Stop A/B Testing button colours. Here is what actually moves conversion rate for DTC Brands.

The button colour test has become a cultural shorthand for bad CRO. It is referenced in almost every conversation about conversion rate optimisation because it represents everything that is wrong with how most teams approach testing: cosmetic changes made based on opinion rather than evidence, tested because they are easy rather than because they address a real conversion barrier, and evaluated against insufficient traffic to produce statistically meaningful results.

The brands with the strongest CRO programmes are not testing button colours. They are testing things that have a plausible mechanism for significantly changing conversion behaviour — and they are choosing what to test based on evidence from their own visitor data.

Here is what that looks like in practice.

THE EVIDENCE-FIRST APPROACH

Before any test is designed, there should be a specific observation from visitor data that motivates it.

Not: “We think the headline could be clearer.”
But: “Session recordings show 68% of mobile visitors scroll past the CTA without clicking it. Heatmaps confirm that most engagement is concentrated around the product description rather than the CTA. We believe the CTA is either not visible enough or not compelling enough to prompt action from product-description readers.”

The difference: the first creates a test that might or might not be motivated by an actual visitor behaviour problem. The second creates a test that is specifically designed to address a pattern observed in real visitor data. If the hypothesis is correct, the test will produce a measurable improvement. If it is wrong, the test result itself is informative — you learn something about your visitors that was not previously understood.

THE FIVE ELEMENTS THAT CONSISTENTLY MOVE CONVERSION RATE

In order of typical impact for DTC ecommerce brands:

1. The product page headline and primary value proposition
   The first text a visitor reads after the product name. If this text is specification-led (“Magnesium Glycinate 400mg — 90 Capsules”) rather than benefit-led (“The sleep supplement that absorbs properly — because magnesium glycinate is the form your body actually uses”), it is a significant conversion opportunity. A visitor who does not understand within 5 seconds why this specific product is the right choice for their specific situation is significantly more likely to leave.
2. Trust signals in the above-the-fold experience
   Social proof — review count and average rating — has a disproportionate impact on purchase decisions for new customers. In the research supporting this, review stars visible before the add-to-cart button produce measurably higher conversion than review stars placed below it. The mechanism: perceived risk reduction at the exact moment the purchase decision is being formed.
3. The mobile CTA accessibility
   Already covered in Article 12. The sticky add-to-cart bar is one of the most consistently validated CRO improvements for mobile product pages — and one of the few that is effectively universal rather than dependent on specific audience behaviour.
4. Objection handling copy on the product page
   The top 3–5 reasons a visitor in your specific category would not buy from you should be explicitly addressed on the product page. Not in a FAQ buried at the bottom — addressed in the main product narrative or in a visible FAQ accordion above the fold on mobile. Visitors with unresolved objections do not convert. Visitors whose objections are addressed directly and honestly — “if you have tried magnesium before and it did not help your sleep, here is why this formulation is different” — convert at meaningfully higher rates.
5. Checkout friction reduction
   The checkout completion rate — what percentage of people who start checkout actually complete it — is the most directly manipulable conversion metric available. Field reduction (removing non-essential required fields), payment option addition (Apple Pay, Google Pay), and shipping cost revelation timing (showing shipping cost on the product page rather than at checkout) all produce measurable checkout completion improvements with low implementation effort.

THE HYPOTHESIS STANDARD

Before running any test, write a hypothesis using this format:

“We believe that [specific change] will [improve/reduce] [specific metric] for [specific audience] because [specific evidence from our own data]. We will measure success by [primary metric] improving by at least [minimum detectable effect]% within [timeframe].”

Every element matters:
Specific change: not “improve the CTA” but “add a guarantee badge directly below the add-to-cart button on mobile product pages.”
Specific evidence: not “best practice suggests” but “exit survey data shows 23% of non-buyers cite uncertainty about returns as their primary reason for leaving.”
Minimum detectable effect: this forces calculation of the sample size required — ensuring the test runs long enough to produce statistically meaningful results.

A test without a hypothesis is an experiment with no way to know what it proved.