Key Concepts

This page defines the foundational terms used throughout VersionForge. Bookmark it as a reference -- every other article in the knowledge base builds on these concepts.

Core Terminology

Connection

A connection is a configured link to an external system, either a source you extract data from or a target you load data into. Each connection stores the system type (e.g., NetSuite, Workday HCM, Adaptive Planning), the authentication credentials, and connection-specific settings like base URL or API version.

You create connections once and reuse them across multiple sync profiles. Credentials are encrypted at rest and never exposed in the UI after initial entry.

Sync Profile

A sync profile is the complete mapping configuration that defines how data moves between a source connection and a target connection. It includes:

• Which source entity or dataset to extract
• Which target model, sheet, or dataset to load into
• Field-level mappings between source and target schemas
• Transform rules (value translations, defaults, concatenation)
• The diff key -- the field(s) that uniquely identify a row for comparison across snapshots
• Schedule settings (manual, daily, weekly, or cron)

Think of a sync profile as the blueprint for a repeatable data pipeline.

Run

A run is a single execution of a sync profile's pipeline. Each run progresses through a sequence of stages: Extract, Stage, Diff, Transform, Review, Load, and Verify. Runs are immutable records -- once complete, they cannot be modified, only referenced.

Every run is assigned a unique ID and retains the full snapshot, diff, approval decisions, and cross-check results. This gives you a complete audit trail for any point in time.

Diff

A diff is the row-level comparison between the current snapshot and the previous one for the same sync profile. VersionForge computes the diff by matching rows on the configured diff key, then comparing every mapped field value.

The diff categorizes each row as one of three types:

• Add -- A row exists in the new snapshot but not in the previous one.
• Update -- A row exists in both snapshots but one or more field values changed.
• Delete -- A row existed in the previous snapshot but is absent from the new one.

For updates, the diff highlights exactly which fields changed and shows both the old and new values. This precision is what makes Safety Gate reviews fast and reliable.

Safety Gate

The Safety Gate is the human-in-the-loop review step that sits between the transform stage and the load stage. No data is pushed to your target system until a reviewer explicitly approves it.

In the Safety Gate, you see every proposed change from the diff, organized and color-coded by type. You can approve individual rows, reject rows you want to exclude, or override specific field values. Bulk actions are available for large changesets, and risk scoring highlights rows that are statistically unusual (e.g., a salary that changed by 300%).

The Safety Gate is what separates VersionForge from a blind ETL pipeline. It ensures that a human with business context signs off on every change.

Override

An override is a manual correction applied to a specific field value during Safety Gate review. When a reviewer sees a value that is technically correct in the source but needs adjustment before it reaches the target, they can edit that field directly in the review interface.

Overrides are logged as part of the run's audit trail. The original source value, the override value, and the identity of the reviewer are all recorded. This makes it easy to trace why a value in the target differs from the source.

Cross-Check

A cross-check is an automated reconciliation step that runs after data is loaded into the target system. It compares aggregate totals from the source extract against the corresponding totals in the target to confirm they tie.

Cross-checks catch problems that row-level diffs cannot: missing rows that fell through due to a filter mismatch, rounding differences introduced by transforms, or load failures that silently dropped records. If the variance exceeds the configured tolerance, the cross-check fails and flags the run for investigation.

Tolerance

Tolerance is the acceptable variance threshold for a cross-check. It defines how much the source total and target total can differ before the cross-check is marked as failed.

The default tolerance is $1.00, which accommodates minor floating-point rounding. You can configure tolerance per sync profile -- tighter for high-sensitivity accounts (e.g., $0.01 for cash) or looser where business rules introduce expected differences (e.g., $100.00 for a profile with known allocation rounding).

Staging

Staging is the intermediate storage layer where extracted data lives before diffing and transformation. When VersionForge pulls data from a source, it writes the raw snapshot to staging in chunks, creating an immutable record of exactly what the source returned at that point in time.

Staging serves two purposes. First, it decouples extraction from processing -- if the diff or transform step fails, you do not need to re-extract from the source. Second, it provides a historical archive of source snapshots that you can compare across runs for audit or debugging.