> For the complete documentation index, see [llms.txt](https://docs.humblytics.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.humblytics.com/split-testing-overview/how-to-setup-a-split-test.md).

# How to Set Up a Split Test

Create and launch an A/B test in three steps.

{% hint style="info" %}
**Before you start:** the Humblytics tracking script must be installed on the page you want to test. See [Getting Started](/how-to-get-started.md).
{% endhint %}

Go to **Optimize → Experiments** and click **New Experiment**. Setup has three steps: **Setup**, **Variants**, and **Review**. Your progress saves as a draft, so you can leave and click **Continue** on it later.

<mark style="background-color:pink;">ADD GIF HERE</mark>

## Step 1: Setup

**1. Pick what to optimize.** This is the goal that decides the winner. Pick one.

| Group      | Goal                                                                                                                                        | Counts as a win when a visitor...                  |
| ---------- | ------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------- |
| Conversion | [Form Submissions](/split-testing-overview/creating-a-b-page-variants-for-split-testing/conversion-goals/increase-form-submissions.md)      | submits a form                                     |
| Conversion | [Click-through Rate](/split-testing-overview/creating-a-b-page-variants-for-split-testing/conversion-goals/increase-click-through-rate.md)  | clicks a link or button                            |
| Conversion | [Destination Page](/split-testing-overview/creating-a-b-page-variants-for-split-testing/conversion-goals/reach-destination-page.md)         | reaches a page you choose, like `/thank-you`       |
| Conversion | [External Destination](/split-testing-overview/creating-a-b-page-variants-for-split-testing/conversion-goals/reach-external-destination.md) | goes to a URL on another site, like a booking page |
| Conversion | [Revenue](/split-testing-overview/creating-a-b-page-variants-for-split-testing/conversion-goals/increase-revenue.md)                        | buys something                                     |
| Engagement | [Visit Time](/split-testing-overview/creating-a-b-page-variants-for-split-testing/engagement-goals/increase-session-time.md)                | stays engaged on the page                          |
| Engagement | [Bounce Rate](/split-testing-overview/creating-a-b-page-variants-for-split-testing/engagement-goals/reduce-bounce-rate.md)                  | doesn't leave right away                           |

**Destination Page** and **External Destination** ask for the destination URL. You can use `*` as a wildcard, for example `/blog/*`.

**2. Pick a test type.** Humblytics detects your platform and shows what applies to it.

* **No-Code Element Testing:** change the page visually. No deploy or republish needed.
* **Page-Level Testing:** compare two live URLs. Both pages must be published before launch. [How to build a variant page](/split-testing-overview/creating-a-b-page-variants-for-split-testing.md).

**3. Plan the test (optional but recommended).** Write your hypothesis, for example *"If we shorten the form, more people will submit it, because it looks faster."* Then enter your daily visitors, baseline conversion rate, and the smallest improvement you care about (**MDE**). Humblytics shows the visitors needed per variant and an estimated duration. [More about these numbers](/split-testing-overview/using-the-humblytics-a-b-sample-size-calculator.md).

Click **Continue**.

## Step 2: Variants

**No-Code Element Testing**

1. Choose the **Page to test** from the list, or type a URL.
2. Click **Open Editor**. Humblytics loads your live page in a preview. Click an element to edit its text, styles, image, visibility, or order. The **Editor** tab in the side panel holds the controls; AI can suggest copy or styling changes there too.
3. Click **Save Changes** in the editor toolbar (or press **⌘S** / **Ctrl+S**). Each save writes the full list of edits for this variant to your experiment draft on the server. If the server is unreachable, Humblytics keeps a copy in your browser until the next successful save.
4. When you're finished, click **I'm Done — Finish Editing** to return to the Variants step. Make at least one saved change, or Variant B will be identical to the control.

**Reopen the editor**

When you open the editor again on the same draft, Humblytics reloads a fresh copy of your page and replays every saved edit onto the preview. Changed elements are highlighted, the preview scrolls to the first change, and the **Editor** tab opens on that element so you can keep working without starting from a blank page.

**Changes history**

Use **History** in the editor toolbar to see every saved instruction for the current variant. **Revert** rolls back to an earlier point in the list; **Remove** deletes one change. Both actions save immediately and refresh the preview so what you see matches what's stored. Clearing every change saves an empty list to the draft.

**Page-Level Testing**

1. Enter the URL for each version. The first one is your **Control (Original)**.
2. Click **Add Variant** to test more than one alternative.
3. Optional: add notes or screenshots so you remember what changed.

{% hint style="warning" %}
**Copy each URL from your browser's address bar after the page loads.** Matching is exact: `/pricing` and `/pricing/` are different URLs. If your site redirects one to the other and you enter the wrong one, that variant gets no traffic.
{% endhint %}

**Optional settings (both test types)**

| Setting                         | Options                                                                                                                                                                                                                        |
| ------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| **Custom Traffic Distribution** | How visitors are split. Default is 50% / 50%. Shares must add up to 100%.                                                                                                                                                      |
| **Visitor assignment**          | **Per visit** (default): each visit is assigned again. Faster results, best for high-traffic, short tests. **Sticky visitor:** a person always sees the same version. Best for pricing, multi-step journeys, and longer tests. |

Click **Review**.

## Step 3: Review and launch

Humblytics checks your setup before launch:

* A goal is set.
* Every variant has a page URL.
* Variant B has at least one edit (no-code) or its own URL (page-level).
* Traffic shares add up to 100%.
* The sample target is realistic for your traffic. If not, you'll see a low-traffic warning. You can still launch, but consider a higher-traffic page or a bigger change.

Click **Start Experiment**. Visitors start being split right away.

## Troubleshooting

**Saved text or styles don't show when I reopen the editor.** After **Save Changes**, reopening the editor should replay your edits on the preview automatically. If the preview looks like the original page, hard-refresh the Humblytics dashboard (**Cmd+Shift+R** on Mac, **Ctrl+Shift+R** on Windows) so you're on the latest app, then open the editor again. Confirm you clicked **Save Changes** before **I'm Done — Finish Editing**.

**A variant gets no traffic.** Almost always a URL mismatch. Check the trailing slash, `www` vs. no `www`, and that the page loads on your live domain with the tracking script installed.

**The traffic split looks uneven.** A variant probably links back to the control page, for example a `#start` anchor that points to the control URL. Visitors leave the variant and land on the control. Use links that stay on the variant's own page.

**Can I run several tests at once?** Yes, on supported plans. Avoid testing overlapping elements on the same page, or you won't know which change caused the result.

**Will a test slow down my site?** No. Changes load through the same small script, and a short snippet hides the page for a moment (2 seconds at most) so visitors don't see the original flash before their version.

Next: [How to Analyze Split Test Data](/split-testing-overview/how-to-analyze-split-test-data.md).