A/B testing the best colour for a CTA button

A call-to-action (CTA) button is often treated as a finishing touch, yet colour can shape whether a visitor notices it, understands its purpose and feels ready to click. A/B testing replaces visual instinct with evidence: visitors are divided between a control and a treatment, and their behaviour reveals which version performs better for a defined conversion goal.

There is no universally superior button colour. The result depends on contrast, wording, page design, audience and the action being requested. Australian audiences deserve a test built around their actual context, from one-handed mobile browsing during a Sydney train commute to comparison shopping across major retailers and online marketplaces. A disciplined test can reveal which colour helps without sacrificing accessibility, brand consistency or downstream conversion quality.

Frame the hypothesis around user intent

Start with the decision a button is meant to make. “Book now” and “Get the guide” require different levels of confidence, while “Buy now” usually carries purchase intent. Write a hypothesis that names the current control, the proposed colour and the expected effect on the primary metric, such as completed bookings, qualified leads or purchases. This prevents a bright shade from winning because it attracts curiosity unrelated to the business goal.

Assign visitors to variants consistently. For a straightforward A/B test, keep the control and change only the button colour in the treatment. People should enter the experiment at the same point, receive the same page and have an equal opportunity to convert. New and returning visitors can be included, but they should be analysed separately if their intent differs; combining a first-time visitor’s demo request with an existing customer’s renewal click can distort the result.

Before counting a click as success, inspect what happens after it. A redirect can interrupt trust, so redirect page guidance is useful when the offer, enquiry form or next step depends on continuity. The CTA should describe the destination clearly, especially when the final page uses different colours, changes the price or asks for personal information.

Build an Australian-relevant colour palette

Colour must first be legible and understandable. Under WCAG 2.2 AA, normal-sized text needs a contrast ratio of at least 4.5:1 against its background, while large text needs 3:1. Check the actual rendered button, including its label, border and state on hover or focus. Do not use red versus green as the only cue, because colour-vision deficiencies and low-quality screens can erase the distinction. An icon, wording or shape should still communicate the action.

Local context can also influence colour performance. Blue may be familiar from Australian finance and technology brands, while red is often used to signal urgency in retail promotions. Those associations are not rules. Shoppers may encounter familiar branding from Coles, Woolworths, Amazon Australia, eBay and local stores, so a distinctive brand colour can help a CTA stand apart. Test within the design people actually encounter rather than against a generic colour swatch.

Mobile behaviour deserves particular attention. Commuters on Melbourne trams or Sydney trains often check offers one-handed, sometimes in variable light or while travelling through crowded areas. Button size, thumb reach and contrast may matter more than a small difference between two attractive desktop shades. Include mobile traffic in the design, but ensure the treatment is readable on smaller displays and usable with assistive technology.

When the destination is generic, a stronger CTA cannot repair vague context. The proceed-style experience shown by Siamese Lipstick is a useful warning: pair the colour treatment with a specific label and a landing page that confirms the offer. If the brand palette conflicts with the selected CTA colour, test a restrained accent rather than a wholesale departure from the site’s identity.

Control the test so the result is credible

Create the smallest clean experiment that can answer the question.

• Use the current button appearance as the control.
• Create one treatment with the proposed CTA colour.
• Keep copy, size, position, page and offer unchanged.

Then lock the measurement rules before launch.

• Set the sample size and minimum detectable effect in advance.
• Record visits, eligible impressions, clicks and conversions by variant.
• Freeze the design and avoid changing either version mid-test.

Run the test for long enough to include meaningful weekday and weekend patterns. Australian audiences span AEST, AEDT and several state time zones, so daily traffic can fluctuate by when each region begins work. A short weekday test may overrepresent Sydney or Melbourne office traffic and miss shoppers elsewhere. If the audience is small, extend the study rather than declaring a result from a handful of clicks.

Analyse the outcome with a method suited to the sample and the number of conversions. Look for a confidence interval around the difference in conversion rate, not just a p-value or a higher raw count. Check for novelty effects, where the new colour performs well briefly because it is unfamiliar. Segment results by device, campaign source and new versus returning status, but treat subgroup findings as clues unless they were planned and adequately powered.

Measure clicks, actions and downstream value

Use two levels of measurement. The click-through rate shows whether the treatment attracts more interaction, while the conversion rate shows whether those interactions produce the intended outcome. A colour may increase clicks by creating urgency but also increase cancellations, low-quality leads or returns. Report both figures alongside revenue per visitor, lead quality or completed bookings so the experiment reflects value rather than activity alone.

Track the complete path, including redirects and final outcomes. Suppose a campaign sends users toward Viber download pages. The button click is one event; reaching the intended destination, completing the requested action or returning later is another. Deduplicate repeated clicks, distinguish bots and internal traffic, and compare the same conversion definition for both variants. Otherwise, a treatment can appear successful merely because it attracts accidental clicks.

Segment only to understand differences. Compare mobile with desktop, paid with organic traffic, and traffic from Sydney, Melbourne, Brisbane, Perth, Adelaide and regional areas. Prices shown in Australian dollars, GST-inclusive totals and local delivery messages may alter behaviour on a landing page, so keep those elements consistent. Never select a winner from a tiny geographic subgroup when the overall result is inconclusive.

Confirm the result before changing the site

Require both statistical and practical confidence. A tiny improvement can be detectable with a large sample while having no worthwhile commercial effect. Set a business threshold before the test, inspect the confidence interval and confirm that the result survives sensible exclusions for bots, internal visits or accidental double clicks. If two colours are genuinely close, retain the control and use the result to rule it out rather than manufacturing false certainty.

Check the legal and customer-experience details before rollout. If the test uses analytics, advertising pixels or personalised variants, handle personal information consistently with the Privacy Act 1988 and applicable Australian Privacy Principles. Consent and disclosure matter for tracking technologies. Campaigns sent by email must also follow the Spam Act 2003, while button copy and destination claims must remain accurate under the Australian Consumer Law. A colour can be compliant by itself; misleading urgency or an unclear offer remains a problem.

Roll out gradually and watch the same metrics after release. Preserve the original control as a fallback, document the audience, device mix, test dates and decision threshold, and keep the winning colour associated with its tested wording and page context. Seasonal campaigns, new traffic sources and site redesigns can change the effect. The final button should be the version with credible evidence, inclusive design and a measurable commercial gain, rather than merely the shade with the highest early click count.