Fivetran Cost Optimization & MAR Management

Fivetran's pricing changed on March 1, 2025: monthly active rows (MAR) are now counted and priced per connection rather than pooled across the account, the Starter and Private Deployment plans were retired, and rows re-synced because of a Fivetran-side change became free. Further pricing updates followed in 2026. For a single-connector account the difference is small. For the typical account we see, with fifteen to sixty connections of very different sizes, it changed the bill materially, and usually upward.

The good news is that MAR is one of the most controllable line items in a data stack. Almost every account we audit is paying for rows nobody queries.

What we do

MAR audit. We pull usage per connection from the Fivetran usage dashboard and the Fivetran Platform Connector's metadata tables, and profile it by connection, schema, and table over the last three billing cycles. The output is a ranked list: which tables generate the MAR, how much of it is churn on high-update tables, which connections are paused, idle, or duplicated, and whether any re-syncs were billed that should have been free.

Levers, applied in order of payback.

• Sync frequency. A source that updates nightly does not need a fifteen-minute sync. Lowering frequency on cold sources is the cheapest win and rarely affects a dashboard anyone looks at.
• Column and table deselection. Audit logs, event tables, and wide CRM objects often account for most of a connection's MAR while feeding nothing downstream. We trace lineage to your dbt models and BI before deselecting anything.
• Splitting or consolidating connections. Under connection-level pricing, the shape of your connections matters. Sometimes a noisy table belongs in its own connection with its own schedule; sometimes three small connections to the same source should be one.
• Re-sync hygiene. Full re-syncs on large tables are the classic surprise invoice. We set guardrails on who can trigger them and verify that Fivetran-initiated re-syncs were not billed.
• Dead connection cleanup. Paused connections do not cost MAR, but forgotten active ones do. Dev and sandbox connections left running against production-sized sources are the usual culprits.
• Destination choice. For some workloads the Managed Data Lake Service changes the cost profile; see our managed data lake page.

Forecasting and renewals. We model MAR growth from your source growth rates so that your renewal commitment is based on a forecast, not last month's spike. If you are mid-contract, we identify what can be cut before the true-up.

Monitoring. We leave behind a MAR dashboard built from the Platform Connector's incremental_mar data and alerts for week-over-week MAR jumps per connection, so the next surprise is caught on day two rather than on the invoice.

What a typical engagement looks like

1. Week 1: audit. Read-only access to your Fivetran account and destination. We deliver the ranked MAR report and a savings estimate per lever.
2. Week 2: changes. We apply the agreed changes with your team, connection by connection, with a rollback plan for each.
3. Week 3: verify. We compare the next usage cycle against the baseline and hand over the monitoring.

Most accounts recover the cost of the engagement inside the first billing cycle after the changes land. We do not promise a percentage because every account is different, and we will tell you early if we do not think there is much to find.

Free MAR audit

If you share a usage export, we will send back a one-page summary of where your MAR is going and what we would cut first, at no charge. Contact us to get started, or read our tutorial on connection-level MAR pricing for the background.