E-commerce A/B Test Experiments

50+ conversion experiments across D2C brands, including the ones that failed.

Project facts

Client
Seven D2C brands
Period
2022 to 2025
Duration
Ongoing
Testing tools
Convert, Shogun, Google Optimize
Brands tested
KettlebellKings, Flygrubs, Silvon, RDN, Baby Merlin, TribeWod, HappyPo

Not every test wins

Not every test wins. In fact, roughly 40% of our tests failed. But that is the point.

We ran disciplined experiments with clear hypotheses, measured rigorously, and learned from both wins and losses.

Impact across the portfolio

How I run a test

  1. Identify the opportunity via analytics (high traffic, low conversion pages)
  2. Form a hypothesis based on competitive research or user behaviour data
  3. Design a variant with the specific change isolated
  4. Run until statistical significance (minimum 2 weeks, 1000+ actions)
  5. Document results regardless of outcome
  6. Implement winners, learn from losers

Six experiments

Flygrubs collection page: subscription highlighting

Flygrubs (chicken feed, US market) · Collection page

Hypothesis: Highlighting subscription benefits and bestseller products on the collection page will increase conversion and subscription sign-ups.

Rationale: Why the collection page? It had the most organic and paid traffic. More traffic means bigger impact.

Mobile
MetricControlVariantChange
Conversion rate6.53%7.68%+17.5%
Add to cart rate13.80%15.03%+9%
Avg session duration4:324:41+3%
Odds to win–95%–
Desktop
MetricControlVariantChange
Conversion rate10.75%12.38%+15.2%
AOV$43.94$46.58+6%
Odds to win–97%–

Outcome: Winner. Implemented site-wide.

Learning: Customers needed clearer subscription value props at the browsing stage, not just on the product page.

KettlebellKings theme and product page upgrade

KettlebellKings (fitness equipment, US market) · Site-wide theme and product detail pages

Hypothesis: Upgrading to the latest theme version and adding A+ content on product pages will improve engagement and conversion.

Dec 2022 to Apr 2023
MetricBeforeAfterChange
Overall CVR1.97%2.48%+25.8%
Sessions with PDP views42.4%53.8%+26.8%

Outcome: Winner. Major conversion lift from the combined improvements.

Learning: Sometimes you need a bigger swing (a full theme upgrade) rather than incremental tweaks.

RDN hero section redesign

RDN (home decor, US market) · Homepage hero section

Hypothesis: A custom hero section based on competitive research will increase product discovery and engagement.

Mobile
MetricChange
Product views+30.88%
Engagement+20.36%
Bounce rate+8.13% (worse)
Desktop
MetricChange
Product views+0.67%
Engagement+1.49%
Bounce rate−0.80% (better)

Outcome: Partial win. Mobile showed a significant product view lift; desktop was flat. We implemented it for mobile and iterated on desktop.

Learning: Mobile and desktop users behave differently. Sometimes you need device-specific solutions.

Silvon homepage improvements

Silvon (bedding, US market) · Homepage

Hypothesis: A new homepage template with better information hierarchy and competitive positioning will improve revenue.

Desktop
MetricChange
Conversion rate+34.40%
Scroll depth+6.78%
Bounce rate+1.31% (slightly worse)
Mobile
MetricChange
Scroll depth+23.92%
Avg session duration+4.84%
Conversion rate−13.64%

Outcome: Split decision. Desktop revenue significantly up; mobile engagement up but revenue down. We implemented the desktop version and continued iterating on mobile.

Learning: Longer sessions do not always mean more purchases. Mobile users may have been overwhelmed by the additional content.

KettlebellKings social proof: video on homepage and product pages

KettlebellKings (fitness equipment, US market) · Homepage and product pages

Hypothesis: Adding customer videos will increase social proof, engagement and conversion.

Mobile
MetricChange
Engagement+7.19%
Avg session duration+7.19%
Bounce rate−0.08% (slightly better)
Revenue goal−1.81%
Desktop
MetricChange
Engagement−5.07%
Avg session duration−0.53%
Revenue goal−1.92%

Outcome: Failed. Engagement improved on mobile but did not translate to revenue. Desktop was worse across the board.

Learning: Video content can increase time on site without improving conversion. Users may have been distracted rather than persuaded. We reverted and focused on static social proof instead.

Baby Merlin product page A+ content

Baby Merlin (baby products, US market) · Product detail page

Hypothesis: Adding UGC images, testimonials and detailed benefits below the fold will improve engagement and conversion.

Desktop vs mobile
MetricDesktopMobile
CVR−10.5%−2.47%
ATC rate−14.7%−14.4%
Session duration−10.3%−13.4%

Outcome: Failed. All metrics declined.

Learning: More content is not always better. For the Baby Merlin audience of new parents, the original concise product page was more effective: they wanted quick answers, not more information to process.

What the tests taught me

Mobile and desktop need separate strategies

Results often diverged significantly between devices.

More content can hurt conversion

Especially for audiences seeking quick decisions.

Engagement is not revenue

Longer sessions and more scroll depth do not automatically mean more purchases.

A ~60% win rate is healthy

If you are winning 90%+ of tests, you are not being ambitious enough.

Compound small wins

3% here, 5% there, 15% there. It adds up to 25%+ over time.

More projects

All projects · Book a call