Which ETL tool should you pick to centralize your data?

Airbyte is the tool that pushed ingestion prices down. It lists 700+ connectors on every plan and comes in two flavors. Airbyte Core is free to run on your own infrastructure; it ships under the Elastic License v2, which is source-available rather than open source in the strict sense, while Airbyte's own connectors stay open source. Airbyte Cloud bills in credits: Standard starts at $20 a month for 5 credits, roughly 1 GB, with syncs at most hourly. Plus starts at $189 for 40 credits, adds 15-minute syncs and SSO, and scales up to $4,999 for 2,000 credits.

What users like: the catalog-to-price ratio, and the connector builder when a source is missing. What they don't: thin docs on less popular connectors, frequent API rate-limit errors, and logs that are "not very comprehensive". One reviewer flags that there is no account-level usage cap. Another says schema changes are not handled well.

Self-hosting is free, not costless. You own every upgrade, and connector updates can change behavior. In practice it's the bang-for-your-buck option against Fivetran, as long as someone on your team is on call.

Fivetran is the benchmark for managed ELT. Pick a source and a destination, and the vendor handles the rest, API changes included. You get 700+ connectors, 15-minute syncs on Standard and 1-minute syncs on Enterprise. The free plan is real: up to 500,000 monthly active rows, meaning rows inserted or updated during the month.

The catch is the unit. Deletes count toward paid MAR too. Add a $5 base charge per standard connection as soon as it moves a row. On the plus side, a row changed ten times in a month counts once, which helps on very busy tables. Fivetran also merged with dbt Labs in 2026.

Reviews praise reliability and complain about the bill: several users say costs get hard to predict as soon as you add tables. Several users point out there are no hard billing controls, and that some connectors don't expose every field or object they need.

Hevo is built for teams that don't write code. Everything is point and click, and support gets a thumbs up in most reviews. The free plan covers 1 million events a month for 5 users, with limited connectors and hourly scheduling. Starter runs $299 a month ($265 billed annually) for 5 to 50 million events, 10 users and 190+ connectors. Professional is $849.

The limits show up with scale. One reviewer says Hevo "couldn't keep up with the high number of databases we wanted to sync frequently". Others mention transformations that time out, thin NetSuite and Jira connectors, and on-demand events that get pricey once you exceed your plan.

Worth knowing: Hevo offers a contract buyout if you're locked in with another vendor. Ask for it before you sign.

dbt doesn't extract or load anything. It transforms what's already in your warehouse: you write SQL models, dbt runs them in the right order, runs data quality checks on them and documents the result. That's why it belongs here. In a modern stack the ingestion tool copies raw data, and dbt turns it into clean tables. They work together.

The Developer plan is free for one person, up to 3,000 models built a month. Starter is $100 per user per month, with 5 seats included and 15,000 models. The dbt Core repository, the version you run yourself, is published under the Apache 2.0 license.

The limit: you need real SQL skills and a modeling mindset. A marketing team on its own won't get much out of it. And with Fivetran and dbt Labs merged in 2026, watch how bundles evolve if you use both.

Azure Data Factory is Microsoft's integration service. You pay only for what runs: $1 per 1,000 activity runs and $0.25 per DIU-hour (the compute unit used to copy data) in the West Europe region. The first 5 low-frequency activities each month are free. For a small nightly load, the bill stays tiny.

Three meters stack up: runs, DIU-hours for copying, and vCore-hours if you transform with data flows. That last one is where budgets slip: an oversized data flow burns far more than a plain copy.

Also note that Microsoft positions Data Factory in Fabric as the next generation. If you're starting fresh, look at Fabric before building dozens of pipelines on the older service.

Estuary specializes in streaming. It captures every change in your database (change data capture, or CDC) and pushes it to the destination within seconds. Useful when a dashboard has to reflect orders from the last minute.

Pricing is simple and public. The Developer tier is free up to 10 GB a month and 2 connectors, no card needed. Past that you pay $0.50 per GB moved, plus $100 a month per connector instance, dropping to $50 after your sixth. A 30-day trial starts when you go past the free level.

The limit: the catalog is shorter than Airbyte's or Fivetran's, and the per-connector fee adds up fast with many small sources. It shines on a few big databases, not on thirty SaaS apps.

AWS Glue is Amazon's serverless integration service. You write a job, usually Spark, AWS provides the compute, and you pay $0.44 per DPU-hour, billed by the second. AWS's own example: a 15-minute job on 6 DPUs costs 6 × 0.25 × $0.44 = $0.66. The Data Catalog is free for the first million objects and the first million accesses.

For data already inside AWS, that's very cheap. But Glue isn't a plug-and-play connector library like Fivetran. Wiring up Salesforce or HubSpot takes more work, and debugging a Spark job is not an analyst's job.

Weld is built by Weld Technologies ApS, a Danish company. Its pitch is a flat, published price rather than metered rows. Basic is $99 a month for 2 connectors and a daily sync, Premium $389 for 6 connectors hourly, Business $959 for 10 connectors every 15 minutes. Those prices are annual, which saves 20% compared with paying monthly.

The limit is clear: there's no free plan after the 14-day trial, and the number of included connectors is small on lower tiers. With 2 connectors on Basic, a CRM and one ad platform already fill the plan.