What CRO really is (and is not)

0:00 What CRO really is (and is not)

Somebody changed a button from green to orange, conversions went up for a week, and a whole industry of myths was born. If that's what you think conversion rate optimisation is, this lecture will change your mind. By the end, you'll be able to define CRO in a way that keeps you honest, describe the research-to-learning loop that real programmes run, choose the right metric for a business, and explain what's changed in twenty twenty-six, from the end of Google Optimize to consent-limited data and AI in the workflow.

0:39 Why CRO matters

Why does this matter? Because traffic is expensive, and it keeps getting more expensive. Every business pays for visitors in some way: ad spend, content time, partnerships, or discounts. CRO is about getting more value from the visitors you already pay for. And the maths is compelling. If you improve the rate at which visitors become customers, every future marketing pound, dollar or rupee works harder. But only if the improvement is real. Fake wins, from bad tests or cherry-picked data, cost you twice: once in the wasted effort, and again when you build on something that isn't true.

1:22 CRO, defined

Here's a definition that keeps you honest. CRO is the systematic process of increasing the share of visitors who take a valuable action, by understanding why they don't, testing changes, and learning from the results. Notice three words. Systematic, meaning a process, not random ideas. Valuable, meaning the action that matters to the business, not just clicks. And understanding, meaning research comes first. Think of CRO like a doctor rather than a vending machine. You don't put in a tactic and get out a result. You diagnose, treat, and monitor.

2:01 What CRO is not

What CRO is not. It's not copying best practices from a blog, because what works for one audience can hurt another. It's not a list of tricks like fake countdown timers, which damage trust and may break consumer law. It's not only A/B testing, since many improvements are simply fixing broken things. And it's not just the website. Emails, ads, onboarding, pricing pages, apps and even checkout policies are all part of the conversion journey. CRO applies wherever someone decides whether to take the next step.

2:38 The CRO loop

Now the loop real programmes run. One, research: analytics, recordings, surveys, user tests. Two, hypotheses: evidence-based statements of what to change and why. Three, prioritisation: choose what to do first. Four, action: test, build or drop. Five, analysis: read results honestly. Six, learning: document it and feed the next round of research. It's a flywheel. Each turn makes the next one faster and smarter, because you know more about your customers.

3:09 Simple example: choosing the metric

A simple example about metrics. A subscription meal-kit company in London celebrates a test that raised sign-ups for the first box. But the variant pushed a bigger first-box discount. Three months later, far fewer of those customers are still subscribed. Sign-ups was the wrong primary metric. A better choice would have been revenue per visitor over a longer window, or at least paying subscribers after the second order, with refunds as a guardrail. The lesson: pick the metric closest to real business value that you can measure within the test.

3:48 Picking the primary metric

Let's make the metric question practical, because it decides whether CRO helps or hurts. For an online store, revenue per visitor is often the best primary metric, because it captures both conversion rate and order value. If a variant gets more people to buy but pushes them to cheaper items, conversion rate alone would call it a win. For lead generation, measure qualified leads or booked meetings, not raw form fills, since a shorter form can double submissions and halve quality. For subscriptions, look beyond the first payment to retention. And always pair the primary metric with guardrails: refunds, cancellations, support contacts, page speed. The primary metric tells you if you won. The guardrails tell you what the win cost.

4:40 Realistic example: Riyadh furniture retailer (illustrative)

Now a realistic scenario with illustrative details. A furniture retailer in Riyadh hires an agency to make the website convert better. The brief is exactly that, one sentence. The agency rewrites it before starting. Goal: increase revenue per visitor from paid social traffic on mobile product pages. Baseline: last twenty-eight days of revenue per visitor and purchase rate, reconciled with the order system. Constraint: no price changes this quarter. First step: a two-week research sprint with funnel analysis, twenty-five filtered session recordings and a short on-site survey. Only then, a prioritised backlog and test plans for the top three ideas. The client gets clarity, and the agency gets a fair way to be judged.

5:30 Watch me do it: the baseline sheet

Watch me build a baseline sheet, because you can't prove improvement without one. I open a simple table. Sessions and users for the last twenty-eight days, with a note on whether consent mode is modelling some data. The primary conversion rate, taken from the backend order system as the source of truth. Revenue per visitor. Average order value, or lead quality if it's a lead business. Checkout or form completion rate. Conversion split by mobile and desktop. The top three landing pages and their conversion rates. And Core Web Vitals on key templates. Then I compare analytics orders to backend orders. If they differ by more than a few percent, I fix tracking before running any test.

6:21 What's changed

What's changed recently? First, the tool market reset. Google Optimize was sunset in September twenty twenty-three. Teams moved to dedicated platforms like VWO, Optimizely, AB Tasty, Kameleoon and Convert, or product-focused and open-source options like GrowthBook, PostHog and Statsig. Second, data is consent-constrained. Cookie consent rules and browser protections mean analytics is incomplete, so many teams analyse experiments in their own data warehouse too. Third, AI is in the loop, summarising research, suggesting ideas and drafting reports. But AI-generated insights are hypotheses until evidence supports them. None of this changes the loop. It changes the tools around it.

7:04 Who owns CRO?

Who owns CRO? In small businesses, usually the marketer or founder, with help from a developer and a designer. In larger organisations, a dedicated experimentation team working with product, analytics, design, engineering and legal. What matters is not the org chart, but clear ownership of the loop: someone owns research, someone owns the backlog, someone owns test quality, and someone owns the decision to ship. When nobody owns the decision, tests run forever and nothing changes.

7:37 Common mistakes

Common mistakes. Copying tactics without research. Measuring success with the wrong metric, like clicks instead of revenue per visitor or qualified leads. Testing on broken tracking. Declaring wins from before-and-after comparisons, which mix your change with seasons, campaigns and news. Buying an expensive tool before having research or hypotheses. And using dark patterns that lift a number briefly while damaging trust and inviting regulators.

8:05 Recap

Recap. CRO is the systematic process of increasing valuable actions by understanding why people don't take them, testing changes and learning. It runs as a loop: research, hypotheses, prioritisation, action, analysis and learning. Choose metrics close to real business value, with guardrails. And work with today's reality: new testing tools, consent-limited data and AI as an assistant. Try this now: build the baseline sheet from the lesson text for your own site, reconcile analytics with your backend, and rewrite your CRO goal as a one-paragraph brief with a goal, baseline, constraints and first step.