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

# Runs, Results, and evidence

> Start Runs, follow their status, read Results, and open the exact values and mismatch evidence that stay in your Fabric workspace.

A Run executes every Test in one Table Pair as a single Spark job in your
dedicated General Validation workspace. General Validation records the Run's
status and a summary of each Test's outcome. Exact values and mismatched rows
stay in your Results Lakehouse and are read by your browser only when you ask
for them.

## Start a Run

Owners and Contributors can start Runs; Readers cannot. Starting a Run requires
an active trial or subscription and a configured Environment. A Run that is
already in progress when access expires still finishes and records its Result.
The job runs under your own Microsoft identity, so you need read access to the
source and target Tables and write access to the Results Lakehouse in the
dedicated workspace.

To run one Table Pair:

1. Open **Table Pairs**.
2. Open the Table Pair.
3. Click **Run validation**.

The row's ⋯ menu on Table Pairs offers the same Run validation
action. It is unavailable until the pair has at least one Test. General
Validation freezes the pair's Tables, join keys, filters, and Tests for that
Run; your browser then submits the job to Fabric with your own Fabric
permissions and opens the Run's page. Later edits to the
pair or its Tests do not change a Run that already exists. A Run is one
attempt; to try again, start a new Run.

To run every Table Pair at once, click **Run all** in the workspace header. It
submits one Run for each Table Pair that has Tests, one after another, skips
pairs without Tests, opens **Results**, and reports how many Runs reached
Fabric together with any Table Pair whose Run was rejected. Run all is
hidden for Readers and unavailable until the Environment is set up.

## Run status

Runs, Table Pairs, and Tests show the same status badges. The text label is
always visible alongside the color.

| Status | Color | Meaning |
| - | - | - |
| Prepared | Blue | General Validation has frozen the Run's configuration. No Fabric job has been recorded for it yet. |
| Submitted | Blue | Fabric accepted the job and its job ID is recorded. Fabric queues the job and decides when it starts on your capacity. |
| Running | Blue | Fabric reports the job as running in your dedicated workspace. |
| Passed | Green | Every Test passed. |
| Failed | Red | At least one Test failed and no Test errored. |
| Error | Amber | At least one Test could not be evaluated, or the job did not finish and no Result was saved. |

Each Test is evaluated independently, so one Test's error does not stop the
others. If the job cannot read a source or target Table, every Test in the Run
reports an error. Fabric, not General Validation, decides when the job runs.
While a Run is in progress its page refreshes on its own, no more than once a
minute, and you can leave the page and return later.

A Run whose Tests fail also shows as failed in Fabric monitoring. Open the Run
in General Validation to see whether a Test failed or the execution itself had
a problem. The Run page also shows a notice when Fabric rejected the
submission or could not finish the job. In those cases open the job in Fabric
for its details, because General Validation records only a fixed reason code
and never Fabric's message.

## Results

Results lists Runs newest first, including Runs of archived Table Pairs,
with each Run, its Table Pair, when it started, its status, its Test count, and
how many Tests passed, failed, or errored. Refresh Runs reloads the list.
Results contains no action to start a Run.

Each Run has its own link, shown as the first eight characters of its Run ID,
that you can bookmark or reload. Opening a Run shows its overall status and,
for each Test, its verdict, the number of comparisons, rows, or values
compared, the mismatch count, and a reason. Most reasons state whether the
compared values, rows, or sets agreed. The following reasons need more
explanation.

| Reason | Meaning |
| - | - |
| Within allowed bad records | The Test passed because its mismatches were at or below its Max bad records setting. The mismatch count stays exact. |
| No matching join keys | A VALUE Test found no rows whose join keys exist on both sides, so it failed. |
| Could not read the Tables | The job could not read a source or target Table. Every Test in the Run shows this reason. |
| The Test could not finish | A mapped column is missing, the function does not support the column type, or a cast failed. |
| Could not save Result details | The job could not write the Test's evidence to the Results Lakehouse. |

Refresh Result asks Fabric for the job's status and imports the latest
summary: verdicts, counts, and reason codes. When a summary is imported,
General Validation confirms that it belongs to the Run it was prepared for and
that it contains only verdicts, counts, and reason codes. Results reflect what
the job in your workspace wrote; they are validated for consistency, not
independently attested. A Reader can refresh and read a Run, but only an Owner
or Contributor records the imported summary in General Validation.

If a submission is interrupted before the job ID is recorded, the Run shows
that Fabric may have started it. Click **Refresh Result** first. If the job
exists in Fabric, an Owner or Contributor expands **Recover a submitted Run**,
pastes its ID into Existing Fabric job ID, and clicks **Attach existing
job**. This records the existing job; it does not submit another Run. General
Validation never retries a submission on its own, and a new attempt always
requires a new Run.

## View details

Click **View details** on a Test to see its exact aggregate values, shown as
Source value and Target value, or a sample of up to 100 mismatched
rows. Your browser reads these directly from OneLake with your own Microsoft
identity. They are shown only while the details view is open: **Close
details** or leaving the Run clears them from the page. They are never sent to
General Validation.

To open details you need access to the dedicated General Validation workspace
in Fabric. Membership in your General Validation Organization does not grant
that access; a Fabric workspace administrator shares the workspace with
teammates who need it. If the page asks for OneLake consent, click **Allow
OneLake access and then Retry details**.

The row sample is evidence for investigation, not an export. The mismatch
count in the Results table is the exact total even when the sample is smaller
or evidence capture was turned off for the Test. If an artifact cannot be
previewed in the browser, the panel links to **Open results Lakehouse in
Fabric** so you can inspect the Run's folder there.

## Where artifacts live

Every Run writes its outputs to the Results Lakehouse, named
GeneralValidationResults, in your dedicated workspace under
`Files/GeneralValidation/results`, in a folder named with the Run ID. The
folder holds a summary, a details file with the exact values, and Parquet
mismatch evidence that you can query in Fabric with a notebook or the SQL
endpoint. Each Run also keeps the exact instructions it executed; you can
review them in the Run's folder or in Fabric's Spark monitoring. **View in
Fabric** on a Run opens the workspace.

You own these files and control their retention. General Validation never
deletes them on its own; there is no automatic cleanup, and product updates and
Reinstall on Environment leave them in place. Remove on
Environment deletes the dedicated workspace and everything inside it,
including the Results Lakehouse, Results, and row evidence, and affects any
job still running there. After removal, Run metadata remains in General
Validation, but exact values and evidence can no longer be previewed.

## What General Validation stores

For each Run, General Validation stores its identifiers, its frozen
configuration (Table references, join keys, filters, Tests, and the
instructions prepared for the Run), its status,
timestamps, each Test's verdict, reason code, compared count, and mismatch
count, and a pointer to the Run's folder in your Results Lakehouse. It never
stores rows, exact values, Fabric or OneLake tokens, or Fabric's error
messages.

Editing a Table Pair or a Test leaves past Runs and their Results intact. The
Latest Validation badge on the Table Pair and the Latest Result badge
on each Test show Not run until the new definition runs. Deleting a Table
Pair archives it; its Runs and Results remain and still appear in Results.

For error messages and what to send to support, see
[Troubleshooting and logs](https://docs.generalvalidation.com/administration/troubleshooting.md). For the full data
boundary, see [Security and the data boundary](https://docs.generalvalidation.com/administration/security.md).


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