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

# Publish the results

> Publish a configuration version and deliver accepted content to a destination.

Accepted content does not go anywhere until you have an active destination and a real review session.

## Publish the version

When a draft performs well in preview, publish it. Publishing makes that version the live one and names the version it displaces, so rolling back is a single action.

<Warning>
  Publishing should never be the first time a version runs over real data. If you have not previewed it, go back a step. [Preview and review](/speedtrain/getting-started/preview-and-review)
</Warning>

## Set up a destination

Go to **Settings → Destinations** and create one. A destination maps accepted content to an Enterspeed source, from where it flows into the rest of your data supply chain.

<Steps>
  <Step title="Choose the connector shape">
    **Per-output** is the recommendation for anything new — each generated output is delivered as its own entity. **Merged** combines outputs into one.
  </Step>

  <Step title="Point it at an Enterspeed source">
    One destination delivers to one source. Take the ingest details from the destination and use them on the Enterspeed side.
  </Step>

  <Step title="Set it active">
    Only **active** destinations deliver.
  </Step>
</Steps>

## Run a real review session

Preview sessions never deliver. To publish content, start a review session — bound to an index or a stage output, on the published version.

As items are accepted, they are delivered to the active destinations.

<Warning>
  Delivery is **never retroactive**. Content accepted before a destination existed, or while it was inactive, is not sent when you activate it later. Set up the destination first.
</Warning>

## Verify

There is no per-item delivery receipt, so verify on the Enterspeed side: check that source entities are arriving in the expected source, and that the shape matches what your schemas expect.

<Tip>
  Judge the pipeline by its end outcome — does the downstream publication actually succeed? — rather than by the quality scores alone. Good scores on content that fails to land is not a working job.
</Tip>

## Keeping it running

A **workflow** connects a content source to a task configuration so new and changed documents are processed automatically, rather than you starting sessions by hand. [Workflows](/speedtrain/key-concepts/workflows)