product Run conversion as a repeating loop, gather behavioural insight, prioritise hypotheses by evidence, build and run the A/B test, analyse honestly, ship the winner, and feed it back into the next round.
The objective
Most tests are flat, that's expected. The win comes from running the loop fast and honestly: watch real users, prioritise by evidence, ship the winners, and let data overrule opinion every time.
Throughput and honesty beat clever ideas. Gather behavioural insight with Hotjar + 
GA4, prioritise hypotheses by evidence (ICE/PXL), run the A/B in VWO, call the flat ones flat, and ship only validated winners. Not doing: HiPPO-driven testing, calling winners before significance, or A/B-ing pages without the traffic to settle.
Cycle & hand-offs: Gather insight runs monthly and starts the cycle. Hand-offs: Gather insight → Prioritise hypotheses: straight away; Prioritise hypotheses → Build & run the test: straight away; Build & run the test → Analyse: once the test has reached its planned sample size or end date; Analyse → Ship winner & bank insight: straight away. Any other link waits for a person. Sign-off steps stay with a person; everything else can run on its own once armed.
How it runs
The workflow, exactly as it runs inside . It plays through step by step,
or pick any node on the map.
GA4; find the step bleeding the most users.
Hotjar / 
Clarity on that step, log rage-clicks, dead-clicks, stalls.VWO (or 
Optimizely); for B2B, personalise the variant with 
Mutiny.GA4 after rollout.Setup
The tools this system drives, and the context you bring once. Connect your own accounts, or swap in your team's equivalent.
Proven in market
CRO & experimentation loop is built and verified by Growth Division, the GTM team behind 130+ companies. Here is what the same play has produced for their clients.
600% lower CPA, 2 channels validated
Higher conversion across the journey
.avif)
+118% demo bookings
Growth Division The small, hard-won details that decide whether it works, from every run.
The biggest wins are usually friction you can already see in replay, not clever copy. Watch first, then test the obvious fix, and let it run to significance before you believe it.
Prioritisation doesn't make ideas better, it stops you spending scarce test slots on low-evidence ones. Most tests are still flat; you just waste fewer of them.
Low-traffic steps don't suit a clean A/B, make the obvious replay-driven fix and read it on the funnel instead. Expect a lag: the verdict takes weeks of cohorts before the line settles.
Get started
Live in an afternoon. Clone it, connect your tools, set your inputs, and turn it on.
Clone the CRO & experimentation loop into your workspace, wired end to end.
GA4, Hotjar, Clarity +3 more.
Your ICP, targets and brand rules, once.
It runs monthly on autopilot, checking in with you only where it matters.
Next launch: Week of 3 Nov 2026 · 1 slot left
In 30 minutes we find where your go-to-market is leaking and map the first system we’d launch. You leave with a plan whether or not you work with us.