Portfolio · CRO & Experimentation
Fifteen years of conversion-rate and ecommerce operations work, written up as the experiments themselves — hypotheses, statistics, the tests that failed, and the ones I killed rather than reported. Every figure traces to a named experiment ID.
case studies — two employers, 4 mastheads
documented experiments behind them
documented win rate across both programmes
storefronts / mastheads / markets
Six iterations rebuilding the product detail page. The UK rollout is the strongest result in the programme.
Removing price from above the fold on campaign landing pages — the largest single effect in the programme.
One treatment, three storefronts, run in parallel to test whether a claim travels.
Unit-based tiers replaced with spend thresholds across 9 experiments, with a quantified discount-efficiency finding.
An iFrame PDP re-platformed to native Liquid, gated behind two A/B validations — and a +39.9% desktop 'win' exposed as an instrumentation artefact.
An image-led mobile menu. Clickthrough fell significantly — the signature of a shorter path, not a weaker one.
Quick-buy cut conversion in both iterations. Stopped in 4 days — significance wouldn't have changed the decision.
Right asset, wrong audience: a PDP founder tile that converts cold visitors while taxing loyal ones. The segment read became a targeting brief.
Trustpilot → +render → +USP icons: a laddered test that asked how much trust signal pays, not just whether it works.
Financing anxiety answered upstream of the cart — payment visibility at the point of decision, not checkout.
Killing the page-hops inside the bundle journey — +19.1% in AU, +5.6% in the UK, same week, same verdict.
A three-channel flash-sale programme across paywall, email and landing page — different price points win on different paywall types.
Churn-propensity-tiered renewal pricing — and the pivot to randomised testing when the model's precision fell short.
One variable — the 30-second auto-dismiss timer removed. Expiry friction was costing engagement, not creating urgency.
Three prices instead of take-it-or-leave-it: +70% conversions at SMH, +102% at The Age, both at 100% winning probability.
Revenue, traffic and baseline figures have been generalised to protect commercially sensitive data. Relative changes, confidence levels and experiment IDs are as measured.