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

# Run reconciliation

> Execute a saved reconciliation setup and generate output.

Running a reconciliation is the easy part — the work happened back in setup. This step takes your saved sources, schemas, matching keys, and rules, compares everything, and hands back what doesn't line up. That output is where the real decisions start.

## Before you start

Confirm:

* The source files or records are available.
* Required connection records are active when the run reads from SFTP, NetSuite, or another configured system.
* The selected schemas match the input shapes.
* The saved run points at the right source systems, schemas, primary ID fields, and rules.
* You are in the correct active company or active tenant context.

## What Darpan needs for a run

| Input              | Why it matters                                                                                    |
| ------------------ | ------------------------------------------------------------------------------------------------- |
| Source data        | The files, text payloads, or connection-backed records being compared.                            |
| Schemas            | The known shape of source data, especially for JSON inputs and field selection.                   |
| Primary ID fields  | The values Darpan uses to match records across systems.                                           |
| RuleSet and rules  | The comparison logic Darpan applies after source records are matched.                             |
| Saved run setup    | The reusable run configuration that ties sources, schemas, primary ID fields, and rules together. |
| Connection records | Saved access details for systems that provide or receive run data.                                |

## Steps

1. From **Ask Darpan**, open **Run Reconciliation**.
2. Select the saved run or source strategy required by the workflow.
3. Confirm schemas, primary ID fields, and RuleSet selection.
4. Provide source files or connection-backed inputs.
5. Start the run. Darpan opens the run so you can watch it work.
6. Wait for generated output.

If the run setup does not exist yet, start with [Set up a reconciliation run](/guides/setup-reconciliation-run).

## Watching a run

Starting a run takes you straight to it, so a long extract is something you can watch rather than guess at. The run shows its current stage, the files it has pulled so far, and results as they land.

A run moves through a fixed sequence of stages, and the stage it is on tells you where the time is going:

| Stage           | What is happening                                                                                                                 |
| --------------- | --------------------------------------------------------------------------------------------------------------------------------- |
| `RESOLVE`       | Working out what to compare: the saved run, its sources, and the date window.                                                     |
| `EXTRACT_FILE1` | Pulling records from the first source.                                                                                            |
| `EXTRACT_FILE2` | Pulling records from the second source.                                                                                           |
| `COMPARE`       | Matching records and applying rules.                                                                                              |
| `WRITE_OUTPUT`  | Writing the result you will review.                                                                                               |
| `VERIFY`        | Rechecking differences against sources that support point lookups, so an export timing skew is not reported as a real difference. |
| `NOTIFY`        | Sending run-completion alerts.                                                                                                    |

To be told when it finishes instead of watching, choose **Notify me** — see [Run-completion alerts](/guides/run-completion-alerts).

## Stopping a run

A run started against the wrong date window does not have to finish. On a running run, choose **Cancel run**, then confirm with **Stop run**.

Stopping is cooperative rather than instant. Darpan records the request and the run ends itself at its next checkpoint — a stage boundary, or a progress tick during a long paged extract — so it may keep reporting as running for a few seconds. The button reads **Stopping run…** in the meantime.

<Warning>
  Any work completed so far is discarded. A stopped run finishes as cancelled and produces no reviewable output — it is not a pause, and it cannot be resumed.
</Warning>

## Expected result

A successful run creates saved output that can be opened from result, history, or generated-output views. The output separates records missing from the first source, records missing from the second source, and matched records that differ by rule (the run result shows a bucket for each source file plus a **Rule differences** bucket).

## Reusing a run

Use a saved run when the same comparison will happen again. Saved runs keep setup stable so the next execution can focus on fresh source data and reviewable output.

Use **Create Reconciliation Flow** when the setup itself is still being defined. Use **Open Runs Settings** when the setup exists but needs a name, schema, source, or rule update before the next run.

Use [Set up an automation](/guides/setup-automation) when the saved run should execute on demand or on a recurring schedule.

<Warning>
  If a run completes with zero matches, do not assume the data is clean. Check join keys, source identifiers, schema selection, and file encoding.
</Warning>