Fivetran vs Airbyte vs Hevo

Fivetran, Airbyte and Hevo all document Salesforce extraction into an analytical destination, but the decision turns on how their Salesforce connectors treat historical scope, formula fields, change detection and exceptions. Fivetran describes broad object replication and excludes formula fields from its ordinary destination tables. Airbyte offers several sync modes and a configurable lookback. Hevo's Edge connector asks you to choose historical duration before creation and describes separate handling for compound and formula fields. Test those differences with your own Salesforce objects before choosing a pipeline.✓ 1✓ 2✓ 3

• Start with Fivetran if your evaluation centers on selectable history mode, granular blocking and a clearly documented re-import path for objects without useful timestamps.✓ 1
• Start with Airbyte if you need to compare append, deduped and full refresh modes on the same Salesforce object, then inspect API-limit behavior in a test run.✓ 2
• Start with Hevo Edge if the initial historical window, column selection and destination treatment of compound or formula fields are the hard decisions.✓ 3

3 Salesforce extraction paths at a glance

The historical row reflects Fivetran's production and Sandbox distinction and history mode, Airbyte's Start Date default, and Hevo Edge's duration setting.✓ 1✓ 2✓ 3

The incremental row follows Fivetran's timestamp requirement and re-import path, Airbyte's mode list and lookback, and Hevo Edge's timestamp and full-fetch descriptions.✓ 1✓ 2✓ 3

The formula and failure rows reflect each vendor's own caveats. They are a reading guide, not a claim that the three settings produce equivalent warehouse tables. The differences matter when they match your objects, permissions and destination design.✓ 1✓ 2✓ 3

Fivetran: inspect the timestamp and formula boundary

Fivetran says its Salesforce connector pulls underlying CRM data into a SQL-enabled environment. Its feature list includes history mode for all tables, data blocking at column, table and schema levels, and re-sync at table or connection level. These are useful controls to verify in a trial: choose a small set of objects, block one sensitive field, and confirm the destination schema and history behavior. A feature listing does not tell you which fields your own Salesforce user can see.✓ 1

Most Salesforce tables use incremental syncs in Fivetran's account of the connector. It relies on reliable timestamps such as SystemModStamp, LastModifiedDate, CreatedDate or LoginTime. When an object lacks those fields, Fivetran says it re-imports the table. A buyer should therefore separate the number of objects from the way each object moves. A large re-import table and a small incremental table can place different demands on a reporting workflow; measure both during a pilot rather than assuming one cadence describes the entire connector.✓ 1

Where Fivetran falls short

The clearest documented limit is the treatment of Salesforce formula fields. Fivetran says they cannot be synced incrementally with reliable integrity, so it excludes them from ordinary synced destination tables and recommends a transformation approach when they are needed. It also warns that a change in derived fields does not update the incremental column, so an ordinary incremental sync will not capture the resulting value change. If a revenue report depends on a calculated Salesforce field, test the transformation and the resulting historical interpretation before relying on that report.✓ 1

Migration difficulty

Fivetran documents separate Salesforce production and Sandbox connectors. That permits a controlled evaluation of schema choices, but migration work still lies in mapping the destination tables used by existing dashboards and queries. Inventory every formula-driven measure and every object that needs a re-import. Then compare sample rows before and after a source update, including one change that touches only a calculated field. Those checks follow directly from the connector's stated timestamp and formula limits.✓ 1

Airbyte: choose a sync mode and watch the cursor

Airbyte reads standard and custom Salesforce objects as separate streams. Its documentation lists incremental append plus deduped, incremental append, full refresh overwrite and full refresh append. The connector's object list depends on both the queryable property of an object and the authenticated user's permissions. In a pilot, choose an ordinary object, one custom object and a permission-restricted object. The last case helps distinguish a missing connector capability from a Salesforce access setting.✓ 2

Historical scope deserves attention before the first load. Airbyte says that leaving its Start Date blank replicates the last two years by default. It also provides an incremental lookback window, defaulting to ten minutes, to reread recent data when Salesforce API results lag. Record the date boundary explicitly in the test plan, then change one record shortly before and after a sync. The aim is to confirm whether the resulting destination matches the reporting period your team expects.✓ 2

