# Adding a Live Chat Button to the Page: A/B Test Idea

Page: https://alidemirbas.com.tr/en/lab/ab-test-library/tests/adding-a-live-chat-button-to-the-page-ab-test-idea

## Hypothesis

If hesitant visitors can ask their question without leaving the page, more of them may go on to buy. Purchase conversion rate (CR) could improve, as long as the button does not take screen space away from the primary action.

## When to run this test

Repeated pre-sale questions in support tickets, session recordings of visitors hunting for contact details, or survey answers from people who could not find help suggest visitors leave before buying because a question goes unanswered. The expected behavior is that some visitors who start a chat go on to purchase in the same session, and page exit rate falls. If the team cannot respond consistently during support hours, do not turn the button on yet; set up capacity and a response standard first, because unanswered chats do harm.

## Primary metric

**Purchase conversion rate (CR)**: Compare purchases per eligible assigned visitor, not only among those who start a chat.

## Other metrics

- **User-started chat rate**: Measure chats initiated by the user among eligible assigned visitors who see the button.
- **Support response time**: Among started chats, measure time to the first human response and account for support hours.
- **Page exit rate**: Compare exits among eligible assigned visitors in the same session window.
- **Primary CTA completion rate**: Compare completion of the action linked from the main CTA using the same denominator.

## What to check before and during the test

- **Button**: Add one live-support entry point without opening the chat automatically.
- **Service**: Keep team availability, working hours and response standards consistent.
- **Placement**: Use one fixed position.
- **Page scope**: Define eligible pages before launch.
- **Mobile**: Check that the button does not overlap content or purchase controls.

## What to watch when reading the result

- A rise in user-started chat rate is not a win by itself. The decision rests on purchase conversion rate across all assigned visitors; do not ship if support response time gets longer or primary CTA completion rate drops.
- Keep the button in the same fixed spot for the whole test; if presence and position change together, you cannot tell which one moved the result. Make sure the button never covers the primary CTA on any device, paying extra attention on mobile, and that starting a chat does not require personal information. Otherwise the test measures setup friction rather than the button.
- Scope the result to the eligible pages you defined before launch and the support hours in effect during the test. Review visitors who arrived outside covered hours separately, and validate again before adding the button to other pages.
- Purchases are measured across everyone assigned, not just the visitors who chat, so the button's effect can look small. If the difference is unclear, start from the current conversion rate and the lift you most want to catch; the [sample size calculator](https://alidemirbas.com.tr/en/calculators/sample-size-calculator) will then give you the visitors each variant needs.

## Frequently Asked Questions

### What should be checked with support before launch?

Ensure consistent staffing, hours and response standards. A live-help button is not useful if requests go unanswered. Preserve the service standard throughout the test.

### Are more chat starts enough?

No. Use purchase conversion as the primary outcome and monitor time to the first human reply, reporting uncovered hours separately.

### Can the support button affect the main action?

Yes. Compare main-CTA completion using a common denominator. Avoid covering primary controls and check mobile overlap with content and purchase actions.

### What should the chat entry flow require?

Use one user-initiated entry point and request only information needed for service. Explain any verification requirement. Exclude automatic opening from this comparison. Provide the required privacy notice separately from any necessary consent.