Comparison
Fabric Data Factory is Azure Data Factory rebuilt inside Fabric and billed against Fabric capacity. Etlworks gives you the same pipelines plus real CDC, EDI, and API integration, on any cloud or on-prem, at a fixed price.
The verdict
Feature breakdown
| Capability | Etlworks | Microsoft Fabric Data Factory |
|---|---|---|
| Pricing & commercial | ||
| Starting price (monthly) | $300 | Fabric capacity (F-SKU), reserved or pay-as-you-go |
| Pricing model | Fixed per tier | Capacity units consumed per activity |
| Cost transparency | High — flat rate | Low — CU burn varies by activity and concurrency |
| Vendor lock-in | Monthly or annual, no contract | Fabric and OneLake native |
| Integration scope | ||
| Sources | 260+ | Large connector set, Microsoft-weighted |
| Destinations | Warehouses, databases, SaaS, NoSQL, files, APIs, queues, IoT, email | OneLake, Fabric Warehouse, Azure SQL, Synapse, Power BI |
| ETL capabilities | ETL, ELT, Reverse ETL, wildcard processing | Pipelines and Dataflows Gen2 |
| API management | Full | Partial — Azure API Management, licensed separately |
| EDI processing | X12, EDIFACT, HL7, FHIR | Logic Apps B2B add-on, separate service |
| On-prem deployment | cloud only, on-prem reached via data gateway | |
| Embeddable | ||
| Transformations | ||
| Visual mapping | drag-and-drop designer with live preview | Dataflows Gen2 and Power Query |
| Scripting languages | SQL, JavaScript, Python, XSLT, shell | M, SQL, notebooks |
| Nested and hierarchical data | JSON, XML, Avro, Parquet — read, write, normalize, flatten by dragging | Partial — passes JSON through, unpacking happens downstream |
| Warehouse pushdown (ELT) | transform before load or in the warehouse, same engine | OneLake and Fabric Warehouse |
| Reusable logic | macros, templates, and 3,900+ prebuilt flow templates | dataflows and pipeline templates |
| Lookups and enrichment | Lookup Builder for cross-source lookups | |
| Data validation | validation rules with per-step error handling | |
| Orchestration & workflow | ||
| Scheduling | Cron expressions and fixed intervals, with per-schedule parameters | Schedule and event triggers inside Fabric |
| Event-driven triggers | HTTP listeners and webhooks, message queues, file and email events | Fabric event streams and Activator |
| Continuous execution | looping schedules for CDC and queue consumers | Partial — Event Streams and streaming dataflows |
| Visual workflow builder | Composer canvas, 200+ flow types | pipeline canvas, inherited from Data Factory |
| Nested workflows | nested flows with conditional and looped steps | Invoke Pipeline, ForEach, If, Until |
| Run external tools | shell and SSH scripts, CLI, JavaScript, Python, SQL, HTTP calls | Partial — notebook, Web, and Azure Function activities |
| Parallel execution | overlapping schedules run as independent, separately cancellable instances | parallel activities, bounded by capacity units |
| Retries and error handling | per-step exception handling with notifications | activity retry policies and failure paths |
| Run monitoring | per-schedule status, run history, automatic Flow Findings reports | Monitoring hub |
| CDC & Streaming | ||
| CDC engine | Debezium-compatible, built-in (no Kafka required) | Partial — Mirroring and copy-job incremental, not a general CDC engine |
| Database CDC sources | MySQL, Postgres, SQL Server, Oracle, MongoDB, DB2, others | SQL Server, Azure SQL, Cosmos DB, Snowflake mirroring |
| Streaming queues | Kafka, EventHubs, Kinesis, SQS, PubSub, ActiveMQ, RabbitMQ | Event Hubs, Event Streams |
| IoT brokers | MQTT brokers | Partial — via Event Hubs and IoT Hub |
| Real-time replication | Log-based CDC, full, incremental | Mirroring into OneLake |
| Change tracking modes | Log-based, trigger-based, timestamp/high-watermark | Incremental, mirroring, watermark |
| Developer experience | ||
| REST API | full API for flows, connections, schedules, and runs | cloud API |
| CLI | full CLI with built-in SQL | cloud CLI |
| MCP server | built-in — connect Cursor, Claude, or ChatGPT to your instance | |
| Client libraries | Python, Bash, and PowerShell clients | cloud SDKs in many languages |
| Version control | built-in — automatic history, diff, and revert on every artifact | Partial — infrastructure-as-code, not artifact history |
| Embeddable / white-label | ||
| Compliance & security | ||
| SOC 2 Type 2 | audited; report under NDA, SOC 3 public | plus ISO 27001, PCI, FedRAMP |
| HIPAA | supported with a BAA | covered by the cloud BAA |
| GDPR / DPA | compliant, DPA available | |
| SSO and MFA | SAML SSO, optional 2FA, JWT stateless auth | cloud IAM and directory integration |
| Role and artifact-level access | six roles plus tag-based scoping of flows, connections, and schedules | fine-grained IAM policies |
| Encryption and data handling | TLS in transit, encrypted at rest, customer-managed PGP, SSH tunnels, IP allowlisting; rows are not persisted by default | platform KMS, customer-managed keys |
| Audit logging | admin actions logged, access logs monitored | cloud-native audit trail |
| Security testing | monthly vulnerability and penetration scans, static analysis blocking every build, periodic third-party audits | continuous, platform-wide |
| Gen AI | ||
| AI agent | Built-in agent (Simba) — builds and edits flows from chat | Partial — Copilot drafts pipelines and dataflows inside Fabric |
| Agent capabilities | Reads metadata, reads/samples data, writes JS & SQL, schedules, deploys, monitors | Code and formula generation, summarization |
| Natural-language flow building | ‘Vibe-build’ — create flows by describing what you want | Partial — Copilot in Data Factory |
| AI-driven mapping | Auto-suggests source-to-destination mappings | Partial — Copilot suggestions |
| Built-in analytics | Agent runs analysis on flow data and pipeline behavior | Power BI is part of Fabric |
| Chat across product | Same agent context on every screen | Copilot across Fabric workloads |
| CLI for agent | Full CLI access for run/deploy/monitor/manage | Partial — Fabric CLI and REST APIs, not agent-driven |
| Trains on customer data | Never | Per Microsoft terms |