Fivetran

Fivetran maintains a library of 700+ pre-built connectors that automatically extract data from SaaS applications, databases, ERPs, and file systems using source-specific APIs and protocols. Each connector handles authentication, pagination, rate limiting, and incremental change detection (CDC/API deltas) without requiring custom code. The platform manages connector versioning, updates, and backward compatibility centrally, ensuring pipelines continue working as source APIs evolve.

Fivetran automatically detects schema changes in source systems (new columns, type changes, deletions) and applies corresponding migrations to the destination schema without manual intervention. The system uses source metadata introspection (information_schema queries, API schema endpoints) to compare current schema against the last known state, then generates and executes DDL statements (ALTER TABLE, CREATE TABLE) on the destination. Customers can configure handling for breaking changes (e.g., column type narrowing) via policies.

Fivetran provides data quality monitoring capabilities (details sparse in documentation) that track data freshness, row counts, schema changes, and sync errors. Customers can configure alerts for anomalies (e.g., unexpected row count changes, failed syncs, schema drift). Alerts are delivered via email or webhooks. Fivetran also tracks sync history and provides dashboards showing connector status, last sync time, and error logs. However, built-in data quality checks (e.g., null validation, referential integrity) are not explicitly documented.

Fivetran tracks data lineage automatically: which sources feed into which tables, which transformations process which tables, and which activations consume which tables. Metadata includes connector names, table names, column definitions, sync history, and transformation dependencies. Fivetran integrates with data governance catalogs (details sparse) to expose lineage and metadata. Customers can use this metadata for impact analysis (e.g., 'if I change this source, which downstream tables are affected?') and compliance reporting (e.g., 'which data sources feed into this sensitive table?').

Fivetran monitors sync health and provides alerts for failures, schema changes, and data anomalies. The platform tracks sync status (success, failure, partial), row counts per sync, and execution time. Users can configure email or webhook alerts for sync failures, and Fivetran automatically retries failed syncs with exponential backoff. The platform provides a dashboard showing connector health across all pipelines, with drill-down into sync logs and error messages. Fivetran also detects schema changes and alerts users to potential breaking changes.

Fivetran allows a single connector to load data into multiple destinations (data warehouses, data lakes, etc.) simultaneously, with independent sync schedules and transformation pipelines per destination. This enables teams to maintain multiple analytics environments (dev, staging, production) or serve different use cases (BI, ML, data science) from a single source connector. Data is loaded in parallel to all destinations, and Fivetran manages schema consistency across destinations.

Fivetran implements incremental loading strategies tailored to each source's capabilities: CDC (Change Data Capture) for databases with transaction logs, API-based delta detection (modified timestamps, cursors), and full-table reloads with deduplication for sources without incremental support. The system tracks the last sync state (high-water mark, cursor position, or transaction log LSN) and uses it to fetch only new/changed rows on subsequent syncs, reducing data volume, compute cost, and sync time. Deduplication logic handles late-arriving or out-of-order changes.

Fivetran integrates with dbt (data build tool) to orchestrate SQL-based transformations on loaded data. Transformations are defined as dbt models (SELECT statements) and run on a schedule (15-minute minimum on Standard, 1-minute on Enterprise) after data is loaded. Fivetran handles dbt project orchestration, dependency resolution, and execution on the destination database, eliminating the need for separate scheduling tools. Transformation results are materialized as tables or views in the warehouse, and Fivetran tracks lineage and execution history.

Fivetran's Activations feature (powered by Census acquisition) enables reverse ETL: pushing transformed data from the warehouse back to business applications (Salesforce, HubSpot, Marketo, etc.) via pre-built activation connectors. Activations use the same connector architecture as forward ETL, with built-in deduplication, upsert logic, and error handling. Data is synced on a schedule (15-minute minimum on Standard, 1-minute on Enterprise) and Fivetran tracks activation status, row counts, and errors. Activation costs are metered by Monthly Active Rows (MAR) pushed to destinations.

Fivetran's Managed Data Lake Service loads data into open-format data lakes (Apache Iceberg, Delta Lake) on cloud object storage (S3, GCS, Azure Blob) instead of traditional data warehouses. Data is stored in Parquet format with Iceberg/Delta Lake metadata, enabling schema evolution, time-travel queries, and ACID transactions. Fivetran manages partitioning, compaction, and metadata optimization automatically. Customers can query the lake using any SQL engine (Spark, Presto, Trino, Athena) without vendor lock-in to a specific warehouse.

Fivetran uses a usage-based pricing model metered by Monthly Active Rows (MAR) — the number of rows synced in a calendar month. Each connector has a per-MAR cost (e.g., Salesforce $0.00 for first 500K rows, then $0.0005 per additional row). Transformations are metered by Monthly Model Runs (MMR), and Activations by MAR pushed to destinations. Fivetran provides cost estimation tools and usage dashboards to track spending. Free tier includes 500K MAR connections, 3.5K MAR activations, and 5K MMR transformations; Standard and Enterprise plans have unlimited usage with per-unit pricing.

Fivetran provides role-based access control (RBAC) to manage who can view, edit, or execute connectors, transformations, and activations. Standard plan includes basic roles (Admin, Editor, Viewer); Enterprise plan adds custom roles with granular permissions (e.g., can edit connectors but not transformations). Fivetran integrates with identity providers (SCIM, SAML) for user provisioning on Enterprise+ plans. Audit logs track all user actions (connector edits, transformation runs, activation syncs) for compliance and troubleshooting.

Fivetran offers deployment flexibility for enterprise customers: Standard multi-tenant SaaS, VPN tunnels for hybrid deployments (Enterprise+ plans), and private networking (Business Critical plan) for air-gapped or highly-regulated environments. VPN tunnels allow Fivetran to access on-premises databases without exposing them to the internet. Private networking uses dedicated network paths (AWS PrivateLink, GCP Private Service Connect) to isolate traffic. Customer-managed encryption keys (CMEK) are available on Business Critical plan for data at rest and in transit.

Fivetran provides a Connector SDK (language and framework unspecified in documentation) that allows developers to build custom connectors for niche or proprietary sources and destinations. Custom connectors follow the same architecture as pre-built connectors, supporting incremental sync, schema detection, and error handling. Developers can publish custom connectors to Fivetran's marketplace or use them privately. The SDK abstracts authentication, pagination, rate limiting, and state management, reducing boilerplate code. However, Fivetran also offers a by-request program where Fivetran engineers build custom connectors for customers.