Fivetran

Managed connectors that copy data from SaaS apps, databases and files into your warehouse or lake, billed by the rows that change each month.

What it is: ETL & ELT pipelines
Type: Commercial product
Made by: Fivetran + dbt Labs
License: Proprietary; the Python Connector SDK is MIT
Facts checked: 6 October 2026

What it is

Fivetran is a hosted service that copies data from sources into a destination you own. You pick a connector (a SaaS API such as Salesforce, a database, files), give it credentials, and choose a destination: Snowflake, BigQuery, Databricks, PostgreSQL and others, or its Managed Data Lake Service, which writes Parquet files with Iceberg metadata to S3, ADLS or GCS. The first sync copies all history. After that, incremental syncs run on a schedule (every 6 hours by default, every minute on the top plans) and fetch only what changed, from API cursors or, for databases, the change log, such as PostgreSQL’s write-ahead log. Fivetran keeps the cursors and checkpoints, creates and alters destination tables as the source schema changes, and adds columns such as _fivetran_synced and _fivetran_deleted. Transformation happens afterwards in the warehouse, usually with dbt. With Hybrid Deployment, the data is processed inside your network while Fivetran’s cloud orchestrates it.

Pricing is measured in Monthly Active Rows (MAR): the distinct rows, by primary key, inserted, updated or deleted in a calendar month, counted per connection and table. A row that syncs many times in a month counts once, and history from the initial sync is free. The price per MAR falls as a connection’s volume grows, and since March 2025 those tiers apply per connection rather than across the account. Fivetran’s own Standard-plan examples, checked 6 October 2026: a Google Ads connection at a median 88,240 MAR costs $48.67 a month, Marketo at 847,574 MAR costs $424.54. Fivetran merged with dbt Labs on 1 June 2026, and both brands continue.

Why teams pick it

More than 700 managed connectors: you configure them instead of writing and running extraction code
Log-based CDC for databases (PostgreSQL logical replication, and HVR for large Oracle and SAP replication)
Handles schema drift: new columns and tables appear in the destination unless you block them, and soft-delete and history modes keep deleted rows and old versions
Initial historical syncs don’t count as paid MAR, and the Free plan covers 500,000 MAR a month
Writes to every major warehouse or to Iceberg tables in your own object storage, and Hybrid Deployment keeps processing inside your network

Watch out for

Bills are hard to predict: a job that touches every row of a big table makes all of them active for the month, and since January 2026 deletes and every history-mode version count too
Volume discounts build up per connection since March 2025, and each connection with 1 to 1 million MAR carries a $5 base charge, so many small connections cost more than their rows suggest
Reverse ETL is priced far higher per row: Fivetran’s median example is $202.25 a month for 12,007 activated MAR (Snowflake to Salesforce), against $22.06 for 34,479 MAR of Facebook Ads
Closed source with fixed connectors: unusual sources need the Python Connector SDK, and leaving means rebuilding every pipeline elsewhere
Fast syncs are plan-gated: Standard goes down to 15 minutes; 1-minute syncs and cron schedules need Enterprise or Business Critical

Alternatives

What all etl & elt pipelines have in common →

AirbyteCommercialAn ELT platform you can host yourself: hundreds of connectors, each a Docker image speaking one JSON protocol, scheduled by services on Kubernetes.

Apache NiFiOpen sourceA Java dataflow server you program by drawing a graph of processors in a web UI, with buffered queues, back pressure and a record of where every piece of data went.

ArtieCommercialA managed change data capture service that copies every insert, update and delete from production databases into a warehouse, aiming for under a minute of delay.

BruinOpen sourceOne Go command-line tool that ingests data, runs SQL and Python transformations and checks the results, with each step a file in your repo.

DagsterOpen sourceA Python data orchestrator built around assets: you declare the tables, files and models that should exist, and Dagster works out what to run.

dbtOpen sourceSQL `select` statements in git, built into warehouse tables in dependency order, with tests and docs alongside.

EstuaryCommercialA managed data-movement platform that captures changes from databases and SaaS sources into collections kept as JSON files in your own bucket, then keeps warehouses and other systems in sync, in real time or on a schedule.

MeltanoOpen sourceA command-line ELT tool that runs Singer taps and targets from a `meltano.yml` project kept in git.

OLakeOpen sourceAn open-source tool in Go that copies databases, Kafka topics and S3 files into Apache Iceberg tables (or plain Parquet), with full loads, incremental syncs and CDC, and no Kafka, Debezium, Spark or Flink in between.

How it fits with other technologies

What Fivetran uses

What uses Fivetran

A custom connector with the Connector SDK: Fivetran stores the cursor and merges rows · python

# connector.py  (pip install fivetran-connector-sdk requests)
import requests
from fivetran_connector_sdk import Connector, Operations as op

def schema(configuration: dict):
    return [{"table": "invoices", "primary_key": ["id"]}]

def update(configuration: dict, state: dict):
    cursor = state.get("updated_since", "1970-01-01T00:00:00Z")  # state is empty on the first sync
    resp = requests.get(
        "https://api.billing.example.com/v1/invoices",
        headers={"Authorization": f"Bearer {configuration['api_key']}"},
        params={"updated_since": cursor, "order": "updated_at"},
        timeout=30,
    )
    resp.raise_for_status()
    for invoice in resp.json()["data"]:
        op.upsert(table="invoices", data=invoice)  # merged on the primary key
        cursor = max(cursor, invoice["updated_at"])
    op.checkpoint(state={"updated_since": cursor})  # the next sync resumes here

connector = Connector(update=update, schema=schema)

# fivetran debug   -> runs locally into files/warehouse.db (DuckDB)
# fivetran deploy --api-key "$FIVETRAN_API_KEY" --destination warehouse --connection billing_invoices

How the bill is built#

List prices aren’t published per MAR; the pricing page shows examples and a calculator. These are the rules that decide what you pay.

