Radical Cost Transparency: Moving Away from MAR

If your team uses data pipelines, you’re probably familiar with the following conundrum: you expect costs to go up over time, but you have no idea how much they will actually cost. What will happen when more rows arrive each month? What about when new datasets are added, real-time syncs are enabled, or more systems are connected?

This article will explain why pricing models like Fivetran’s MAR (Monthly Active Rows) often miss the bigger picture and don’t scale as you grow. It will also introduce Estuary’s volume-based model as an alternative that keeps costs predictable.

The Hidden Complexity of Monthly Active Rows

MAR uses Fivetran's internal data model rather than the source system layout. So, when data is reorganized, a single record can turn into several rows. And the worst part is, you might not notice this while building pipelines, but you’ll definitely spot it on your bill.

Over time, your systems evolve. You install updates, CRM data changes, deals move through different stages, and product events keep coming in. Each of these changes increases your MAR count.

When Growth Becomes Expensive

Imagine a growing software-as-a-service (SaaS) company that brings together customer relationship management (CRM) data, financial records, and product events.

At first, MAR-based pricing seems logical. Let’s say you need about 2 million MARs each month. You’ll probably pay between $700 and $2,667 (the exact amount will depend on the connector you pick and its complexity). However, as usage grows and you add more integrations, the row changes start to pile up. When you edit a contact in HubSpot, more than one record changes, and a single edit can ripple out to deals, lifecycle stages, custom properties, and engagement logs.

This is particularly true for streaming or CDC workloads. Event-driven systems are designed to send frequent updates, but when each unique event incurs a cost, those real-time pipelines can become expensive fast.

At around 10 million MARs, fees can exceed $10,000 monthly, and big projects can easily amount to $100,000 or more a year. Some customers even say Fivetran costs about five times more than Estuary at the same scale.

What Real Teams Are Saying

If you take some time to read public discussions held among data engineers, you'll spot a clear pattern.

One Reddit user has published a story about a NetSuite finance workflow that updated millions of historical transactions. This internal process caused a spike in MAR usage, burning through three months of annual credits almost instantly. Nothing was broken, though; the system simply generated more row changes than expected.

Another team has shared that reloading Pendo data through Fivetran would have cost them around $30,000, so rather than forking out this amount of money, they rebuilt the pipeline themselves for approximately $3,000. While using Fivetran, they also discovered issues that caused more than a billion rows to vanish (not to mention the pricing plan that kept changing).

BI consultant Justin Butlion has also spoken up about the new pricing on LinkedIn. According to him, renewal quotes for a client with over 30 connectors came in at two to three times the initial contract value. And while he does consider the product to be reliable, he admitted he was taken aback by how quickly the price increased in comparison to other options.

To be fair, some engineers still believe this tool is a good choice for smaller teams since it saves time and cuts maintenance. However, the same pattern often emerges: when row activity grows faster than expected, costs increase quickly.

2026 Pricing Update: When Deletes Start to Count

In 2025, Fivetran changed its pricing so that discounts apply to each connector rather than the whole account. This made forecasting more difficult for teams with many connectors, since they could no longer rely on a single volume discount.

The 2026 update has gone even further.

Starting January 1, inserts, updates, and deletes all count toward paid MAR. At first, this may seem minor, but it really isn’t.

Take deletes, for example. They are very common in real systems: users churn, records are cleaned up, products are archived, and CRM entries are merged or removed. And just like that, all this activity now contributes to your MAR count.

What’s even worse is that if you use history mode, multiple updates made within the same month count toward paid MAR. Therefore, your usage may increase if a row changes more than once a month.

And that’s not all: now, there is also a tiny minimum fee for each connection. For example, a base fee (such as $5) applies if a standard connection generates between 1 and 1 million MAR per month. And for teams using numerous small connectors, the fees can add up.

When combined, these updates reveal several key points:

• Billing now includes a wider range of row activity.
• Running small connectors is no longer almost free.
• Forecasting requires a clearer understanding of how data churns over time.

None of this makes MAR pricing inherently bad, but it does make costs more sensitive to the behavior of your data. Plus, with pricing rules changing year after year, it’s hard not to worry about the unexpected ways your bill might evolve as your systems grow.

