100+ A/B tests on a $4,299 course funnel. One change nearly doubled sign-ups — twice.

Context

Mindvalley sells its certification programmes through a four-stage funnel: masterclass opt-in page → live webinar → replay → sales page, with a $200 deposit as the conversion event. Traffic ran into the hundreds of thousands of sessions across three programmes.

I owned the experimentation programme end to end — hypothesis, test design, build spec, QA, analysis, and the ship/kill decision. Every test was logged against a fixed template covering the full funnel, from landing page conversion through to projected revenue and value per attendee.

Over eight months I ran 100+ A/B tests across the three programmes. The experiments below are a selection — the ones that changed how I read the funnel, including two that looked flat or negative at the top and turned out to be significant wins underneath.

Experiment 1 — Social proof before the ask (Life Coach)

Problem. The Life Coach masterclass opt-in page converted at 15.65%. Every point of opt-in rate compounds through four downstream stages, so this was the highest-leverage surface in the funnel.

Hypothesis. The page asked for an email before establishing that the certification produced real outcomes for real people. Moving graduate proof above the opt-in would raise conversion.

What I tested. Control vs a rebuilt page leading with graduate testimonials. 3,509 vs 3,475 sessions, 9 days, statistically significant on landing page conversion rate.

Result.

What I'd do next. Replicate across Hypnotherapist and Business Coach, hold spend constant, and track through to true ROAS rather than opt-in.

Experiment 2 — Webinar revamp (Life Coach)

Problem. After the testimonials win, the next constraint moved down the funnel — attendance and offer exposure, not opt-in.

Hypothesis. Rebuilding the webinar and its supporting pages around a tighter narrative would lift both opt-in and show-up.

What I tested. Control vs revamped webinar and page set. 6,769 vs 6,828 sessions, 19 days, statistically significant on landing page conversion rate.

Result.

The 81% opt-in lift landing twice, on two structurally different tests against the same funnel, is the finding I trust most from this programme. It says the constraint on that page was credibility, and that the ceiling was far above where the page had been sitting.

What I'd do next. Reshoot the webinar and rerun, holding the page constant, to isolate which half of the change drove the show-up gain.

Experiment 3 — Quiz opt-in page rebuild

Problem. The quiz funnel was leaking between quiz completion and result opt-in — people finished the quiz and never claimed the result.

Hypothesis. The opt-in step was framed as a form, not as a reward for work already completed. Reframing it around the result the user had just earned would lift claim rate.

What I tested. Old vs new opt-in page, sequential test, 29 days per arm.

Result.

Logged as a winner and shipped. Same funnel, separate test on the quiz landing page: 8,042 → 9,145 leads on ~15,000 completions per arm, conversion 53.68% → 57.87% (+8%).

What I'd do next. Segment claim rate by quiz result bucket — the reframe likely works harder for some outcomes than others.

Experiment 4 — Shorter webinar (Life Coach)

Problem. The webinar ran 21:55 before the offer appeared. Drop-off before offer exposure was the largest single leak in the funnel.

Hypothesis. Moving the offer forward to 15:09 would raise both show-up quality and the share of attendees who actually reached the pitch, without costing conversion.

What I tested. 21:55 vs 15:09 webinar cut. 17,354 vs 17,596 sessions, 18 days.

Result.

Shortening the webinar by seven minutes got more people to the offer and moved revenue in the same direction. Shipped.

Experiment 5 — Evergreen vs launch webinar (Hypnotherapist)

Problem. Landing page conversion on the Hypnotherapist funnel was already at 24.88% against a 12–13% platform average. The page wasn't the constraint — the webinar format was.

Hypothesis. A launch-style webinar would convert a smaller number of more committed attendees at a far higher rate than the evergreen format, even if opt-ins stayed flat.

What I tested. Evergreen vs launch format, ~12,000 sessions per arm, 13 days.

Result.

The top of the funnel didn't move at all. Everything below it did. A test read only on opt-in rate would have been logged as a null result and thrown away.

What I'd do next. Rerun at 3–4x volume with deposits as the primary metric, and test the launch format on the Life Coach funnel.

Experiment 6 — Lead quality over lead volume (Hypnotherapist)

Problem. The funnel was optimised for opt-in volume, and opt-in volume was not translating into deposits.

Hypothesis. A more qualifying landing page would cut lead count but raise the deposit rate of the leads it did produce.

What I tested. Existing page vs a rebuilt, more qualifying version. 7,130 vs 7,186 sessions.

Result.

37% fewer leads, and each one 45% more likely to put down a deposit. This is the test I point to when someone asks why opt-in rate is a bad primary metric.

Also shipped

What this programme demonstrates

• Volume and rigour. 100+ documented A/B tests across three premium programmes over eight months, each logged to a fixed template covering the full funnel from session to projected revenue.
• Repeatable wins, not one-offs. Two independent tests on the same funnel both delivered an 81% opt-in lift. The finding replicated.
• Full-funnel measurement. The Evergreen and Lead Quality tests both looked flat or negative at the top and were significant wins at the bottom. Reading only landing page conversion would have thrown both away.
• Correct primary metrics. Deposit per lead and value per attendee, not opt-in rate.