, crucial for high-frequency trading and real-time analytics.
• WAF Bypass Resilience: Continuous adaptation to Cloudflare Bot Management, Enterprise Turnstile, and advanced TLS/JA4 fingerprinting to maintain consistent access without bans or challenges.

This dataset represents a benchmark sample of the capabilities provided by a fully managed enterprise data service, designed to offload the complexities of web data extraction from your internal data engineering teams.