> ## Documentation Index
> Fetch the complete documentation index at: https://docs.pcnaid.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Review and approve a change set

> Inspect field-level diffs, risk scores, approve or reject changes, execute through Shopify, and roll back if needed.

# Review and approve a change set

## What does this do for me?

Review the exact field-level changes before they reach your store. Each item shows the current value, the proposed value, and a risk level. After approval, an executor applies the changes through Shopify with rollback snapshots.

## 1. Find the change set

1. Go to **Change sets** in the sidebar.
2. Select a change set to open its detail view.

The detail page shows:

* The change set name, status, and risk score
* A paginated table of change items with key, operation, field, before and after values, risk level, and status
* Approval history and current approvals
* Import status (for in-progress imports)

## 2. Review the change items

Each change item shows:

* **Key** — product handle or natural key
* **Operation** — what will happen (update, create, delete, add tag, remove tag, set metafield)
* **Field** — which field is changing (title, vendor, price, tags, metafields, etc.)
* **Before** — the current value in Shopify
* **After** — the proposed new value
* **Risk** — low, medium, high, or critical

Risk levels:

* **Critical** — product deletion or archiving
* **High** — status, price, compare-at price, or inventory quantity changes
* **Medium** — metafield changes or product creation
* **Low** — title, vendor, tags, or product type changes

The overall risk score is computed from all items combined (0-100).

## 3. Generate an AI review (optional)

1. On the change-set detail page, select **Generate AI review**.
2. The app samples up to 200 items and sends them to an AI reviewer.
3. Review the AI draft in the **AI review** section. It includes a risk summary, suggested decision, concerns, sampling plan, and rollback notes.

AI review output is a draft. Always review the actual change items yourself before making a decision.

## 4. Approve or reject

Approvers can record a decision:

1. Select **Approve** or **Reject**.
2. Add reviewer notes in the **Reviewer notes** field (optional).
3. Confirm your decision.

Approval rules:

* Change sets with critical items or a risk score of 70+ require two distinct approvers.
* Change sets with more than 10 high-risk items or a risk score of 35+ require one approver.
* Change sets below those thresholds require no recorded approval and are approved automatically.
* A rejection blocks execution until the decision changes.

## 5. Execute the changes

Executors apply approved changes:

1. On the detail page, select **Execute**.
2. Confirm in the dialog that appears.
3. Safe Bulk applies each item through Shopify and captures before-values as rollback snapshots.

During execution, each item updates through Shopify's API. If Shopify rate-limits the request, the remaining items stay pending and the job retries automatically.

## 6. Roll back changes

If you need to reverse an executed change set:

1. Open the executed change set.
2. Select **Roll back**.
3. Confirm the rollback.

Rollback processes only items recorded as successfully executed. For updates, it writes the captured prior value. It removes tags that were added, re-adds tags that were removed, restores or deletes metafields as needed, and restores the prior status of products archived by a delete action. Products created by the change set are archived rather than deleted.

Rollback does not reconstruct the entire catalog state from the time of execution. Review any Shopify edits made afterward: rollback can replace newer edits to the same fields with the captured prior values. The rollback outcome is recorded for audit.

## 7. Track job status

Use the **Jobs** page (under **Settings**) to monitor running, queued, and completed imports, executions, and rollbacks. Each job shows its type, status, outcome metrics, and creation time.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.