Fivetran’s pricing model changed in March 2025, shifting from account-wide to per-connector Monthly Active Row (MAR) pricing. Before this change, buying managed ELT was broadly cost-effective: 50-100 hours per custom connector, 44% of engineer time lost to pipeline maintenance (Wakefield Research), versus a predictable monthly bill at $150/hour engineering cost. The March 2025 change altered that calculus for many teams.
What Changed
The shift from account-wide to per-connector MAR pricing altered how costs scale.
Bulk discounts disappeared. Under the old model, your total MAR volume across all connectors determined your pricing tier. High-volume accounts got discounts that made individual connectors cheaper as you added more. The new per-connector MAR tiering means each connector has its own pricing tier. Adding connectors no longer benefits from volume discounts.
Cost increases hit immediately. Reddit users reported 4-8x cost increases after the change. One user described going from $20/month to $2,000/month as their data volume grew. Teams with many connections reported 70% cost increases. 35% of recent G2 reviewers cite cost as their primary concern with Fivetran.
The minimum floor rose. The minimum annual contract sits at $12,000. That’s before any data actually flows. For small and mid-size teams, the floor alone may exceed what their pipeline infrastructure should cost.
Why Marketing Data Gets Hit Hardest
MAR pricing charges you for every row that changes in a billing period. Marketing data changes constantly, and in ways that make MAR pricing particularly punishing.
Ad metrics update retroactively. Platforms like Meta recalculate attribution over 3-7 day windows. Yesterday’s click count will be different tomorrow. Each retroactive update counts as an active row, even though the underlying event hasn’t changed — the platform just revised its attribution math.
Attribution windows shift. When a conversion gets attributed to an ad click from last week, the row representing that ad’s performance updates. Every attribution recalculation across every campaign generates active rows.
Campaign performance data refreshes daily. Daily spend, impression, and click metrics for active campaigns change every sync. A modest account running 50 campaigns across 5 ad groups each generates hundreds of rows that update on every sync cycle. A large account with thousands of active campaigns can produce enormous MAR counts.
Granular data multiplies the problem. If you’re syncing ad-level or keyword-level performance data (which most analytics teams need for optimization decisions), the row counts explode. Account-level summaries have manageable MAR; the granular data that actually drives decisions does not.
The result is that marketing data — one of the most common use cases for managed ELT — became one of the most expensive categories under the new pricing. What looked like a reasonable expense at $500/month can become a $5,000/month line item once your campaigns scale or your sync granularity increases.
The Predictability Problem
The deeper issue isn’t just cost — it’s cost unpredictability. Under the old model, you could estimate your Fivetran bill with reasonable accuracy. Under per-connector MAR, your bill depends on:
- How many rows each source modifies in a given month
- How often your syncs run (more frequent syncs = more MAR for retroactively updating sources)
- How granular your data is (ad-level vs. campaign-level)
- Whether any source had an unusual data event (a bulk update, a schema migration, a retroactive recalculation)
None of these are fully within your control. A source system’s decision to reprocess historical data can spike your bill without warning. This transforms Fivetran from a predictable operational expense into a variable cost that requires active monitoring.
For teams that chose managed ELT specifically to avoid the operational overhead of building pipelines, having to monitor and optimize their ELT tool’s billing model is an ironic outcome. The “hidden cost of buying” has become very visible.
What This Means for the Build-vs-Buy Calculus
The traditional argument for buying rested on cost predictability and the high cost of engineering time. Both pillars have weakened:
-
Cost predictability is gone. The whole point of paying for managed ELT was avoiding surprise infrastructure costs. Per-connector MAR pricing reintroduces exactly the kind of variable, hard-to-predict cost that managed tools were supposed to eliminate.
-
The comparison baseline shifted. The $520,000 annual pipeline maintenance cost cited in traditional analyses assumed engineers spending weeks per connector. With AI-assisted development and mature open-source tools, the engineering time per connector has dropped dramatically.
-
The breakeven calculation changed. When Fivetran was $500/month for marketing data and a custom connector took 100 hours to build, buying won easily. When Fivetran is $2,000-5,000/month and a custom connector takes 10-20 hours with AI assistance, the math reverses for many teams.
Fivetran remains cost-effective for certain scenarios — compliance requirements, non-technical teams, extreme connector breadth. For marketing data teams loading to BigQuery, the per-connector MAR model has shifted the build-vs-buy breakeven point significantly.