Most businesses treat conversion rate optimisation as something they do once, redesign the page, change the button colour, move on. That approach might produce a one-time lift. It does not produce compound growth. A CRO process is different. It is a structured cycle that runs continuously: research what is happening, hypothesise why, test the hypothesis, learn from the result, and feed that learning into the next cycle. Each loop builds on the one before it. After 10-20 cycles, the cumulative effect is substantial, and it compounds in a way that one-off redesigns never do. At Apexure, we have run this process across 300+ clients over the past decade. The eight steps below are the same ones we follow on every engagement, whether the client is a SaaS startup or a financial services enterprise. The steps are consistent. The specifics change based on the data.

What is the CRO process?

The CRO process is a systematic approach to improving your website or landing page conversion rate through data collection, hypothesis building, controlled experimentation, and iterative learning. It is not “changing things and hoping”. It is the scientific method applied to marketing. Two things make the CRO process different from a typical redesign or optimisation project:

1. It is cyclical. Every completed test generates new data, which feeds new hypotheses, which feeds new tests. The learning compounds.
2. It is data-driven. Changes are based on quantitative metrics (analytics, funnel data) and qualitative evidence (heatmaps, user feedback), not design opinion or competitor imitation.

The 8-Step CRO Process

Step 1: define conversion goals and KPIs

Before you audit anything, define what “conversion” means for this specific page or funnel. A demo booking is not the same as a newsletter signup. An add-to-cart is not the same as a completed purchase. Each conversion type has different benchmarks, different traffic requirements for testing, and different upstream factors. Set primary KPIs (conversion rate, cost per conversion, revenue per visitor) and secondary KPIs (form start rate, scroll depth, engagement rate). Without clear KPIs, you will not know whether a test won or lost, and your CRO process becomes guesswork dressed up as science.

Step 2: conduct a quantitative CRO audit

The quantitative audit tells you what is happening. Pull at least 30 days of data from GA4 to establish a reliable baseline. Focus on:

This audit does not tell you why visitors leave. That comes in Step 3.

Step 3: gather qualitative insights

Numbers tell you where the problem is. Qualitative research tells you why. Without this step, your hypotheses will be blind guesses, and your tests will produce inconclusive results.

Heatmaps and scroll maps (Hotjar, Microsoft Clarity) show you where visitors click, how far they scroll, and which elements they engage with. If visitors are clicking something that is not clickable, your visual hierarchy is misleading. If 70% of visitors never scroll past the hero section, your below-the-fold content might as well not exist.

Session recordings let you watch real visitors move through your site. Look for rage-clicking, u-turns (scrolling back up after reaching a section), and form hesitation. We have found more conversion killers in 20 session recordings than in 20 reports.

Surveys and feedback give you the visitor’s own words. A single on-page poll, “What stopped you from completing this form?”, can surface objections that no amount of analytics will ever reveal.

Step 4: map the user journey and identify friction

Combine the quantitative and qualitative data into a picture of the visitor’s actual journey, not the journey you designed, but the one they experience. Where do they hesitate? Where do they get confused? Where do they abandon? Build a simple journey map: Entry point → Key pages → Conversion point → Post-conversion. Annotate each stage with the data from Steps 2 and 3, drop-off rates, friction points, user feedback. The friction points you identify become the raw material for hypotheses in Step 5. Common ones we find in audits: message mismatch between ad and landing page, forms asking for too much too soon, trust signals buried below the fold, CTAs that blend into the page, and mobile experiences that were clearly an afterthought.

Step 5: build Data-Backed hypotheses

A hypothesis is not “let’s try a different headline.” It is a structured statement linking a specific problem to a proposed change and a measurable expected outcome.

Format: “Based on [data/observation], we believe that [change] will [outcome], which we will measure by [metric].”

Example: “Based on heatmap data showing 65% of visitors clicking the hero CTA but only 12% completing the form, we believe that reducing the form from 7 fields to 3 will increase form completion rate by 20-30%, measured by form submission events in GA4.” This format forces precision. It prevents the team from testing changes based on gut feeling, design preference, or what a competitor is doing. Every test should trace back to a specific data point from Steps 2-4.

Step 6: prioritise with a framework

With a list of 10-15 hypotheses, you need to decide what to test first. Without a framework, the loudest voice in the room wins, which is how teams spend months testing hero images when the real conversion killer is the form. At Apexure, we use the EPIC framework: each hypothesis is scored 1-5 on Experimentation, Priority, Impact, and Cost. The highest combined score runs first.

Step 7: design and run controlled tests

Choose your testing method based on traffic volume and the scope of the change:

Testing discipline matters as much as test design:

• Run to 95% statistical significance. Stop early and your result is noise.
• Change one variable per A/B test. Otherwise you cannot attribute the result.
• Calculate sample size upfront. Know how long the test needs to run before you start.
• Account for external factors. Seasonality, promotions, and budget changes all affect results. Run tests for at least 2-4 weeks.

Step 8: analyse, document, and iterate

Analysis is not “which version won.” It is understanding why it won, what that tells you about your audience, and how it shapes the next cycle. After every test, answer four questions:

1. Was the result statistically significant? If not, the test is inconclusive, not a failure.
2. What segments responded differently? Break results by device, traffic source, and geography.
3. What does this tell us about the audience? A benefit-focused headline beating a feature-focused one tells you something about your visitors’ priorities.
4. What should we test next? Every completed test should generate at least one new hypothesis.

Document everything. Maintain a test log with the hypothesis, variant details, duration, sample size, result, confidence level, and the learning. After 20-30 tests, the patterns in that log become your most valuable CRO asset. This step feeds directly back into Step 1. The cycle continues. Each loop is faster and more informed than the last because you are building on accumulated knowledge, not starting from scratch.

The CRO process tool stack (2026)

Common CRO process mistakes

How Apexure runs the CRO process

We follow this exact 8-step process on every engagement. The specifics vary by client, a SaaS startup with 5,000 monthly visitors gets a different treatment than an enterprise with 500,000, but the structure is always the same.

Results from our CRO process

Related reading:

• Landing Page CRO Strategy: A Practitioner’s Guide

• The Complete Guide to Conversion Rate Optimisation

• Landing Page A/B Testing Framework Playbook

• CRO Testing: Everything You Need To Maximize Conversions

• CRO Consultant: What They Do, What They Cost & When to Hire One

• CRO and UX: The Ultimate Duo to Maximize Conversions

Sources 8 Last updated Mar 2026

View all sources

• How to Create a CRO Processcxl.com
• CRO Steps and Tips for 2026contentsquare.com
• CRO Strategy for 2026hubspot.com
• CRO Process in 5 Easy Stepsvwo.com
• CRO Strategy Developmentinvespcro.com
• 111 Key CRO Statistics for 2026loopexdigital.com
• The CRO Process Blueprintspeero.com
• How to Implement a CRO Processunbounce.com