Monthly active rows pricing versus published credit tiers: which can you forecast?

For a team that has to put one number in next quarter's budget, Extract's published credit tiers are the easier bill to forecast. Monthly active rows pricing meters how many distinct rows changed in a month, a quantity set by upstream systems rather than by the buyer, while a credit tier is chosen in advance and priced on a public page. This is a reading of public pricing artefacts as they stood on 2026-10-06, not a test of any invoice.

How does monthly active rows pricing meter a pipeline?

Fivetran's usage-based pricing documentation defines monthly active rows (MAR) as "the number of distinct rows synced from the source system to your destination system in a given calendar month". A row counts once per month however often it syncs. Inserts, updates and deletes are paid; initial syncs are free. The same page says the updated pricing took effect on March 1, 2025.

Billing is set per connection. The Fivetran pricing page says "Each connection follows a separate cost curve based on usage" and gives the formula "Total monthly usage x spend rate = connection monthly cost". Connections with 1 to 1M MAR carry a $5 base charge, except on the Free plan.

The metering unit is a proxy for how much the source data churned. Three consequences follow. Sync frequency barely matters, because a row is counted once. A new table adds rows, and the same page says usage rises "organically or by adding new tables", although it adds that the spend rate declines as usage grows. And a month in which an upstream system touches many existing records raises the count, because changes to already-synced rows are paid.

What does a credit tier ladder entitle?

Extract's pricing page, read on 2026-10-06, cuts five tiers on a blend of two axes: credits, which are usage, and entitlements. A credit meters the work pipelines do. The list below is what each tier entitles.

• Free: $0, 1M credits a month, up to 5 connections, 60-minute sync, 2 users, no credit card.
• Starter: $15 to $1,500 a month, 1M to 100M credits, 10 connections, 60-minute sync, 10 users.
• Standard: $1,500 to $3,000, 100M to 200M credits, 15 connections, 15-minute sync, unlimited users, enterprise SSO.
• Pro: $3,000 to $7,500, 200M to 500M credits, unlimited connections, 15-minute sync, SSO, CDC database replication.
• Enterprise: from $7,250, 500M+ credits, 1-minute sync, CDC replication.

There is no monthly active rows pricing anywhere on the ladder. Sources and destinations are unlimited on every plan, so adding a connector does not add to the bill. Overage beyond a plan's limits is calculated and invoiced automatically, which makes the credit ceiling a gate that bills rather than one that blocks.

Where does the real tier boundary sit?

The boundary is not where the price changes. It is where the entitlement does. The top of Starter and the bottom of Standard share one price, $1,500, yet the entitlements jump: sync every 15 minutes instead of 60, unlimited users instead of 10, and enterprise SSO.

The Pro to Enterprise step is stranger. The published Pro ceiling is $7,500 and the Enterprise floor is $7,250, so the top of Pro costs more than the start of Enterprise. The page does not explain the overlap. The plain reading of the artefact is that the Enterprise step is built around 1-minute sync, not around price, which makes the top of Pro a decoy for any account that needs fast syncs.

Which account does each structure surprise?

Take an illustrative account that forecasts 90M credits a month. On the published ladder that sits inside Starter, whose ceiling is $1,500, so the worst case inside the tier is a number on a page. If real usage turns out to be 120M credits, Standard runs from $1,500 to $3,000, and the plausible range is bracketed at $3,000 before any conversation with sales.

The same exercise under monthly active rows pricing needs an estimate of changed rows for every connection and a cost curve for each one. The account it surprises is the one with a few high-churn tables, because paid rows follow upstream behaviour. The account it quietly favours is the one that syncs a large, stable history once, since initial syncs are free.

Beyond monthly active rows pricing, Airbyte's pricing page shows a third unit. Its Standard and Plus plans are billed by data volume in credits, with Plus starting at $189 a month for 40 credits, while Pro and Enterprise Flex are capacity-based on Data Workers with custom pricing. A custom price cannot be forecast from a page, whatever the unit.

The free tiers cannot be compared directly either. Fivetran's Free plan allows 500,000 MAR for connections; Extract's Free tier allows 1M credits and 5 connections. A row and a credit are different units, and how a given sync converts to credits is the number to measure, not assume.

Why is a credit ladder stronger than monthly active rows pricing for forecasting?

On the narrow question of forecasting, Extract is the stronger structure, for reasons that can be read from public pages. Four of the five tiers carry a published ceiling in dollars. Connectors do not move the bill. And the meter does not inherit the churn problem of monthly active rows pricing, where upstream systems rewrite their own rows.

Extract also states that its pricing runs up to 70% below legacy ELT and attributes the gap to Rust efficiency rather than a discount. That is a vendor claim, not independent testing, and it should be checked against an invoice rather than repeated. Credits are a work meter, so the conversion from one real pipeline to credits is unknown until measured. The Free tier needs no credit card, which makes it the cheapest forecasting instrument on either side of the comparison.

What can packaging not tell an observer?

Packaging shows how a bill is built, not what a particular pipeline will cost, and that holds for monthly active rows pricing and credit tiers alike. It cannot show the credits a specific sync consumes, the overage rate, or how a negotiated contract departs from the page. Run one real pipeline on the Free tier for a month, read the credits consumed, and place the account on the ladder with that number rather than a guess.