# Data Reconciliation Recipe Permissions

Source: https://userguide.qualytics.io/recipes/data-reconciliation/deep-dive/permissions/

The Data Reconciliation recipe has no permissions of its own. Each step performs an ordinary platform action, and that action's permissions apply. The check is created and activated like any check, the scan and the materialize are ordinary operations, and the anomaly is acknowledged or archived like any anomaly. This page gathers those requirements in one place. For the canonical matrices, see [User Roles](https://userguide.qualytics.io/settings/access/users/roles/overview/) and [Team Permissions](https://userguide.qualytics.io/settings/access/teams/team-permissions/overview/).

> **Note: Two Permission Systems**
>
> **User Roles** (Member, Manager, Admin) control what a user can do across the platform. **Team Permissions** (Reporter, Viewer, Drafter, Author, Editor) control what a user can do on the specific datastores they reach through team membership. They are independent systems with different scopes.

## User Roles (Platform-Level)

Every role can use the recipe. No step is reserved for Managers or Admins.

| Action | Member | Manager | Admin |
| --- | --- | --- | --- |
| Open Recipes and browse a recipe's steps | Allowed | Allowed | Allowed |
| Receive AI suggestions and analyses | Allowed | Allowed | Allowed |
| Create and activate the check | Allowed | Allowed | Allowed |
| Run the scan | Allowed | Allowed | Allowed |
| Write the keep and remove sets | Allowed | Allowed | Allowed |
| Acknowledge or discard the anomaly | Allowed | Allowed | Allowed |

Each of these still requires the matching team permission, listed below.

## Team Permissions (Datastore-Level)

Most permissions are checked on the **left datastore**, the one holding the asset you investigate: the check is created on it, the scan and the materialize run on it, and the keep and remove sets are written to its enrichment destination. The **right datastore** only needs to be readable.

| Action | Reporter | Viewer | Drafter | Author | Editor |
| --- | --- | --- | --- | --- | --- |
| Browse the recipe's steps and pick assets on either side | Allowed | Allowed | Allowed | Allowed | Allowed |
| Receive AI suggestions and analyses | Allowed | Allowed | Allowed | Allowed | Allowed |
| Use an asset as the reference (right datastore) | Allowed | Allowed | Allowed | Allowed | Allowed |
| Load the scan's mismatched rows (Scan, Reconcile) | Not allowed | Allowed | Allowed | Allowed | Allowed |
| Create the Draft check (Review) | Not allowed | Not allowed | Allowed | Allowed | Allowed |
| Save and activate the check (Validate) | Not allowed | Not allowed | Not allowed | Allowed | Allowed |
| Run the scan (Scan) | Not allowed | Not allowed | Not allowed | Not allowed | Allowed |
| Write the keep and remove sets (Materialize) | Not allowed | Not allowed | Not allowed | Not allowed | Allowed |
| Acknowledge or discard the anomaly (Materialize) | Not allowed | Not allowed | Not allowed | Allowed | Allowed |
| Create a computed asset from Select Assets (as owner) | Not allowed | Not allowed | Not allowed | Allowed | Allowed |
| Create a computed asset from Select Assets (any owner) | Not allowed | Not allowed | Not allowed | Not allowed | Allowed |

> **Note: Completing the recipe end to end**
>
> The scan and the materialize are operations, and running operations needs the **Editor** team permission. A user with **Editor** on the left datastore and at least **Reporter** on the right datastore can complete the whole recipe alone. A **Drafter** can build the check, and an **Author** can also activate it, but both need an Editor to run the scan and write the keep set.

## How Both Layers Work Together

To take two assets all the way to a keep set, a user must satisfy both layers:

1. **User Role**: at least **Member**
2. **Team Permission**: **Editor** on the left datastore, and **Reporter** or above on the right datastore

The AI suggestions follow the same rule as viewing: any team permission on the datastores involved, plus the Member role. Whether a given AI step is available at all depends on the AgentQ data sharing level, not on the user's permissions. See [AgentQ Access Controls](https://userguide.qualytics.io/agentq/deep-dive/access-controls/).

> **Info: Admin Bypass**
>
> Users with the **Admin** role bypass team permission checks and can complete the recipe on any pair of datastores.

## UI Behavior Without Permission

| Scenario | What the User Sees |
| --- | --- |
| User cannot dry-run checks on the left datastore (below Drafter) | **Reconcile in Recipe** is not shown on the anomaly details. The recipe itself still opens from the Recipes page, but on Review it can neither check for a conflicting check nor create the Draft check, and shows a permission error. |
| User is below Viewer on the left datastore | The Scan step cannot load the mismatched rows, so reconciliation cannot start. |
| User cannot run operations on the left datastore (below Editor) | **Scan in Recipe** is not shown on the check details. Inside the recipe, **Scan** and **Materialize** fail with a permission error, and you can try again once the permission is granted. |
| User cannot read the right datastore | The reference asset cannot be selected, and validation of a loaded check fails. |
| User cannot create computed assets on a datastore | The **Computed Table** (or **Computed File**) and **Computed Join** cards are not shown in that column; only **Existing Table** is offered. |
| AgentQ data sharing level below what a step needs, or no AI answer | The step reports that its AI suggestions could not load and lets you continue manually. |
| No AI provider configured | No AI banners appear anywhere in the recipe, and the **Get AI suggestions** offer on Reconcile is absent. Every step works manually. |

## Important Notes

- **The check is an ordinary check.** Once the recipe has created it, editing, cloning, archiving, or scheduling it from the datastore page follows the usual [quality check permissions](https://userguide.qualytics.io/settings/access/teams/team-permissions/overview/).
- **Entering from a check or an anomaly checks the same permissions** as building the recipe from scratch. The entry-point buttons are hidden when you lack them; see [Entry Points and Resuming](https://userguide.qualytics.io/recipes/data-reconciliation/deep-dive/entry-points/).
- **Masked fields stay masked.** The recipe never asks to reveal masked values. Fields masked on either asset appear masked in the mismatch explorer and are written masked to the keep and remove sets, whatever your team permission.

## See Also

- **How It Works**

  Left and right, the phases, the prerequisites, what the run reads and writes, and AgentQ's role.

  [How It Works](https://userguide.qualytics.io/recipes/data-reconciliation/deep-dive/how-it-works/)
- **Build and Validate**

  Select Assets, Row Identifiers, Compare Fields, Tolerances, Review, and Validate, step by step.

  [Build and Validate](https://userguide.qualytics.io/recipes/data-reconciliation/deep-dive/build-and-validate/)
- **Scan and Reconcile**

  The full scan, resolving each mismatch, writing the keep and remove sets, and the completion summary.

  [Scan and Reconcile](https://userguide.qualytics.io/recipes/data-reconciliation/deep-dive/scan-and-reconcile/)
- **Entry Points and Resuming**

  Scan in Recipe, Reconcile in Recipe, and how progress is kept between visits.

  [Entry Points and Resuming](https://userguide.qualytics.io/recipes/data-reconciliation/deep-dive/entry-points/)
- **Examples**

  Real scenarios showing two assets compared, their mismatches resolved, and the keep set written.

  [Examples](https://userguide.qualytics.io/recipes/data-reconciliation/deep-dive/examples/)
- **Best Practices**

  Guidelines for choosing sides, identifiers, tolerances, and resolutions that hold up.

  [Best Practices](https://userguide.qualytics.io/recipes/data-reconciliation/deep-dive/best-practices/)

*Last updated: September 19, 2026.*
