A high-fidelity synthetic data warehouse simulating the complete operational stack of a global beauty & cosmetics ecommerce company.
Built for:
- 🤖 Autonomous analyst benchmarking
- 📊 BI & analytics engineering
- 🧠 Agentic AI systems
- 🏗️ Data warehouse testing
- 🔍 Causal inference & root-cause analysis
- 📈 Attribution & experimentation modeling
Inspired by real-world enterprise commerce systems — but fully synthetic, open, and research-friendly.
Note: Only sample data available here. For actual dataset, please visit HuggingFace Datasets: https://huggingface.co/datasets/sidhasadhak/beautycommerceos
BeautyCommerceOS is not a toy ecommerce dataset.
It is a fully simulated enterprise operating system containing:
- clickstream events
- identity stitching
- orders & payments
- inventory & fulfillment
- attribution systems
- influencer commerce
- experimentation infrastructure
- finance & accounting layers
- KPI semantic conflicts
This dataset was designed to emulate the analytical complexity of modern commerce organizations.
warehouse/
├── bronze/
│ ├── sessions.parquet
│ ├── events.parquet
│ └── anonymous_users.parquet
│
├── silver/
│ ├── products.parquet
│ ├── skus.parquet
│ ├── brands.parquet
│ ├── suppliers.parquet
│ ├── warehouses.parquet
│ └── identity_map.parquet
│
├── gold/
│ ├── orders.parquet
│ ├── order_items.parquet
│ ├── carts.parquet
│ ├── payments.parquet
│ ├── refunds.parquet
│ ├── campaigns.parquet
│ ├── attribution.parquet
│ ├── inventory_snapshots.parquet
│ ├── shipments.parquet
│ ├── invoices.parquet
│ ├── pnl_daily.parquet
│ └── ...
Most ecommerce datasets only contain:
- orders
- products
- customers
BeautyCommerceOS models the entire enterprise.
Including:
- sessions
- clickstream events
- funnel progression
- add-to-cart flows
- attribution leakage
- carts
- orders
- order items
- payments
- refunds
- subscription behavior
- inventory snapshots
- warehouse transfers
- stockouts
- supplier delays
- shipments
- delivery failures
- campaigns
- ad impressions
- ad clicks
- influencer commerce
- multi-touch attribution
- ROAS distortion
- A/B testing
- treatment/control groups
- traffic allocation
- experiment contamination
- invoices
- journal entries
- revenue recognition
- P&L reporting
- KPI definition conflicts
This warehouse intentionally includes:
- attribution ambiguity
- operational inconsistencies
- delayed reconciliation
- refund lag
- coupon abuse
- stockouts
- payment retries
- KPI disagreements
- semantic conflicts
Because:
real enterprise data is messy
The goal is to benchmark systems against reality — not clean textbook tables.
| Attribute | Value |
|---|---|
| Domain | Beauty & Cosmetics Ecommerce |
| Format | Parquet |
| Total Size | < 10 GB |
| Geography | US, UK, DE, FR, CA |
| Time Coverage | ~2 years |
| Data Type | Fully synthetic |
| Warehouse Style | Bronze / Silver / Gold |
| Granularity | Event-level |
| Identity Model | Anonymous + stitched users |
- Why did TikTok ROAS improve while profit declined?
- Which influencers drive low-retention customers?
- Which stockouts caused the highest revenue loss?
- Which warehouses generate the most delayed deliveries?
- Why does finance revenue differ from marketing revenue?
- How much revenue was recognized late?
- Root cause analysis
- Causal reasoning
- Cross-functional diagnosis
- Metric reconciliation
- Semantic consistency checks
✅ Autonomous AI agents ✅ Analytics engineering ✅ Synthetic enterprise benchmarking ✅ Data quality tooling ✅ BI dashboards ✅ Attribution modeling ✅ Supply chain analytics ✅ Experimentation systems ✅ LLM evaluation frameworks ✅ Semantic layer testing
This dataset is:
✅ Fully synthetic ✅ Generated programmatically ✅ Contains no PII ✅ Contains no real customers ✅ Contains no payment data ✅ Contains no proprietary enterprise records
Recommended:
CC BY 4.0
Use freely for:
- research
- benchmarking
- education
- experimentation
- open-source tooling
PRs are welcome for:
- new anomaly generators
- causal benchmark suites
- warehouse extensions
- dbt models
- semantic layers
- synthetic CRM systems
- forecasting pipelines
- observability tooling
Potential future modules:
- customer support systems
- CRM lifecycle automation
- loyalty programs
- subscription commerce
- feature stores
- retail/POS systems
- forecasting infrastructure
- fraud systems
- schema drift simulation
BeautyCommerceOS was designed to answer one question:
Can autonomous systems reason across the complexity of a real enterprise?
Not just dashboards. Not just SQL. But causality, inconsistency, ambiguity, and operational reality.