# Reconcile an Anomaly in the Recipe

Source: https://userguide.qualytics.io/recipes/data-reconciliation/how-tos/reconcile-in-recipe/

*Requires the Master Data Management (MDM) module.*

Open an anomaly raised by a Data Diff check at the recipe's **Reconcile** step, with the anomaly's mismatched rows already loaded. This is how a scan run outside the recipe, by hand, on a schedule, or from a Flow, hands its mismatches to the recipe. See [Entry Points and Resuming](https://userguide.qualytics.io/recipes/data-reconciliation/deep-dive/entry-points/).

> **Note: Permission Required**
>
> The button appears only on anomalies of a Data Diff check that are not archived, and only when you can dry-run checks on the datastore (the **Drafter** team permission or above). Loading the rows needs **Viewer**, and writing the keep set afterwards needs **Editor**. See [Permissions](https://userguide.qualytics.io/recipes/data-reconciliation/deep-dive/permissions/).

## Steps

**Step 1**: Navigate to the datastore holding the asset you investigate and open the **Anomalies** tab.

**Step 2**: Open the anomaly raised by the Data Diff check and click **Reconcile in Recipe** .

**Step 3**: The recipe opens on the **Reconcile** step with the anomaly's mismatched rows loaded, and the **Anomaly** badge highlighted under **Working Assets**. Pick a side for each mismatch in the **Choose** column, or click **Get AI suggestions** to let AgentQ propose them.

**Step 4**: When every mismatch is resolved, click **Next** and write the keep and remove sets on the **Materialize** step, as in [Run the Data Reconciliation Recipe](https://userguide.qualytics.io/recipes/data-reconciliation/how-tos/run-recipe/). The anomaly is acknowledged when the keep set is written.

*Last updated: October 19, 2026.*
