managed pipelines

Fivetran

Fivetran is a managed pipelines product that keeps warehouse tables in sync. This profile covers what it does, how it is priced, and where it stops being the right choice.

What the product actually does

Fivetran is a managed data pipeline service. You point it at a source — a SaaS application, a database, an event stream — and it creates and maintains the connectors, handles schema changes automatically, and lands the data in your warehouse. The selling point is that you do not operate any of it.

Automatic schema handling is the feature that justifies the price. When a source adds a column, Fivetran adds it downstream without a human noticing. When a source changes a type, Fivetran handles the migration. Teams that have maintained hand-written pipelines through a source API change understand what this is worth.

How pricing works once you are inside

Fivetran prices on monthly active rows, which is a metric designed to be fair and which is frequently surprising. A row counts as active in a month if it was inserted or updated in that month. A table where every row is touched nightly — a status field, a last-seen timestamp — bills as the full table every month, regardless of how little the meaningful data changed.

That single mechanic drives most Fivetran cost overruns. The standard mitigations are column hashing to suppress no-op updates, history-mode configuration to avoid re-syncing unchanged rows, and excluding high-churn columns you do not need. All of them require knowing the mechanic exists before the first invoice.

Where it breaks down

Custom sources. Fivetran has a large connector catalogue, but when your source is an internal API or an obscure vendor, you either wait for a connector or build one with the SDK — and a hand-built connector loses much of the automatic-schema advantage that justified the price.

Cost at scale is the other wall. The per-active-row model is comfortable at moderate volume and becomes very expensive at high volume, which is why teams frequently run Fivetran for SaaS sources and hand-built pipelines for their own databases.

What switching costs

Moderate. The data is in your warehouse, but the connector configurations, the schema-history behaviour and the state tracking are Fivetran's. A replacement has to re-establish all three, and getting state tracking wrong causes duplicates or gaps.

Who should not use it

High-volume teams with in-house platform engineers. At that scale the economics favour building, and the automatic-schema advantage matters less because you control both ends of the pipeline.

Sources

Fivetran public documentation and pricing page, read October 2026. Active-row behaviour described from the product's own billing documentation.