---
title: "How to set up a Cancel Flow A/B Test"
slug: "how-to-set-up-a-cancel-flow-a-b-test"
updated: 2026-08-19T20:25:17Z
published: 2026-08-19T20:25:17Z
canonical: "help.skio.com/how-to-set-up-a-cancel-flow-a-b-test"
---

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

# How to set up a Cancel Flow A/B Test

## Before you start

- The flow must be **Active**. There's no minimum volume requirement to start a test.
- You can only run one A/B test per flow at a time.
- Entry conditions for tests running at the same time must not overlap exactly. Skio warns you if two tests target identical entry conditions.
- Once a test is published, the flow graph, traffic split, entry conditions, name, and description are locked for the life of the test — including while it's paused. Only the stopping condition and the notify toggle stay editable after publishing. To make any other change, you'll need to cancel the test.

> [!WARNING]
> Results reflect your whole flow's save rate, not just the one change you made. For the clearest read, keep your Variant as close to the Original as possible.

## How to start an A/B test in Cancel Flows

> [!NOTE]
> Learn more about key terms and test statuses in our [Getting started with Cancel Flow A/B Test guide here.](/help/docs/getting-started-with-cancel-flows)

### Step 1: Go to the A/B Testing tab

1. In your **Skio Dashboard**, go to **Retain** > **Cancel Flow**.
2. Click the **A/B Testing** tab, next to **Flows**, to see all your existing and past tests.

![Cancel Flow page showing the Flows and A/B Testing tabs, with an arrow pointing to the A/B Testing tab](https://cdn.us.document360.io/a953bba6-c70d-43ce-bc77-d7f32a8fc514/Images/Documentation/base64-converted-image-1786512930574.png)

The A/B Testing tab sits next to your regular Flows tab on the Cancel Flow page.

### Step 2: Start a test from an existing flow

1. Click back to the **Flows** tab.
2. Find the flow you want to test in the list, then click the **⋮** (more options) menu next to it.
3. Select **Start A/B test**.

![Flows list with the options menu open next to a flow, showing Disable, Archive, Duplicate, Promote to default, and Start A/B test](https://cdn.us.document360.io/a953bba6-c70d-43ce-bc77-d7f32a8fc514/Images/Documentation/base64-converted-image-1786512930575(1).png)

Start a new test from the options menu next to any eligible flow.

### Step 3: Configure your test

The **Create new test** panel opens. Fill in the following:

- **Test name:** Give your test a name that identifies which flow and change it covers. Skio suggests a default name based on the flow you're testing.
- **Traffic split:** Use the slider to set the percentage of customers routed to the **Original (A)** versus the **Variant (B)**, from 1% to 99%. It defaults to 50/50, and Skio enforces whatever split you set when assigning customers to a variant.
- **Variant source:** Select **Duplicate current flow** to clone your original flow so you can edit the copy for your variant.
- **Goal definition:** Choose the **Winning metric** Skio should use to determine a winner, such as `Churn reduction`. Skio picks the variant with the highest save rate for this metric.
- **Completion & notifications:** Turn on **Notify me to stop test** if you want an alert when it's time to wrap up. Choose whether the test should end on a **Specific date** or once it reaches a chosen **Data confidence** level, and set your **Statistical Significance Goal** (for example, `95% Confidence`).

### Step 4: Edit your Variant flow

1. After you save your test settings, Skio creates a duplicate of your flow as **Variant B** and opens it in the flow builder.
2. Make the change you want to test. This can be anything from swapping a treatment's offer to reordering cancellation reasons. Skio highlights what you've changed as you go: added nodes get a green outline, modified nodes get an orange outline, and removed nodes appear dimmed out. Anything you haven't touched keeps its normal appearance.
3. Save your changes to the Variant. Your **Original (A)** flow is left untouched and keeps running as-is.

> [!WARNING]
> Editing Variant B extensively before launch is fine, but keep in mind that the more your Variant differs from the Original, the harder it can be to tell which specific change drove any difference in save rate.

### Step 5: Launch your test

Click **Publish test** to move your test from **Draft** to **Live**. You can do this from the test's edit view, or from the **⋮** menu next to the test in the list. You'll need a stopping condition (a specific date or a data confidence goal) saved before Skio will let you publish.

## Managing a running test

Open a test from the **A/B Testing** tab to manage it. From the test detail view, you can:

- **Pause:** Stop routing new traffic to either variant while keeping the data you've already collected. You can resume a paused test later.
- **Edit termination:** Change when the test is scheduled to end, whether by date or confidence level. This is editable at any point after publishing.
- **Cancel test:** Permanently stop and archive the test.

