At American Student Assistance, the trouble began with a missing transformation feature. The nonprofit added dbt to its ingestion setup, then struggled to follow the work across tools. Data could arrive perfectly well. Making the next task happen was another matter.
Anyone who has watched a dashboard refresh too early knows the peculiar embarrassment of this situation. Every component can be doing its job while the overall result is wrong. Somewhere between “the data arrived” and “the report is ready,” a person becomes the scheduling system.
- Rivery, now Boomi Data Integration, joins ingestion, transformation, and orchestration in a managed cloud service.
- Engineers can use SQL and Python; analysts get a visual interface and reusable workflow kits.
- Usage pricing rewards careful workload estimates. A cheap credit does not guarantee a cheap month.
01A pipeline is a relay race
The product occupies an unglamorous but consequential patch of enterprise software. It takes data from applications and databases, loads it into cloud warehouses or lakes, and helps teams turn the raw material into something usable. Snowflake, BigQuery, and Redshift belong in this picture. The service sells the plumbing around the destination.
Consider a marketing report. Advertising data must arrive before a revenue model runs; that model must finish before a dashboard updates. Rivery’s orchestration supports dependencies, branching, loops, and scheduling. Put those decisions in a workflow and the sequence becomes visible, rather than something preserved in an engineer’s memory.
The return trip matters, too. Reverse ETL sends warehouse results back into operational applications. An enriched customer record can reach a CRM; a workflow can trigger Tableau or deliver a Slack notification. The analysis leaves its glass cabinet and acquires an actual job.
02Consultants who got tired of repeating themselves
Rivery’s founders, Itamar Ben Hemo, Aviv Noy, and Alon Reznik, came from business intelligence consulting. That background offers a useful explanation for its breadth. Consultants see the connector, the transformation, the deployment, and the customer waiting at the end of all four. Their work rarely stops when a table lands.
Ben Hemo later described the decision to make Rivery independent: “keeping this product within the company was holding us back.” The commercial platform launched in 2019. A $16 million Series A followed in March 2021, then a $30 million Series B led by Tiger Global in May 2022.
Itamar Ben Hemo
Aviv Noy
Alon ReznikBy that second round, Rivery reported more than 300 customers worldwide. Its December 2021 pay-as-you-go program had opened another route into the product. Customers could start with usage rather than a lengthy negotiation. Its named users span Bayer, American Cancer Society, Preqin, and Entravision: the connecting thread is scattered data, not a single industry.
03The experiment that changed the shopping list
American Student Assistance’s team considered Rivery and Matillion, preferred managed SaaS, and tested Facebook Ads data. That proof of concept settled the choice. Working with Archetype Consulting, it migrated by business domain, taking roughly a month for each.
The resulting workflow coordinated ingestion, dbt Cloud transformations, and Tableau refreshes. The published case study reports a halved software bill. Much of the migration effort went into transformation logic, rather than setting up ingestion. There is the useful distinction: moving a connection is easier than moving the assumptions attached to it.
The competitive argument is therefore about coordination as much as connection count. Fivetran, Matillion, and Informatica appear in customer evaluations. Rivery offers a combined workflow; buyers can also assemble separate tools. Existing skills, security requirements, and the actual sources determine which arrangement earns its keep.
“we looked at pricing, and we picked Rivery”
Eric Sanders / Entravision
Published customer testimonial
04The ninety-cent question
The published monthly Base rate is $0.90 per Boomi Data Unit credit. Database, file, and webhook movement is metered by transferred volume; most application sources are charged by execution, with additional volume rules. Python consumption follows runtime, instance size, and bandwidth. Warehouse compute adds another bill.
A rate card becomes meaningful only when it meets a workload. More frequent polling can change application costs even when business activity barely changes. Large initial loads behave differently from incremental updates. Ask for an estimate that includes both, then compare measured consumption during the trial.
Illustrative credit subtotal only. Excludes warehouse compute, taxes, and any other charges. Check edition and workload rules.
05Why Boomi wanted the other half of the journey
Boomi acquired Rivery in December 2024 and introduced Boomi Data Integration within its Enterprise Platform in May 2025. The strategic fit is easy to see. Application integration coordinates business systems; warehouse pipelines prepare data for analysis. Companies need both, and the boundary becomes awkward when yesterday’s analysis must influence today’s transaction.
Change data capture helps by moving database changes without repeatedly copying everything. Data Connector Agent addresses another chore: generating customizable YAML configurations for REST sources that lack a ready-made connector. An AI-produced configuration still needs testing against the intended data and permissions.
The September 2026 release extends that direction. Boomi announced full API coverage for data flows and sources, Companion support for building from the IDE, and automatic endpoint discovery from a source name. The company is giving engineering teams more ways to operate the same machinery.
06Give the trial something difficult to do
A useful evaluation begins with a stubborn report. Trace its sources, load the data, run its transformations, and trigger the final refresh. Deliberately interrupt a dependency. Compare the result with the original system. Record the effort required to diagnose and repair the failure.
This approach suits teams that want managed cloud operations and reusable workflows. It becomes less attractive where full self-hosting is essential, or where an established engineering stack already handles coordination cheaply. “Near real-time” also deserves a measured latency target, not an admiring nod.
Even recovery has boundaries. Boomi’s December 2025 documentation says checkpoint recovery for supported database synchronizations operates within a 48-hour execution limit. Test the initial load as well as the ordinary daily run. A demonstration with a small table tells you little about a weekend spent copying the large one.
The reader can copy the experiment without copying the purchase. Test the handoffs, count the work they consume, and price the whole route. A dashboard can look wonderfully composed while someone underneath it is still passing buckets.