Here’s a quick overview of Fivetran pricing changes so you could get a clearer picture.

Estuary as an Alternative: The Price Depends on How Much Data You Use

Instead of charging you for each active row, Estuary charges based on how much data you move.

The idea is simple. You pay $0.50 for each GB that moves, no matter where it’s headed. Each connector costs $100, for the first six, and every additional one costs $50.

There are no row multipliers or penalties for latency. The number of normalizations stays the same, and you only pay for the data you use, no matter how complex it is or how often it changes.

Compared to other ELT and ETL vendors, this is one of the least expensive pricing models.

Rivery, for example, charges between $7.50 and $12.50 per GB, depending on the plan. If you move around 10 GB each month, Estuary costs less. Once you reach 1 TB of data usage per month, the cost is about one-tenth of what most other companies charge.

A Practical Cost Comparison (MAR vs. GB)

Let's compare the real-world costs of MAR and GB.

Here’s the situation: each month, 800 GB of data are moved, and there are 2 connector instances. Estuary charges $0.50 per GB to ingest your data and another $0.50 per GB to materialize it. The captured-to-materialized ratio isn’t always exactly 1:1, but for a single source to a single destination, you can generally expect about $1 per GB in total.

With that in mind, moving 800 GB at $1 per GB for ingestion and materialization comes to $800. The two connectors at $100 each add another $200, which brings the monthly total to $1,000.

Now, imagine that same workload under the MAR pricing. It’s common for a system to update tens of millions of rows each month when product events are combined with CRM activity. This can easily drive costs to $15,000 to $20,000 per month.

That’s quite a difference, and the gap gets bigger as activity increases.

With Estuary, the math stays simple no matter how big you get. You decide how much data you need, then you multiply and add connector instances.

That's all there is to it.

Capture Data Once, Manage Its Delivery, and Keep Costs Under Control

Modern pipelines don't just move data from one place to another; they send it to many places at once. For example, one operational database can send data to a data lake like Snowflake, BI tools, and even search systems at the same time. This many-to-many setup is now common.

With Estuary's streaming architecture, you capture the data once and store it in your cloud bucket. From there, you can send it to many places without pulling it back each time you add a new destination. You only pay for the data transferred to each source and target. This setup makes it easier to predict both costs and effort as the size of your project grows.

Cost management doesn’t stop at ingestion. Real-time pipelines can spike warehouse computing costs if no one is monitoring them. For example, if lots of small updates flow into Snowflake, your warehouse instances might be busy processing them, driving up your bills along the way.

Estuary separates the continuous CDC capture from the warehouse materialization. Data can flow into Snowflake as it happens, but you choose when to make it available for use. For instance, you can sync only during business hours, so the warehouse doesn’t run when it's not needed.

This simple approach gives you more control: capture data once, deliver it when it makes sense, and stop your warehouse from costing you more than it should.

Growth That's Easy to Predict (and Easy to Get Into)

With volume-based pricing, forecasting is simple. Use twice as much, and you can estimate how much the data volume will grow. Add a new source, and it’s easy to anticipate how much data it will send. Turn on real-time sync, and the price per GB stays the same.

With Estuary, you also:

• Get a free tier with 10 GB of data per month and up to two connector instances
• Try the full version of the Cloud for 30 days
• Pay for how much you use each month, without vendor lock-in

This lets you test real-time pipelines without signing long-running contracts first.

You don't have to guess how nested data structures will flatten into normalized rows, nor do you need to explain how the internal row accounting logic works. You're just measuring how much data you move.

Conclusion

MAR pricing works well in small, low-change environments. However, pipelines are becoming increasingly complex. They now handle streaming workloads, change data capture (CDC), multi-connector setups, and evolving schemas. As a result, it’s tough to predict and manage costs.

Estuary is different in this regard. It charges based on how much data you use. The connector prices are simple, and scaling up doesn't trigger hidden multipliers, surprise charges, or delays. With predictable costs, teams can build reliable, real-time pipelines without worrying about the next invoice.