Color psychology in marketing is the idea that specific colors trigger specific feelings and actions, so a "red" or "orange" button should convert better than a blue one.
Real A/B tests don't support a single best color: what wins is a button that contrasts with the rest of the page, and the same color swap can win on one site and lose on another.
Treat color as a visibility problem first and a mood problem a distant second.
This page sets out what research on color actually shows, then what button color tests in our database show, with the numbers and who ran each test.
Is there a best button color for conversions?
No. Every "best color" claim we could trace comes from one site, one page and one test, and the winning color differs from site to site.
Look at the color swaps in our archive.
Pink beat green for Say It With Diamonds (11.49% vs 10.81% add-to-cart, per Elevate A/B Testing).
Green beat yellow for a Shopify store in a Clean Commit test (+4.6% conversion, not significant).
Red and orange beat green at GSM.nl (+5% clicks, sales inconclusive). Every color both wins and loses somewhere.
What the winners have in common is contrast.
The biggest color results in the database come from making the main button stand out from everything around it:
- Conversion Fanatics switched the add button on a black-and-grey women's apparel store from black, which blended into the page, to bright pink: +57% conversion rate. No sample size was published.
- Blend Commerce found that every button on Jackson's product page (Add to Cart, Rewards, Chat, Keep Me Updated) shared one color and weight. Giving Add to Cart its own color lifted conversion rate 18.40% and add-to-cart clicks 8.14%.
- King Living moved a grey checkout CTA to a more visible color, alongside clearer copy, and reported +15% revenue. The copy changed too, so color can't take all the credit.
None of these tests found a magic hue.
They found a button that was hard to see, then made it easy to see.
That matches the oldest relevant finding in psychology: the isolation effect.
In 1933 Hedwig von Restorff showed that an item which differs from its neighbors is remembered better than the items around it (Psychologische Forschung, vol. 18).
Our UX laws page covers how designers apply it.
Blend Commerce
@elevate_ab
Conversion Fanatics
Clean CommitA practical rule for picking a button color
- Pick one accent color that appears nowhere else on the page, and give it only to the primary action.
- Make secondary buttons visibly quieter: outline, grey or text link.
- Check contrast. WCAG 2.2 requires user interface components to have a contrast ratio of at least 3:1 against adjacent colors (Level AA). A button that fails this is hard to see for many shoppers, not just those with low vision.
- Keep the button text readable on the button color. Clean Commit's green variant shipped with "contrast-appropriate text," which matters as much as the fill.
For button placement, size and copy, see our add-to-cart button tests and CTA copy tests.
Copy changes tend to move more than color changes.
What button color A/B tests actually show
Button color is the most famous A/B test in marketing.
HubSpot still repeats the story that Performable's red button converted 21% better than its green one.
The write-up gives no sample size or significance level, and one result on one page says nothing about your page.
Our database has a better spread. Here are the color-only tests, with the result as each team reported it:
| Test | Change | Result | Ran by |
|---|---|---|---|
| Women's apparel store | Black add button to bright pink on a black and grey site | +57% conversion rate | Conversion Fanatics |
| Jackson's | Add to Cart given its own color | +18.40% conversion rate | Blend Commerce |
| High Voltage SEO | New color on "Download Checklist" (colors not disclosed) | +77.71% clicks | Convert Better |
| Beamax | Blue outlet link to red; a red banner was a third arm | +53.13% click-through | Beamax |
| Say It With Diamonds | Pink vs green buttons | 11.49% vs 10.81% add-to-cart | Elevate A/B Testing |
| OnSwitch | "Best Value" text green to red | +5.6% sales | OnSwitch |
| Shopify store | Yellow to green buttons sitewide | +4.6% conversion, not significant | Clean Commit |
| GSM.nl | Orange Buy Now buttons sitewide | +5% clicks, sales inconclusive | GSM.nl |
| OnSwitch | Yellow offer text and a red-boxed button | -10% conversion | OnSwitch |
| Encyclopaedia Britannica | Red or orange article links instead of blue | about -10% engagement | Encyclopaedia Britannica |
Three patterns stand out.
The big lifts are on clicks, not sales. High Voltage SEO's +77.71% and Beamax's +53.13% measure clicks.
A brighter button draws more clicks almost by definition; whether those clicks become revenue is a separate question.
GSM.nl saw exactly this: clicks rose 5%, but the sales effect was not significant.
