# Applying Filters Instantly or With a Button A/B Test

Page: https://alidemirbas.com.tr/en/lab/ab-test-library/tests/applying-filters-instantly-or-with-a-button-ab-test

## Hypothesis

With instant filtering, each selection shows its result right away, so visitors may close in on the product they want faster; applying filters together lets them set several preferences before the list moves. If visitors usually narrow down with one filter, instant filtering could lead on successful product discovery; if they combine several, the Apply button could.

## When to run this test

Visitors who leave the list right after filtering, session recordings of people losing their place as the list refreshes on each selection, or visitors who miss the Apply button show that filter use on your product listings is not turning into found products. The behavior to expect is that visitors in one option reach a product page more often and with fewer retries. If the problem is a slow results service, a weak filter set or the product order, fix that first; keep the design, product order and list unchanged during the test.

## Primary metric

**Successful product discovery after filtering**: Among all eligible assigned visitors, compare those who reach a product list or product page within the same time window. Define the success event before launch.

## Other metrics

- **Filter usage rate**: Compare eligible assigned visitors who apply a filter.
- **Product interaction in results**: Measure product views or selections from the same defined list impression.
- **Results refresh time**: Measure time from the filter event to the updated product list.
- **Listing exit rate**: Compare visitors who leave the list after interacting with filters.

## What to check before and during the test

- **Application**: Compare results that update after each filter selection with results that update after one Apply action.
- **Selection stability**: Check that selections do not reset and controls do not move unexpectedly.
- **Performance**: Use the same results service and measure the effect of repeated requests.
- **Outcome**: Review filter interaction alongside purchases.
- **Device**: Review desktop and mobile separately.

## What to watch when reading the result

- Decide on successful product discovery after filtering. Instant filtering may raise the filter usage rate on its own, but that rise alone does not earn a rollout unless it carries through to product interaction in results and to purchases. Compare the listing exit rate with the same definition too.
- Watch the results refresh time in the instant option; a slow refresh or unnecessary animation and transitions would have you measuring a performance loss rather than the application model. Do not trigger refreshes so often that they wear users out, and check that selected filters stay visible in both options.
- Keep the filter set the same so the difference can be tied to how filters are applied. Scope the finding to the listings you tested; read desktop and mobile separately, because a gain on one device can hide a loss on the other.
- When the gap is narrow, put each arm's assigned visitors and the visitors who found a product into the [A/B test significance calculator](https://alidemirbas.com.tr/en/calculators/ab-test) to see whether the difference between the two rates is larger than random variation would produce. If it is not significant, note that the two options did not separate in this sample and consider the one with the lower repeated-request cost.

## Frequently Asked Questions

### Does more filter use mean instant filtering won?

No. Read interaction changes alongside purchases. Use the predefined successful product-finding rate as the primary outcome.

### What technical checks does instant filtering need?

Updates must preserve selections and avoid unexpected control movement. Use the same results service and measure repeated-request cost and update time. Avoid excessive refreshes or animation.

### Can desktop and mobile results be combined?

Review each separately, including list exits. A gain on one device can conceal a loss on another.