Pattern #77: Filled Or Ghost Buttons

Pattern #77  Tested 7 timesFirst tested by Luke Hay Recently tested by Andrey Andreev on Mar 22, 2025

With A Total Of 4 Evidence For Version B It Is Very Likely Worse

LOSSES
-5
-4
-3
-2
-1
FLAT
+1
+2
+3
+4
+5
WINS

Measured by the sum of negative and positive tests.

A B
Filled Or Ghost Buttons (Variant A) Filled Or Ghost Buttons (Variant B)

Expected Median Effects Of B

-9.6%

Progression

(7 tests)

-

Leads

-

Signups

-

Engagement

0.3%

Sales

(4 tests)

-

Revenue

-

Retention

-

Referrals

-1.1%

ANY PRIMARY

(7 tests)

Tested on

Tests

Pattern #77: Filled Or Ghost Buttons
Was Tested On Online.metro-cc.ru by Andrey Andreev

Replaced

Isolated

Test # 526 on Online.metro-cc.... by $conducted_test->test->user_->first_name . ' ' . $conducted_test->test->user_->last_name Andrey Andreev    Apr 17, 2024 Test link

Variation B Was Worse

-1 Evidence

  • Measured by completed searches   |   p-val 0.00000001

  • Measured by completed purchases   |   p-val 0.01390546

Zoom In
  • 45.0% (201591 successes out of 447656 visits)
  • 0.8% (3691 successes out of 447656 visits)
A
  • 40.7% (182583 successes out of 448353 visits)
  • 0.8% (3489 successes out of 448353 visits)
B
Global

In this experiment, a feint gray background search bar was tested against a higher contrast one containing a blue border and a filled blue color button style.  Impact on searches and purchases was measured. (Test was inverted to fit the filled-vs-ghost button pattern).

Replaced

Isolated

Test # 582 on Online.metro-cc.... by $conducted_test->test->user_->first_name . ' ' . $conducted_test->test->user_->last_name Andrey Andreev    Mar 22, 2025 Test link

Variation B Was Insignificant

+0 Evidence

  • Measured by adds to cart   |   p-val 0.00000001

  • Measured by completed sales   |   p-val 0.26761044

Zoom In
  • 65.9% (409735 successes out of 621571 visits)
  • 1.3% (7979 successes out of 621571 visits)
A
  • 64.4% (399731 successes out of 620899 visits)
  • 1.3% (8110 successes out of 620899 visits)
B
Listing

In this experiment, the plus and minus quantity icons near the add to cart button were made more visible. In the variation they gained a higher contrast from a full background color. Impact on add to cart and sales was measured.

Replaced

Isolated

Test # 538 on Volders.de by $conducted_test->test->user_->first_name . ' ' . $conducted_test->test->user_->last_name Daria Kurchinskaia    Jun 20, 2024 Test link

Variation B Was Insignificant

+0 Evidence

  • Measured by clicks on checkbox   |   p-val 0.04084492

  • Measured by completed purchases   |   p-val 0.31060228

Zoom In
  • 45.0% (12619 successes out of 28032 visits)
  • 41.2% (11542 successes out of 28032 visits)
A
  • 44.2% (12361 successes out of 27993 visits)
  • 40.8% (11408 successes out of 27993 visits)
B
Checkout

In this experiment, a less visible (ghost button style) legal confirmation box, was tested against a more visible one (filled state with higher contrast). Impact on error rates (from submitting an incomplete form) and sales was measured.

Replaced

Isolated

Test # 310 on Backstage.com by $conducted_test->test->user_->first_name . ' ' . $conducted_test->test->user_->last_name Stanley Zuo    Jul 25, 2020 Test link

Variation B Was Insignificant

+0 Evidence

  • Measured by visits to a detail page   |   p-val 0.00000001

  • Measured by membership sales   |   p-val 0.68327197

Zoom In
  • 12.7% (37853 successes out of 298012 visits)
  • 0.4% (1233 successes out of 298012 visits)
A
  • 9.4% (27929 successes out of 296985 visits)
  • 0.4% (1249 successes out of 296985 visits)
B
Listing

In this experiment, the style of a button leading to view detailed casting calls on a listing page was changed. In the A version the style was a filled high contrast blue background, and the B variation had a feint "ghost button" style. 

Replaced

Isolated

Test # 307 on Volders.de by $conducted_test->test->user_->first_name . ' ' . $conducted_test->test->user_->last_name Michal Fiech    Jul 17, 2020 Test link

Variation B Was Worse

-1 Evidence

  • Measured by clicks on the secondary button   |   p-val 0.00000001

Zoom In
  • 8.4% (2273 successes out of 27167 visits)
A
  • 6.5% (1724 successes out of 26451 visits)
B
Thank You

This experiment measured a shallow click goal on a button that would encourage to repeated the action that was just completed (in this case a contract cancellation). In the control version (A) a thank-you screen shows a filled button style, and the variant (B) there was a ghost button. As a note, I also flipped the A-B in this experiment for the purpose of matching it to our ghost button pattern, which means that Volders in fact was starting  out with a ghost button to begin with. 

Replaced

Isolated

Test # 243 on Goodui.org by $conducted_test->test->user_->first_name . ' ' . $conducted_test->test->user_->last_name Jakub Linowski    Jun 05, 2019 Test link

Variation B Was Insignificant

+0 Evidence

  • Measured by visits to pricing page   |   p-val 0.78386322

Zoom In
  • 8.4% (1053 successes out of 12493 visits)
A
  • 8.3% (1046 successes out of 12553 visits)
B
Home & Landing

In this experiment, full red background buttons were changed to ghost buttons (red outline and transparent fill)

Replaced

Isolated

Test # 199 on Freshegg.co.uk by $conducted_test->test->user_->first_name . ' ' . $conducted_test->test->user_->last_name Luke Hay    Sep 13, 2018 Test link

Variation B Was Worse

-1 Evidence

  • Measured by clicks on any of the primary buttons

Zoom In
A
B
Home & Landing

In this experiment, standard (filled) buttons were tested against ghost buttons.

Leaks

Leak #39 from Booking.com   |   Jan 3, 2020 Listing

Booking A/B Tests Ghost Buttons, But Their Hopes Fade Away As Predicted

A - Dec 18, 2019
B - Dec 18, 2019 REJECTED

Ghosts buttons. We know that this style of buttons is net negative based on this evidence-based pattern. So when Booking ran this experiment, it's no surpirse that it was eventually rejected - a predictable replication. View Leak

-0.5 Evidence

Leak #106 from   |   Apr 17, 2024 Global

Etsy A/B Tests This High Contrast Button Style

A - Oct 11, 2023
B - Oct 11, 2023 IMPLEMENTED

Etsy ran a little A/B test where they probably tried to encourage more searches with the help of a stronger visual cue in their search bar. Instead of using a  ghost-like icon only approach for the primary search, they stylized it as a high contrast button. Fast forward a few months and it now looks like the variation was implemented.  View Leak

-0.5 Evidence

Related Patterns

For each pattern, we measure three key data points derived from related tests:

REPEATABILITY - this is a measure of how often a given pattern has generated a positive or negative effect. The higher this number, the more likely the pattern will continue to repeat.

SHALLOW MEDIAN - this is a median effect measured with low intent actions such as initiating the first step of a lengthier process

DEEP MEDIAN - this is derived from the highest intent metrics that we have for a given test such as fully completed signups or sales.