Where Airbyte falls short

Airbyte's connector can stop after reaching Salesforce's daily API limit while marking the current sync successful; its next incremental sync resumes from the prior point. A green job status alone can therefore be an incomplete answer to “is the warehouse current?” Check the latest replicated record or cursor alongside the job state. Airbyte also syncs formula field outputs but says that a formula change without another record update requires a stream reset and historical backfill to refresh old values. Put one such change into acceptance testing if calculated fields matter.✓ 2

Migration difficulty

Airbyte uses the Salesforce Bulk API by default and documents REST fallback for objects the Bulk API cannot handle. That makes an object-by-object inventory valuable before replacing a pipeline. For each selected stream, record the sync mode, starting date, last observed cursor and destination row behavior after an update. If you choose full refresh for one object and incremental append with deduplication for another, migration cannot be judged by a single total row count. A sample of changed and deleted records provides a more useful check.✓ 2

Hevo Edge: settle historical scope before creation

Hevo's Salesforce Bulk API V2 Edge page sets a three-month default historical sync duration and says that choice cannot be changed after pipeline creation. Its first run loads historical data for selected objects. Later runs use a specified frequency and timestamp fields to fetch modified records. If none of those fields is present, Hevo fetches every record for that object on each incremental run. Define the earliest reporting date and list any unusual Salesforce objects before creating the Edge pipeline.✓ 3

The page says Hevo ingests all columns by default but lets you select columns on the Object Configuration page. It uses Salesforce's Bulk API for replication and REST when Bulk is unsupported for an object. Compound fields are split into destination columns rather than replicated as one column. For example, a warehouse model that expects a single address structure may need to rebuild it from subfields. Check field naming in the destination before switching dependent models to the new tables.✓ 3

Where Hevo Edge falls short

Hevo describes formula definitions in destination system tables and views that recalculate logic after incremental loads. That is a different design from simply loading a static formula output. Review the generated views with the team that owns downstream SQL, especially if they need to explain how a past value was calculated. Hevo also replicates deleted records from the Salesforce Recycle Bin and warns that a pipeline disabled for more than 15 days cannot replicate deleted data from that period. The Recycle Bin may purge older records sooner when the organization reaches its limit, and Hevo advises running the pipeline within two hours of deletion for correct capture. Test the delete state and document the recovery procedure before relying on the pipeline for audit-style reports.✓ 3

Migration difficulty

The initial duration setting makes historical backfill planning a front-loaded task in Hevo Edge. Write down the oldest required opportunity, account and lead dates before setup. Then compare selected columns, compound-field subcolumns and a formula-derived measure between Salesforce and the destination. Run a delete test while the pipeline is active. Each check corresponds to a behavior Hevo documents; it is more specific than declaring a connector easy or hard to migrate in every environment.✓ 3

How to choose from a real Salesforce sample

Use one small but deliberately varied set: a standard object, a custom object, a calculated field, a compound address, an update made near the sync boundary and a deleted record. Record the source value, destination value, last successful run and actual freshness of the row. Fivetran's timestamp and formula rules, Airbyte's cursor and API-limit behavior, and Hevo Edge's historical and delete rules suggest different failure probes. This is an evaluation plan drawn from the cited documentation, not a measured ranking of the vendors.✓ 1✓ 2✓ 3

The comparison is specifically about Salesforce extraction. A team's warehouse destination, access policy and reporting model can change the purchase decision. Ask each vendor to show the exact objects and fields you will use, then retain the sample results and configuration choices so the chosen pipeline can be reviewed later. The connector pages establish what is documented, while your own test establishes whether it fits your data.✓ 1✓ 2✓ 3

FAQ

How we researched this

We read the three vendors' Salesforce connector documentation on October 12, 2026. We compared only the behaviors each vendor describes for its own product and did not run a live pipeline or infer a universal speed ranking. The saved snapshots preserve the precise passages used in the fact ledger.✓ 1✓ 2✓ 3