The Same Traffic. The Same Product. 40% More Revenue.
A six-month CRO engagement for Complement Essential — a premium vegan supplement brand hosted on Shopify. 21 experiments, 17 winners, and a sustained 40% lift in product page conversion rate.
Strong Science.
Leaking Conversions.
Complement Essential is a vegan multivitamin built around 8 nutrients plant-based diets consistently miss — backed by 10 peer-reviewed studies and endorsed by MDs. Founded in 2017, they're the highest-ranked plant-based company on the INC. 500 list.
The product is exceptional. The page wasn't doing it justice.
"Every one of those baseline numbers had money sitting behind it — money the page was leaving on the table every day."
Where things stood
when we started.
Hover each card to see where it ended up. Every one of these numbers had money sitting behind it.
Four stages.
In order. Always.
Heatmaps, session recordings, funnel drop-off data, review mining, and competitive benchmarking — before a single hypothesis is written.
Find where the page loses people and why. Not every gap is a design problem. Most are structural — the right argument in the wrong place.
Each hypothesis scored on revenue impact, evidence strength, and speed. 34 scored. 21 made the cut. No gut-feel exceptions.
Minimum 2-week run. 95% statistical significance. No early calls. Post-test analysis before the next begins.
Three things
most audits miss.
The piecemeal vs. all-in-one comparison — "buying 8 supplements separately costs $180+, Essential costs $89" — was buried inside a product image carousel. Fewer than 18% of visitors ever swiped to see it. Their best value proof was functionally hidden.
A "Spin to Win" gamified popup was running on an $89 science supplement endorsed by cardiologists. It signals "this price is negotiable" — a brand problem masquerading as a conversion tool.
"Free USA shipping on orders over $95" was the first thing visitors read — anchoring them on a price point before they had any context for why $89 was reasonable. The first impression was cost, not benefit.
Every gap fell into
one of three buckets.
The compounding effect
of six months of evidence.
13 experiments killed in 3–4 days.
Every loss sharpened the next test.
These tests were called early — statistical data was clear within days that they were either flat or harmful. No test ran longer than 4 days once significance pointed the wrong direction.
Muted autoplay to reduce reading load. Killed day 3 — mobile LCP jumped 2.8s, bounce rate spiked immediately.
Performance cost cancelled any engagement benefit before visitors reached the content.
Manufactured urgency — "offer ends in X hours". Killed day 2 — trust scores in micro-surveys dropped sharply.
This audience researches before buying. Fake urgency reads as dishonesty, not motivation.
Surprise free gift callout adjacent to the CTA. Killed day 3 — ATC rate unchanged, subscription rate dropped 4%.
Attracted deal-seekers. Lowered buyer quality without lifting volume.
Collapsed ingredient explanations to summary bullets. Killed day 4 — search traffic CVR dropped 9% immediately.
High-intent visitors need depth. Page length is an audience question, not a design preference.
Discount offer triggered on exit intent. Killed day 2 — returned same brand-cheapening signal as Spin to Win.
Discount conditioning at exit is even more damaging than at entry — it rewards abandonment.
"Shared by 12,000+ plant-based eaters" badge by the buy button. Killed day 3 — no measurable impact on any metric.
Share counts carry no purchase intent signal. Visitors don't buy because others shared — they buy because others bought.
Kinetic type cycling through benefit statements in the hero. Killed day 3 — scroll depth below hero dropped.
Animation in the hero signals "advertisement" to a sceptical audience. It reduced rather than increased trust.
Added Essential + Omega Complex bundle as a third purchase option alongside one-time and subscribe. Killed day 4 — choice paralysis reduced ATC rate 6%.
Three options in the buy box is one too many. The bundle was right but the placement was wrong.
Media logos (Forbes, Well+Good, Verywell Health) added above the fold. Killed day 3 — no CVR change, distracted from product messaging.
Press logos convert brand awareness traffic. Complement's visitors arrive product-aware. The signal missed its audience.
"43 people viewing this right now" social urgency signal. Killed day 2 — trust survey scores dropped, felt manipulative.
FOMO mechanics require impulse-prone buyers. This audience deliberates. Live counts read as pressure, not proof.
Proactive chat — "Have questions about Complement?" triggered after 30 seconds. Killed day 4 — chat open rate 1.2%, interruption increased bounce.
Visitors on a long-form product page are reading, not asking. Interrupting a reader mid-argument is the wrong moment to offer help.
Table showing Complement's doses vs. competitor products. Killed day 3 — naming competitors introduced doubt about products visitors hadn't considered.
Competitor comparisons introduce alternatives into a page that should be closing a decision, not opening new ones.
A floating pill badge highlighting subscription savings, separate from the sticky CTA bar. Killed day 4 — two persistent elements competed visually and both performed worse.
The sticky CTA bar already handled subscription messaging. Doubling persistent elements divided attention without adding new information.
Three principles that held
across every test.
The piecemeal cost comparison was buried in a carousel. CRO often isn't about creating new ideas — it's about finding the thing that already works and making it visible. Before writing a single hypothesis, look for what's already true but invisible.
Countdown timers, spin-to-win wheels, fake urgency — all of it backfired. Vegan health consumers are unusually skeptical and research-driven. Every tactic borrowed from impulse-purchase playbooks actively damaged trust. The page that won was more honest, more specific, and more considered — never more aggressive.
The quiz, segmented review sequencing, returning-visitor module — none required machine learning or a CDP. They required understanding who was arriving and why, then designing the page to respond to that. Behavioural logic, not technology.
40% Higher Conversions.
Same Product. Same Traffic.
Every test was grounded in behavioural evidence. Every winning variant made the page more honest, more relevant, or more specific — never more aggressive.