Skip to main content
loser-13.5% lift

General: Ghost / Outline Buttons Lose to Filled Buttons

Hypothesis

Ghost buttons (outline with transparent fill) test the principle that minimal design improves conversion — they don't.

CTALanding PageCross-Industrybutton designghost buttonoutline buttonCTAaffordance

Test Results

Key Learning

Problem: Each additional form field adds friction to the general, increasing the chance users abandon before completing their submission.

What was tried: Ghost buttons violate Fitts's Law and affordance principles: a button must look like a button to be clicked. The outline style reduces visual weight and click magnetism. Always use filled, high-contrast CTA buttons. Ghost buttons may be fine for secondary actions but should never be used for primary CTAs. (-13.5% change)

Why it failed: Not every CTA change improves conversion. Users may have preferred the original because it was clearer, more familiar, or better positioned.

How to Apply This to Your Site

This test showed that general: ghost / outline buttons lose to filled buttons led to a -13.5% drop in conversions. The change was tested on a landing page page in the cross-industry industry. Avoid replicating this exact approach — instead, consider testing the opposite direction or a more subtle variation.

Before you test: Consider that cta tests typically require adequate traffic to reach statistical significance. Run your test for at least 2 full business cycles to account for weekly traffic patterns.

This result reached 95% statistical confidence, meaning there is a very low probability the observed effect was due to chance. Results at this confidence level are generally considered reliable for making business decisions.

What Was Tested

Tests confirm ghost buttons consistently underperform: Test #199 on Freshegg.co.uk (-20% progression), Test #243 on (-1.1% progression with red ghost vs red filled). Both show that reducing button visual weight decreases click rates. This is one of the most replicated anti-patterns in CRO.

Methodology

Confidence Level
95%
Lift Range
-25.0% to -2.0%

Build On These Learnings

Save your own experiments, spot winning patterns across your test history, and stop repeating what's already been tried.

Related Experiments

Explore More Experiments