Review Queue Overview
How the Safety Gate review queue works and how to read, filter, and act on pending changes.
Review Queue Overview
The Safety Gate is VersionForge's human review layer. Every change detected by the change detection engine -- every add, update, and delete -- lands here before it reaches your target system. Nothing syncs until you approve it.
The Core Guarantee
No data reaches your planning target without explicit approval. This is the foundational promise of the Safety Gate. Whether a sync detects 3 changes or 30,000, every change sits in the review queue until a human (or an auto-approve rule) approves it.
This matters because FP&A data has real financial consequences. A wrong compensation number in Adaptive Planning can cascade through headcount models, forecasts, and board decks. The Safety Gate gives you a checkpoint to catch problems before they propagate.
What You See in the Queue
When you open the Safety Gate for a pipeline, you see a table of pending changes. Each row represents a single data change and includes:
- Change Type -- An icon and label indicating ADD (new row), UPDATE (existing row changed), or DELETE (row removed from source)
- Row Key -- The unique identifier for the affected row (e.g., employee ID, GL account + period)
- Changed Fields -- For updates, each modified field is shown with its old value and new value side by side
- Risk Level -- A badge showing LOW, MEDIUM, or HIGH risk based on VersionForge's risk scoring algorithm
- Source -- The connector and extraction timestamp
For ADD changes, all fields are shown as new values with no "old" column. For DELETE changes, all fields are shown as old values being removed. Only UPDATEs display the before/after comparison.
Reading the Queue Effectively
Field-Level Detail
Click any row to expand its full detail view. You see every field in the row, not just the fields that changed. Changed fields are highlighted. This lets you review a change in its full context -- for example, seeing the employee's department and location alongside a compensation update.
Risk Indicators
Each change carries a risk score. High-risk changes are flagged with a red badge and appear at the top of the queue by default. Risk scoring considers:
- Field sensitivity -- Compensation and financial amounts are weighted higher than department names
- Magnitude -- A 50% salary change scores higher than a 2% adjustment
- Pattern deviation -- Changes that fall outside historical norms are flagged
See Risk Scoring & Tiers for a deep dive on how scores are calculated.
Filtering and Sorting
The queue supports several filtering and sorting options to help you work efficiently:
| Filter | What It Does | |---|---| | Change type | Show only ADDs, UPDATEs, or DELETEs | | Risk level | Show only HIGH, MEDIUM, or LOW risk changes | | Field name | Show only changes where a specific field was modified | | Search | Free-text search across row keys and field values |
You can combine filters. For example, show only HIGH-risk UPDATEs where the compensation field changed.
Default Sort Order
The queue sorts by risk level descending by default, so the most critical changes appear first. You can re-sort by change type, row key, or timestamp.
Acting on Changes
For each change in the queue, you have three actions:
- Approve -- Accept the change and allow it to proceed to the Load stage
- Reject -- Block the change from syncing. The row stays in its current state in the target system.
- Override -- Modify the change before approving it. See Overrides for the three override types (Correct, Enrich, Suppress).
Rejected changes are logged but not retried. If the same change appears in the next sync (because the source data still differs from the target), it will re-enter the queue as a new pending change.
Queue Lifecycle
Once every change in a sync batch is approved, rejected, or overridden, the approved changes move to the Load stage automatically. You do not need to click a separate "submit" button -- the pipeline resumes as soon as the queue is fully resolved.