What Fivetran does
Pick a connector, set a destination, let it run. Fivetran pulls data from source systems, writes it to the warehouse and follows schema changes automatically. No loader of your own, no maintenance window when an API renames its fields.
Since the Census acquisition the return path runs through the same platform, branded Fivetran Activations. Ingestion and activation now sit with one vendor.
If you're weighing the alternatives inside a marketing context, the funnel.io review covers that ground. This is about Fivetran itself.
The dbt Labs merger, and what it actually changes
On 1 June 2026 Fivetran and dbt Labs closed the merger they announced in October 2025. George Fraser runs the combined company, Tristan Handy is President. Both products keep running independently for now, so existing pipelines and models are untouched.
For an architecture decision it still matters. The standard advice ran like this for years: Fivetran loads raw, dbt models on top, and because the two came from different vendors the transformation layer stayed independent of the loader. That argument is gone.
What survives hangs on one detail: dbt Core is open source, and version 2.0, released alongside the merger, ships the Fusion engine under an open licence too. Your models stay portable even though the vendor is now the same one running the loader. The question shifts from "do we own the transformation layer" to "what happens to the price once both line items sit on one invoice".
That isn't a warning, it's an observation with a task attached: if you run both tools today, look at your next contract cycle earlier than usual.
How we scored it
Four axes, the same ones every tool in our catalogue gets.
Ecosystem 5.0. More than 750 sources, over 200 destinations, plus reverse ETL from the same vendor. Nothing in the managed segment matches that breadth.
Reliability 4.5. This is what teams actually buy. Pipelines run, schema changes get followed, and nobody checks on Monday whether the sync completed.
Value for money 3.0. The weakest axis, for one reason: you cannot work out the price in advance. More on that below.
Ease of use 4.5. Click a connector, choose a destination, done. The half point comes off where sync modes and cursor fields still want manual attention.
What MAR pricing really means
Billing counts monthly active rows, meaning rows that change during the billing month. Inserted and updated rows count. Unchanged rows don't, and neither does the first sync of a source.
That sounds fair, and for many sources it is. The catch sits in tables that get touched constantly: campaign statistics corrected retroactively, a CRM recalculating fields nightly, a shop system with a status change on every order. The same row goes active month after month.
Four plans are public: Free, Standard, Enterprise and Business Critical. Fivetran names a price for none of them. Here is what the pricing page does publish, as of 15 August 2026:
| What is published | Value |
|---|---|
| Free plan | 500,000 MAR for connections, 3,500 MAR for activations |
| Base charge per standard connector | $5 per month between 1 and 1M MAR, not on the free plan |
| Price curve | separate per connector, the rate declines as volume rises |
| Annual contract | up to 22 percent discount |
| Published worked example | Facebook Ads at 34,479 MAR, $17.23 list price per month |
That worked example deserves a second look. $17.23 for one ad source is cheap, and it is meant to read that way. The amount then scales along a curve you only see in production, and the total across twenty sources including CRM and shop appears nowhere.
Hence the advice we give every team: take the free plan, connect your two or three highest-volume sources, measure for a month. After that you negotiate with numbers instead of estimates.
Funnel harmonises schemas before write, what lands in the warehouse is already unified. Shortest path to the AI/BI layer, but lock-in on Funnel's mapping.
Where Fivetran wins
When sources reach beyond marketing. ERP, product database, support system, finance. Once those need to land in the same warehouse, a marketing hub is the wrong tool and Fivetran's breadth pays off.
When nobody wants to mind pipelines. Operations is the real reason to buy. Measured against half a data engineering role, the licence is often the cheaper line item, even at uncomfortable MAR numbers.
When governance has to be provable. Versioned models, tests, lineage, an audit trail back to the source. In regulated industries that isn't comfort, it's a precondition.
Where it hurts
The price cannot be determined up front. No list price per plan, a curve per connector, a discount grid for annual contracts. For a business case that needs approval before purchase, this is the hardest part.
Volatile tables cost disproportionately. Retroactive corrections in campaign data are the classic case. Set sync frequency and scope carelessly and you pay for a volume of data nobody reads.
For pure marketing it is too much tool. Eight ad accounts, one shop, one newsletter platform: that doesn't need a platform that also connects SAP.
The vendor landscape has narrowed. Loader, transformation and activation now sit with one company after Census and dbt Labs. Operationally convenient, one negotiating position fewer.
Fivetran vs funnel.io vs Airbyte: which fits when?
- If your sources are almost entirely ad platforms and the goal is marketing reporting, then → funnel.io. Harmonisation is built in there, Fivetran delivers raw. The direct head-to-head: funnel.io vs. Fivetran.
- If data from several domains converges and a data team owns the models, then → Fivetran. This case is what justifies the price level.
- If you need many niche sources or self-hosting in the EU is a given, then → Airbyte. No licence, but you run it.
- If you have two or three Google-native sources, then → load straight into BigQuery. The Data Transfer Service costs nothing.
- If your main need is pushing back out of the warehouse into operational tools, then → compare Fivetran Activations against Hightouch. Since the Census acquisition that is a real choice rather than a default.
Transparency
We are not a Fivetran partner and take no commission on referred licences. The verdict above comes out of architecture work on setups where Fivetran was on the shortlist, some of which it won and some of which it lost to funnel.io or a purpose-built loader.
Our partnerships are listed in the imprint, so you can see where we have an interest and where we don't.
What to decide it on
- How many of your sources sit outside marketing? At zero or one, a marketing hub is the cheaper answer. From three upwards it tips towards Fivetran.
- Which tables change daily? That list drives the invoice, not the number of connectors. Take the three biggest and measure a month on the free plan.
- Who owns the transformation layer? Without someone maintaining dbt models you are paying for governance nobody uses. A harmonising tool is the more honest choice then.