Once you publish a test, the flow graph, traffic split, entry conditions, name, and description are locked for the life of the test — this stays true even while paused. The only things you can still change after publishing are the stopping condition and whether you're notified when the test ends. To make any other change to either flow, cancel the test first.

## Viewing test results

### From the Flows tab

1. From the **Flows** list, click **See variants** next to a flow that's part of a test, or open the **⋮** menu and select **See details**.

![Flows list row showing the See variants dropdown and a See details option in the options menu](https://cdn.us.document360.io/a953bba6-c70d-43ce-bc77-d7f32a8fc514/Images/Documentation/base64-converted-image-1786512930576.png)

Open a test's results from See variants or See details.

1. The test detail view shows your **Goal definition**, **Ongoing metrics** (Sessions, Post-save revenue, and Avg save rate lift for each variant), and a side-by-side visualization of both flows with per-step Saved, Cancelled, and Skipped rates.

![Test detail view showing Variant A and Variant B flow builders side by side, with ongoing metrics for sessions, post-save revenue, and average save rate lift](https://cdn.us.document360.io/a953bba6-c70d-43ce-bc77-d7f32a8fc514/Images/Documentation/base64-converted-image-1786512931015.png)

The test detail view compares Variant A and Variant B side by side, with live metrics and per-step save, cancel, and skip rates.

### From the A/B Testing tab

1. From the **A/B Testing tab**, click on the **⋮** menu in the top right corner of an existing test card and select **See details**.
2. The test detail view shows your **Goal definition**, **Ongoing metrics** (Sessions, Post-save revenue, and Avg save rate lift for each variant), and a side-by-side visualization of both flows with per-step Saved, Cancelled, and Skipped rates.

![](https://cdn.us.document360.io/a953bba6-c70d-43ce-bc77-d7f32a8fc514/Images/Documentation/CleanShot 2026-08-12 at 11.19.09@2x.png)

You can also see a quick summary of **Lift**, **Confidence**, **Sessions**, and **Post-save revenue** for any test directly from its card in the **A/B Testing** tab, without opening the full detail view.

## Ending a test and promoting a winner

A test ends either because you end it manually or because it reaches the stopping condition you set (a specific date or a data confidence goal). Reaching a stopping condition doesn't end the test automatically — you still need to open it and choose how to conclude it.

### Ending a test early

You can end a test at any time from its detail view, even before it's reached its stopping condition.

1. Open the test from the **A/B Testing** tab.
2. Click **End test** in the top-right corner. If you want to keep collecting data instead, click **Extend test**.

1. Confirm by clicking **End test** in the dialog. This can't be undone.

> [!WARNING]
> **Warning:** Ending a test always marks it as `Inconclusive`, even if one variant is currently outperforming the other, and you won't be offered the option to promote a variant. If a winner has already been determined and you want to promote it, use **Promote winner** instead of **End test** — see Promoting a winning variant below.

### Extending a test

If a test hasn't reached its stopping condition and you want more time to collect data, extend it instead of letting it end.

1. Click **Extend test** in the top-right corner of the test detail view.
2. Choose a new date from the calendar.
3. Click **Extend test** to confirm.

### Choosing what to do once a test reaches its stopping condition

Once a test reaches its stopping condition, Skio marks its results ready for review, but doesn't end the test or promote anything on its own. If you turned on **Notify me to stop test**, you'll get an email letting you know results are ready — you'll still need to open the test and choose an action yourself.

Which buttons you see depends on the outcome:

- **A variant reached your significance goal:** you'll see **End without promotion** and **Promote winner**.
- **You're using a retention-optimization goal and it's been met:** you'll see **End test** and **Promote treatment**.
- **Neither variant reached a significant result:** you'll only see **End test**. There's no option to promote a variant in this case.

### Promoting a winning variant

When a winner is available, click **Promote winner** (or **Promote treatment**) to open the Conclusion & Promotion dialog and choose which variant becomes your new default Cancel Flow.

1. Review the **Winner** and **Underperformed** labels and each variant's save rate.
2. Select the variant you want to promote, or check **Pick another flow** to keep a different existing flow as the default instead of either variant.
3. Click **Promote [Variant]** to confirm.

> [!NOTE]
> **Note:** Promoting a variant doesn't archive, disable, or delete the other variant. Skio creates a new version of your original flow using the winning variant's content and makes that version active going forward, so all customers meeting the flow's entry conditions see it from that point on.