Destination-Based Pricing vs Row-Based Pricing for Data Export
Row-based billing taxes growth, but destination pricing has its own traps.

Choosing between destination-based and row-based pricing is a bet on how your data volumes will move over the life of a contract. Row-based pricing wagers that a buyer's data change rate stays predictable and stable over time. Destination-based pricing wagers that growth in data volume shouldn't automatically grow the bill. The data integration market is heading toward $33.24 billion by 2030, and that growth means more teams are signing these contracts, at higher stakes, than at any point before. Each vendor meters something different (rows, events, credits, compute-hours, destinations) and the unit advertised at the lowest price is rarely the one that stays cheap once a real workload hits it.
Where row-based pricing's assumptions break down
Row-based pricing looks simple on a pricing page, but it hides several distinct metering choices, and most buyers only learn which one governs their contract after signing it. Monthly Active Rows, or MAR, counts a row once per month no matter how many times it's inserted, updated, or deleted. That makes MAR forgiving for stable, append-only sources, and expensive for high-churn tables like orders, sessions, or inventory, where the same row gets touched repeatedly. Event-based billing takes a different approach: every write to the destination counts as its own billable unit, including repeat updates to the same row. That model is easier to reason about on paper, but it costs more than MAR once a workload involves heavy change-data-capture traffic from operational sources. Stitch built its business on a third variant: flat monthly tiers with a fixed row allowance, stepping up at breakpoints, starting around a modest rate for lower volumes and climbing to a Premium tier for very high row counts.
None of these models are static once a contract is signed, and three changes in particular catch buyers off guard. If MAR is calculated per connection rather than pooled across an account, teams that run many mid-size connectors lose their volume discounts. After this change took effect in March 2025, increases of 40 to 70 percent were common for multi-connector setups. Starting in January 2026, deletes began counting toward paid MAR totals, which makes high-churn tables materially more expensive overnight, with no change in actual business activity behind the jump. History Mode, which tracks every change to a row as a distinct event, erodes the central advantage MAR was supposed to offer: counting a row once no matter how often it changes. G2 reviewers have flagged real cost increases tied to these shifts, especially among teams running many smaller connectors. One reviewer described the per-connector MAR model as something that "makes it impractical for event sources".
The three workload triggers that make row-based costs compound non-linearly

Row-based costs don't track business growth in a straight line. They inflect sharply upward when three common scaling patterns hit at once, and for a growing product, they almost always do. The first trigger is connector growth. Adding new data sources is the normal path for a maturing data team, but connector-level MAR billing restarts the pricing clock with every new source and offers no pooling benefit across the account. That structure makes it costlier to run many moderate-volume sources than to consolidate into a handful of large ones, penalizing exactly the architectural choice most teams make to improve reliability and source coverage.
The second trigger is sync frequency. Moving from hourly batches to near-real-time sync increases how often rows get touched, and that frequency drives MAR counts up directly for any table with meaningful churn. The third trigger is schema change. A schema change on the source side can force a full table resync, and that resync counts against the row allowance for the billing period, at a volume that has nothing to do with how much the underlying business actually grew.
This is the "success tax" dynamic built into any model that bills per row or per active row: product growth itself becomes the thing that gets taxed. The more customers a SaaS product signs, the more its integration bill balloons, and some teams end up throttling their own data flows just to keep costs in check. The numbers involved aren't trivial. Teams that cross into high MAR territory routinely report bills landing an order of magnitude above what their pilot led them to expect, and a platform that started on a modest monthly plan at a low row tier can climb past $40,000 a month once volumes reach the hundreds of millions of rows. This structural risk is visible specifically when connector count, sync frequency, and schema volatility move together, and the next section covers where row-based pricing still fits.
What destination-based pricing meters instead
Destination-based pricing removes volume from the billing equation. The meter becomes the number of destinations, connections, or bytes delivered, rather than the number of rows that moved through them, so growth in data volume no longer grows the bill. Integrate.io is one example of a vendor built around this logic: a flat monthly rate starting at $1,999 a month, regardless of data volume, though the exact rate varies by plan tier. The assumption this model makes is different from the one row-based pricing makes. It assumes the number of downstream systems a team connects to stays more stable and predictable over time than the volume of data flowing through those systems.
Prequel shows what destination-based pricing looks like when you apply it to embedded data connectivity. It prices by destination with no data-transfer fees. A SaaS product can sync very large volumes to a single destination, frequently, without the bill moving. The cost of shipping data to a customer's warehouse stays the same whether that data amounts to a small number of rows or many times more. That model encodes a particular bet: that a product team's ability to serve customers data connectivity shouldn't be taxed by how much data moves through it.
What destination-based pricing does well is give finance one number to forecast and give engineering room to run more frequent syncs, add connectors, and bring on new data sources without turning each change into a pricing negotiation. But the model's core assumption can fail too. If a team's dominant growth pattern is adding many new destinations, rather than pushing more volume through a smaller set of existing ones, destination-based pricing can compound on its own. Most SaaS products show that growth pattern less often, but it still happens, so you should check for it before assuming the flat-fee model is automatically safer.