Louder isn't always better. OnSwitch made one offer red, yellow and repeated, and conversion fell 10%.
Britannica found that non-blue links cut engagement by about 10%, likely because readers expect links to be blue.
A color that breaks a convention can hurt.
Most color-only effects are small. Once you remove tests where the button was nearly invisible to begin with, the remaining swaps land in the low single digits, which is hard to detect.
Worked example: was pink really better than green?
Elevate A/B Testing reported the Say It With Diamonds numbers in full: 667 add-to-carts vs 615, 11.49% vs 10.81%, from 11,495 visitors over 14 days. That's enough to check the result ourselves.
- Visitors per arm, implied by the counts and rates: about 5,805 (pink) and 5,689 (green).
- Relative lift: 11.49 / 10.81 - 1 = about 6.3%.
- Pooled rate: 1,282 / 11,494 = 11.15%.
- Standard error of the difference: sqrt(0.1115 x 0.8885 x (1/5,805 + 1/5,689)) = 0.0059.
- z = (0.1149 - 0.1081) / 0.0059 = about 1.16, two-sided p of about 0.25.
- 95% interval for the difference: about -0.5 to +1.8 percentage points.
By our calculation, this result is not statistically significant at the usual 95% level: the interval includes zero, so green could plausibly be just as good.
Keeping pink was a sensible call because it was the existing design, but this test can't show pink is better.
The company's other numbers fit that reading. Checkout completion showed no notable difference, per the fuller write-up.
This is typical. Detecting a 6% relative lift on an 11% base rate takes far more than 5,800 visitors per arm.
Our guides to statistical significance and minimum detectable effect show how to size a test before you run it.
Small, noisy wins also tend to shrink when shipped, which is the winner's curse.
When a button color test is the wrong test
- Low traffic. If you can't reach a few thousand conversions per arm, a color-only test will usually come back inconclusive. Fix the offer, copy or page layout first.
- The button is already obvious. If the primary button already contrasts with the page, swapping one strong color for another rarely moves sales.
- You only measure clicks. Use add-to-cart, orders or revenue per visitor as the main metric, and treat clicks as a supporting signal.
Browse more results in the full test database or the product page layout and design tests.
Color psychology: what research shows
Separate two things: what lab research shows about color, and what A/B tests show about buttons. They overlap less than marketing blogs suggest.
The claims marketers repeat
The most quoted figure comes from a 2006 review by Satyendra Singh in Management Decision, which says people decide about a product within 90 seconds and that "about 62-90 percent of the assessment is based on colors alone" (Singh, 2006).
It's a literature review rather than a new experiment, and it's about overall product judgments, not button clicks.
Stronger evidence links color to brand perception. In Labrecque and Milne (2012), published in the Journal of the Academy of Marketing Science, blue logos were linked to competence and red logos to excitement.
That's useful for branding. It says nothing about whether a red button sells more.
Replication problems
The research base is thinner than its reputation.
Andrew Elliot, one of the field's leading color researchers, wrote in a 2015 review that the literature on color and psychological functioning "is at a nascent stage of development," that "much of the research remains underpowered," and that this includes the first study of an effect, naming the first red-and-attraction study, which Elliot co-authored.
Specific findings have weakened under retesting:
- Red and test performance. A 2020 meta-analysis of 67 effect sizes by Timo Gnambs found that, after correcting for publication bias, there was no evidential value for an effect of red on intellectual performance. Early studies showed much larger effects than later ones.
- Red, blue and task type. A widely cited 2009 Science paper by Mehta and Zhu found red helped detail tasks and blue helped creative ones. A later attempt failed to replicate its anagram result (Steele, 2014).
Elliot's broader point applies directly to button tests: the meaning of a color depends on context.
Red can mean "sale," "danger" or "error" depending on the page, the category and the culture.
That's why the same red button can win on one site and lose on another.
What this means for your site
- Use color psychology for brand decisions, where the research is strongest, not for button picks.
- For buttons, aim for contrast and a clear visual hierarchy. Those are the changes behind our largest color wins.
- Never assume a color will lift conversion. Even a well-reasoned change, like OnSwitch's louder red offer, can lose.
- If you test color, test it on a high-traffic page, measure sales rather than clicks, and size the test first.
For other biases with stronger test records, see our psychology hub and the cognitive biases guide.