Monthly Active Rows

the meter

Each distinct primary key inserted, updated or deleted in a month, tracked separately per connection and table. A row counts once a month however often it syncs. Tables without a key get a _fivetran_id computed from the row’s values.

Per-connection tiers

since March 2025

The rate per MAR falls as volume grows, but the tiers apply to each connection on its own, not to the whole account. Fivetran said this would be a price cut for most customers; accounts with many small connections lose pooled discounts.

Base charge

$5 per connection

Every standard connection with 1 to 1 million MAR in a month costs at least $5, except on the Free plan.

Deletes and history mode

since January 2026

Deletes count as paid MAR. In history mode, which keeps every version of a row, repeated updates to the same row in one month each count.

What’s free

free

Initial historical syncs, re-syncs of rows identical to the destination, history backfilled when you add a column, and connection trials.

Transformations

per model run

The first 5,000 model runs a month are free on every plan; then $0.01 per run up to 30,000, $0.007 up to 100,000 and $0.002 beyond.

Free plan

500,000 MAR for connections, 5,000 model runs and 3,500 MAR of reverse ETL (Activations) a month.

Annual contracts

from 5% off

Committing to a year gives automatic discounts that grow with the annual list price: 5% at first, rising to 22.6% for contracts above $145,340 a year at list price.

Fivetran’s example list prices#

Median usage and median monthly list price for real connections, as Fivetran publishes them.

Connection	Median MAR	Median monthly list price
Google Analytics 4	21,993	$15.88
Facebook Ads	34,479	$22.06
Google Ads	88,240	$48.67
Marketo	847,574	$424.54
Activation: Snowflake to HubSpot	9,934	$197.96
Activation: Snowflake to Salesforce	12,007	$202.25

Estimates for the Standard plan at a company with 1 to 200 employees, from fivetran.com/pricing, checked 6 October 2026.

Other ways to load data#

Airbyte

source-available and cloud

Connectors for APIs, databases and files into warehouses and lakes, self-hosted or as a cloud service.

Meltano and Singer

MIT

Taps and targets that pass JSON over a Unix pipe; Meltano runs and configures them as code.

Estuary and Artie

managed CDC

Streaming change data capture from databases into warehouses with low latency, for when hourly batches are too slow.

Cloud-native options

one cloud

AWS DMS and zero-ETL integrations into Redshift, Google Datastream and BigQuery Data Transfer Service, and Fabric Data Factory copy jobs and mirroring.

See it in a real stack

No architecture uses Fivetran yet. Drop it into the playground and wire it up yourself.Open playground →

Where it came from2012–2026

George Fraser and Taylor Brown, friends since childhood, founded Fivetran in 2012 and went through Y Combinator’s winter 2013 batch with a data analysis tool for scientists. In 2014, responding to what customers needed, they built their first connector. The bet that followed was that moving data should be a utility: managed pipelines from each source that copy everything into a cloud warehouse, where analysts transform it with SQL afterwards.

Cheap cloud warehouse storage made copying raw data reasonable, and Fivetran priced by the rows that change each month. In 2021 it raised $565 million at a $5.6 billion valuation and bought HVR for $700 million, adding high-volume database replication for systems such as Oracle and SAP. In 2025 it agreed to buy Census, a reverse ETL company, and bought Tobiko Data, maker of SQLMesh and SQLGlot. That October it agreed an all-stock merger with dbt Labs, its partner for a decade; the deal closed on 1 June 2026, with Fraser as CEO and dbt’s Tristan Handy as president.

Why it’s built this way

Fixed, managed connectors instead of a toolkit: Writing and maintaining extraction code for every SaaS API was the most time-consuming part of a pipeline. Fivetran builds and runs each connector, and customers only configure it.
Load raw, transform later (ELT): Storage became cheap enough to keep untransformed copies, so Fivetran loads source tables as they are and leaves modelling to SQL in the warehouse.
Bill by rows changed, not by data volume: Monthly Active Rows count each changed row once a month, and history from the initial sync is free, so the price follows how much the source changes rather than how big it is.
Follow schema changes automatically: Source APIs add fields all the time. By default new columns and tables are created in the destination, so pipelines keep running instead of failing on schema drift.

Timeline

2012
George Fraser and Taylor Brown found Fivetran
2013
Joins Y Combinator’s winter batch
2014
Builds its first connector; opens headquarters in Oakland
2021
Raises $565M at a $5.6B valuation; buys HVR for $700M
2025
Per-connection MAR tiers; buys Census and Tobiko Data; agrees dbt Labs merger
2026
Deletes billed as MAR from January; dbt Labs merger completes on 1 June

Sources27 · facts checked 6 October 2026

Improve this page on GitHub