DOCUMENTATION
Overview
Evaluation domains in DEC Bench.
Domains define the business contexts used to evaluate agent performance on realistic data engineering work.
v0.1: Foo Bar Domain
The v0.1 release ships exclusively on the Foo Bar synthetic domain. This uses dummy data with no production business semantics, keeping the focus on data engineering competency rather than domain knowledge.
- Foo Bar (Dummy): 36 scenarios (13 tier-1, 18 tier-2, 5 tier-3) covering ingestion, schema design, query optimization, debugging, transformation, migration, reliability, data quality, and end-to-end pipelines.
Coming Soon
Future releases will add production-realistic domains:
| Domain | Focus | Status |
|---|---|---|
| B2B SaaS | Product usage events, subscription lifecycle, feature adoption | Planned for v0.2 |
| B2C SaaS | User activity streams, content interactions, session data | Planned for v0.2 |
| UGC | Posts, comments, reactions, moderation signals | Planned for v0.2 |
| E-commerce | Orders, inventory, catalog, customer behavior | Planned |
| Advertising | Impressions, clicks, conversions, bid data | Planned |
| Consumption-Based Infra | API calls, compute usage, storage metering | Planned |
How To Use These Pages
Start with the domain that best matches your workload profile, then map its constraints to competency evaluations under Competencies.