Audit & baseline
Instrument first, test second
Most "CRO problems" turn out to be instrumentation problems: untracked CTAs, mis-firing events, missing funnel steps, two analytics tools disagreeing about the same number. Before any test runs, I audit what's being measured against what's actually happening on the page, fix the gaps, then baseline the funnel for at least two weeks so we have a clean starting line.
The baseline covers landing → primary CTA → form start → form submit → qualified action, plus device splits and acquisition source. Patterns show up fast: a desktop conversion rate that's twice mobile usually means the mobile form is broken; a paid-channel conversion rate well below organic usually means landing-page intent mismatch, not page design. The audit decides which problem to solve first.
For the broader systems context, see the Webflow developer overview, and for the full methodology in long form, maximising conversion rate with Webflow CRO walks through the audit-first approach end to end.
Hypothesis & testing
One variable, one test, one decision
A good test starts with a hypothesis grounded in actual user behaviour — session replay showing rage clicks on a non-button, form-analytics showing drop-off on a specific field, a heatmap showing nobody scrolls past the third fold. "I think the CTA should be bigger" is not a hypothesis; "drop-off jumps from 18% to 47% at the email field because the requirement to confirm is unclear" is.
Every test changes exactly one variable, runs for at least one full business cycle (almost always at least two weeks, longer on low-traffic pages), and is sized against the actual conversion volume so we know in advance whether it can resolve. Multi-variable tests are tempting but produce uninterpretable results unless you have e-commerce-grade traffic.
CRO work in production includes a fitness subscription funnel and a paid-education sales page for an ex-Wall Street trader, and the playbook is documented in CRO best practices.
Server-side vs client-side
Where to run the experiment
Client-side A/B tools (VWO, Convert, Optimize-style scripts) are the fastest to set up and the worst for Core Web Vitals. The flicker between control and variant on slow connections introduces an LCP delay that often swamps the lift you're trying to measure — especially on mobile. They're fine for above-the- fold copy tweaks; they're not fine for layout-level tests.
For anything more invasive than copy, I move the experiment to a Cloudflare Worker that picks the variant at the edge based on a hashed cookie, then serves the chosen HTML directly. No flicker, no client-side JS overhead, and the page stays fully cacheable per variant. Webflow renders both variants as separate published pages; the Worker routes between them. The result is server-side testing without giving up the Webflow editor.
Post-launch iteration
The decision log compounds
The single highest-leverage CRO artefact isn't any individual test — it's the decision log. Every test ships with a hypothesis, a result (winner, loser, inconclusive, or shipped-anyway-because-of-secondary- metrics), and a note on what we'd test next. Six months in, the log is a map of what works on your specific audience that no agency or new hire could reconstruct from scratch.
Losers matter as much as winners. They prevent the team from re-litigating the same ideas every six months and stop you from chasing whatever the latest CRO blog post recommends. The combination of clean instrumentation, single-variable tests, and a maintained decision log is what turns CRO from a tactic into a durable advantage.
CRO compounds best when the underlying site is built for it. See the programmatic Webflow page for landing-page architecture that's easy to test against, and the Airtable + Webflow page for keeping test variants and content data in one place.
Running conversion rate optimisation as a continuous loop, not a one-off project.
- Instrument & BaselineAudit the analytics, fix the gaps, and baseline the funnel for at least two weeks so every later test has a clean starting line.
- Single-Variable TestsOne hypothesis, one variable, sized for the traffic you actually have. Server-side variants where flicker would otherwise eat the lift.
- Decision LogEvery winner, loser, and inconclusive test gets logged. Six months in, the log is the most valuable CRO artefact you own.
Tell me what you need to build or improve.
Share the goal, the problem and what you already have. I will suggest a practical next step.