# Live Activity Notice or Low-Stock Warning A/B Test

Page: https://alidemirbas.com.tr/en/lab/ab-test-library/tests/live-activity-notice-or-low-stock-warning-ab-test

## Hypothesis

A recent-purchase message points to demand, while a low-stock message points to availability, and either could speed up the decision for a different reason. Within the shared product pool, whichever message shoppers find more believable could lift purchase conversion rate (CR) per assigned visitor further than the other.

## When to run this test

Survey answers asking how popular a product is or whether it will stay in stock, and visitors who add to cart and leave without buying, show shoppers hesitating after looking at an item. Whichever message they find more believable should pull add-to-cart and purchases ahead of the other. When missing product details or the price cause the hesitation, look at those first. Ahead of the launch, verify that each message uses that product's own real data, that no single generic number is copied across products, and that the count is never tied to a counter that keeps climbing on its own.

## Primary metric

**Purchase conversion rate (CR)**: Compare purchases per assigned visitor across the shared set of products eligible for both messages.

## Other metrics

- **Add-to-cart rate**: Check whether urgency changes the first action.
- **Time to purchase**: Measure time to purchase; time on the product page is a different measure.
- **Return rate**: Track returns over the same follow-up period in both arms.
- **Perceived brand trust**: If surveyed, ask a direct trust question. NPS measures likelihood to recommend, not trust itself.

## What to check before and during the test

- **Message type**: Compare a real recent-purchase count with actual remaining stock.
- **Presentation**: Keep placement, design and supporting detail consistent.
- **Time window**: Set the activity-counting window before the test starts.
- **Device**: Review mobile and desktop results separately.
- **Eligibility**: Define a shared data rule and fallback in advance for when either message is unavailable.

## What to watch when reading the result

- Purchase conversion per assigned visitor within the shared product pool settles which message wins. Do not confuse time to purchase with time on the product page, measure brand trust with a direct trust question rather than NPS, and a conversion gain that weakens trust is not grounds for a launch.
- Show each visitor only the single message type they were assigned; if both appear together, the result tells you nothing about either one. Check in both arms that the message does not disrupt the price and CTA hierarchy, read mobile and desktop separately, and compare returns once each arm's orders have had time to come back.
- Products that switch to the fallback when one message becomes unavailable can throw the arms off balance. Before comparing purchases, confirm that assigned visitor counts stayed in the planned proportions (sample ratio mismatch); if they did, see whether the purchase gap is significant with the [A/B test significance calculator](https://alidemirbas.com.tr/en/calculators/ab-test). If the gap is still unclear, note that the pool may have been too small, and do not extend the result to products outside it.

## Frequently Asked Questions

### Which products are eligible for the two urgency messages?

Use a shared pool with valid data for both messages. Define fallback behavior when either claim stops being valid.

### Can both messages appear simultaneously?

Not in this comparison. Show only the assigned message, with comparable position and design. Fix the activity-counting period before launch.

### Can NPS measure brand trust?

NPS measures willingness to recommend, not trust itself. Use a direct trust measure for that question and avoid fabricated activity or stock claims.

### Does longer time on the product page establish urgency?

No. Time on page differs from time to purchase. Use purchase conversion within the common eligible product pool as the primary outcome and inspect devices separately.