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

# Build a test from an idea

> Build this turns an approved Signals idea into a real ABTestly experiment draft, prefilled from the hypothesis. When that experiment concludes, its outcome feeds back into future scans as ground truth.

<Note>
  **Beta.** Build this is part of the Signals Beta. See the [Signals
  overview](/signals/overview).
</Note>

An idea is only useful if it becomes a test. **Build this** is the
button that makes the jump, from a hypothesis on the [idea
board](/signals/read-a-scan) to a real experiment you can run.

## What Build this does

Click **Build this** on an approved idea and ABTestly creates a **real
experiment draft**, prefilled from the hypothesis, and opens it in the
[experiment editor](/build/create). It is an ordinary ABTestly
experiment from that point on. Nothing about it is special or locked.
The draft gives you a running start:

* The **page** the idea came from, ready under Where it runs.
* A **name and hypothesis** carried over from the idea, so the test
  remembers why it exists.
* A sensible **starting point** for the change the idea proposed.

From there you are in the normal flow. Write the [variation
code](/build/variations), confirm [targeting](/targeting/overview),
pick a [primary goal](/goals/overview), and start it when you are
ready. Nothing goes live until you do. Everything on the [Create an
experiment](/build/create) page applies, because this is that same
editor.

<Note>
  Starting an experiment needs the **Publish experiments** permission on
  the site, the same as any test. If Start is grayed out, ask an admin,
  see [Team and access](/team/members-and-roles).
</Note>

## The learning loop

This is what makes ABTestly's audit different from a standalone tool.

When the experiment you built runs to a conclusion, its outcome, **won,
lost, or inconclusive**, feeds back into future scans of the same site
as **ground truth**. The next scan already knows what worked on your
pages and what did not, so it stops repeating ideas you have disproven
and leans toward the kinds of changes that have won for you before.

```mermaid theme={null}
%%{init: {'theme':'base', 'themeVariables': {
  'fontSize': '18px',
  'fontFamily': 'ui-sans-serif, system-ui, sans-serif',
  'primaryTextColor': '#0f172a',
  'lineColor': '#94a3b8',
  'edgeLabelBackground': '#f8fafc'
}, 'flowchart': {'nodeSpacing': 50, 'rankSpacing': 70, 'padding': 16, 'curve': 'basis'}}%%
flowchart LR
  I["<b>Approved idea</b>"]
  B["<b>Build this</b><br/>experiment draft"]
  R["<b>Result</b><br/>won, lost, inconclusive"]
  S["<b>Next scan</b><br/>knows the outcome"]

  I --> B --> R
  R -- "ground truth" --> S
  S -- "sharper ideas" --> I

  classDef a fill:#fef3c7,stroke:#d97706,stroke-width:2.5px,color:#0f172a;
  classDef b fill:#dcfce7,stroke:#16a34a,stroke-width:2.5px,color:#0f172a;
  classDef r fill:#ede9fe,stroke:#7c3aed,stroke-width:2.5px,color:#0f172a;
  classDef s fill:#dbeafe,stroke:#2563eb,stroke-width:2.5px,color:#0f172a;

  class I a
  class B b
  class R r
  class S s
```

A standalone audit tool cannot close this loop, because it never sees
the result of its own advice. ABTestly owns both ends, the idea and the
experiment, so the loop closes on its own. The more you run, the more
the audit reflects your site rather than generic best practice.

## From idea to shipped, end to end

<Steps>
  <Step title="Approve an idea">
    On the [idea board](/signals/read-a-scan), move a promising idea to
    Approve.
  </Step>

  <Step title="Build this">
    Create the prefilled experiment draft and open the editor.
  </Step>

  <Step title="Finish and start the test">
    Add the variation, confirm targeting and a goal, and start it. See
    [Create an experiment](/build/create).
  </Step>

  <Step title="Read the result">
    Let it reach a verdict on the [results
    page](/results/reading-a-result): won, lost, or inconclusive.
  </Step>

  <Step title="Scan again">
    The next scan of this site carries that outcome forward as ground
    truth, and its ideas get sharper.
  </Step>
</Steps>

## Next

<CardGroup cols={2}>
  <Card title="Create an experiment" icon="flask" href="/build/create">
    The full editor the draft opens into.
  </Card>

  <Card title="Reading a result" icon="chart-column" href="/results/reading-a-result">
    How the outcome that feeds the loop is decided.
  </Card>
</CardGroup>