# Category Ordering A/B Test on Listing Pages

Page: https://alidemirbas.com.tr/en/lab/ab-test-library/tests/category-ordering-ab-test-on-listing-pages

## Hypothesis

An order built from recent category-page visits may move the categories shoppers open most to the top and get them to the products they want sooner. If that turns into purchases, the data-based order could deliver a higher purchase conversion rate (CR); if it only adds clicks, the hand-picked order may do as well or better.

## When to run this test

Analytics showing visitors reaching popular categories through search instead of the menu, heatmaps with scrolling to the bottom of the list, or category names that come up often in site search suggest that frequently visited categories sit too far down. If the data-based order helps, visitors reach the category they want sooner in the data-based order, and that carries through to purchases. Base both orders on observed browsing behavior rather than guesses; if the real problem is that category names are unclear, look at that first.

## Primary metric

**Purchase conversion rate (CR)**: Compare purchases among all eligible visitors assigned to each category order, not only visitors who click a category.

## Other metrics

- **Category click-through rate**: See whether the categories moved higher in the list receive more clicks.
- **Revenue per visitor (RPV)**: Compare revenue per eligible assigned visitor to capture changes beyond category clicks.
- **Average order value (AOV)**: Review order value alongside conversion and RPV.
- **Exit rate**: Track whether visitors leave the listing experience more often.

## What to check before and during the test

- **Order**: Compare the current category order with the same categories ordered by category-page visits over the 90 days before launch.
- **Keep constant**: Do not change category names, images, product ranking or card design.
- **Discovery**: Review category clicks and product discovery alongside purchases.
- **Conversion**: Include all eligible assigned visitors in the purchase comparison, including those who do not click a category.
- **Device**: Review mobile and desktop separately.

## What to watch when reading the result

- More clicks on promoted categories are not a win by themselves; judge the purchase conversion rate (CR) across all eligible assigned visitors. If conversion rises while revenue per visitor (RPV) or average order value (AOV) falls, or the exit rate climbs, that is not enough to ship.
- The result depends on this category set and on an order built from this data window; validate it again when you refresh the order with new data or move it to another listing page. Read each device on its own. Both versions should keep category spacing and mobile readability intact, and key categories such as New Arrivals and Outlet should stay easy to find; if either version breaks that, do not attribute the gap to the order.
- Each visitor should see the same assigned order for the whole test; otherwise the effects of the two arms blur together. A borderline gap does not show that the two orders work equally well; enter each arm's visitors and purchases into the [A/B test significance calculator](https://alidemirbas.com.tr/en/calculators/ab-test), and if chance could explain the gap, hold off on a decision.

## Frequently Asked Questions

### When is popularity-based category ordering worth testing?

When the current order pushes frequently sought categories too far down. Define the data period and popularity signal beforehand. Preserve access to important low-click categories rather than treating click volume as their only value.

### Does more clicking on promoted categories mean success?

No. Use category-page purchase conversion as the primary outcome. Revenue per visitor and AOV help explain changes in purchasing. Increased category clicks alone do not establish additional sales.

### Which visitors belong in the conversion denominator?

Include all eligible visitors assigned to each arm, not just those clicking a promoted category. Selecting clickers after exposure creates different populations and weakens the comparison.