Ampersand takes a related but distinct approach: it bills by gigabytes of data delivered rather than by row count or destination count. So your cost ties to actual data mass instead of record count, but you still get the same insulation from row-count volatility that destination-based models provide. Its Catalyst tier is built for a fixed number of production customers with a set monthly data allowance.
Where each model holds up
Neither model is the right answer in every case. Each fits a specific workload profile, and the right choice comes down to which variable in a given pipeline is actually predictable.
Row-based pricing holds up well in a few clear situations. It works when source tables are insert-dominant and low-churn, so data arrives and stays put rather than getting rewritten, and MAR counts reflect genuinely new information instead of update noise. It works when a team runs a small number of large, stable sources rather than many moderate-volume connectors, because that structure captures volume discounts instead of fragmenting across per-connector billing. It also fits pilots and proofs of concept well, where low entry costs matter and volume stays controlled by design, limiting long-term exposure.
Row-based pricing breaks first when a specific, recognizable set of conditions line up. It breaks when the source is a high-churn operational table, orders, sessions, inventory, where rows get updated frequently and each update compounds the MAR or event count. It breaks when a team adds connectors as a matter of course, since connector-level billing prices each new source independently with no pooling across the account. It breaks when a product needs fresher data over time, since the push toward real-time sync raises the rate at which rows get touched and counted. And it breaks when schema changes happen often, since a single resync triggered by a schema change can exhaust a month's row allowance in days.
If your data volumes are high or growing and you need to sync frequently, destination-based pricing holds up, because the flat rate doesn't penalize freshness or scale. It holds up when a pipeline is already production-critical and finance needs a number it can actually plan around. It fits particularly well when a SaaS product embeds data export as a customer-facing feature, since per-customer volume can't be controlled from the vendor side, and a model that ignores volume protects margins regardless of how customers behave. It also suits regulated environments, where compliance rules demand frequent, complete data movement, so you can't throttle syncs just to manage row costs.
Destination-based pricing breaks first in narrower circumstances. It breaks when a team's main growth pattern is connecting to many new destinations rather than moving more data through existing ones, since destination count becomes the variable to watch in that case. It also breaks, in a more mundane sense, when the workload is genuinely small and stable: paying a flat fee for a pilot-scale workload that would cost almost nothing under MAR is simply wasteful.
Compute-based pricing is a third category on this map, exemplified by AWS Glue's DPU-hour model. It shares destination-based pricing's insulation from row counts, but it shifts the cost basis to job runtime and cluster size instead. The software itself may be free, but infrastructure and engineering time typically dominate the real bill, which makes this model workable mainly for teams with real infrastructure engineering capacity on staff.
Egress fees and warehouse storage costs as a third billing layer
If you pick between destination-based and row-based integration pricing, you only address part of the total cost picture, because cloud data warehouse and object storage egress fees run on their own metering logic and compound independently of whatever the integration tool charges. Storage pricing at the warehouse tier tends to be fairly uniform across platforms, but egress costs compound sharply for any organization syncing data out to multiple destinations. Regulated organizations running compliance tooling, SIEM systems, and audit platforms that pull from warehouse data see this compounding directly, because every one of those downstream systems generates its own egress charges.
Object storage egress is its own problem. A small application on S3 storing 50 GB while serving 500 GB of downloads a month pays roughly 31 times more in egress than in storage, and the download line ends up dominating the bill entirely. Cloudflare R2's zero-egress pricing model emerged to answer this for SaaS builders whose costs are dominated by egress rather than storage, so you need to calculate total destination-based cost before you choose a warehouse, not after.
The practical consequence is straightforward: a team on flat-fee integration pricing can still end up with a variable, growing bill at the warehouse layer if its sync pattern generates heavy egress. Predictability at the integration layer doesn't carry over to storage automatically. This is one reason the hidden cost multipliers inside row-based billing, per-connector MAR calculations, delete counting, History Mode, tend to surprise buyers only after they've already committed to a contract. Platforms built around destination-based pricing sidestep this entire category of surprise because they decouple the bill from operational events like schema changes, deletions, or sync frequency, so cost behavior stays predictable no matter how the underlying data changes.
Compliance architecture intersects with this cost layer directly. Platforms that avoid storing customer data in the first place reduce their subprocessor disclosure obligations, and a flat-fee, pass-through architecture tends to simplify SOC 2, GDPR, and HIPAA compliance posture alongside whatever cost benefits it offers. Prequel's architecture, fully encrypted and SOC 2 Type II certified, with an optional private-cloud deployment, addresses this layer specifically: customers with strict data residency requirements can avoid the egress charges that come from cross-region data movement by keeping data inside their own infrastructure. For a team weighing pricing models, that compliance posture is as much a part of total cost as the row count or destination count ever was.


