Conversion rate optimisation explained. As a process, not a redesign.

The short answer. A loop, not a project.

Most founders' first encounter with conversion rate optimisation is a one-off exercise: hire someone to rebuild the homepage, launch it, hope for the best. That is a redesign wearing CRO's name. Genuine conversion rate optimisation, explained at its simplest, is a continuous loop run against evidence rather than a single decision made on instinct. It never really finishes, because there is always a next-highest-drop-off point to work on once the current one is fixed.

The distinction matters because the two approaches are measured completely differently. A redesign is judged on whether it looks better. A CRO loop is judged on whether the number moved, on a page that already carries traffic, within a defined window, against a specific hypothesis stated before the change went live. If you cannot say in advance what you expect to happen and why, it is not a CRO test yet, it is a guess with a launch date.

How it works in practice. The four-step loop.

The loop has four stages, and skipping any one of them is where most founder-led CRO efforts quietly fail.

Audit. Start with analytics, not opinion. Pull the funnel for the page in question and find the single step with the sharpest drop-off, whether that is visitors leaving before scrolling, starting a form and abandoning it, or reaching a pricing page and never clicking through. Session recordings and heatmaps, where available, show what the numbers alone cannot: whether people are hesitating, rage-clicking, or simply never seeing the call to action.

Hypothesis. Write down what you believe is causing the drop-off and why, in one sentence: "visitors abandon the form because it asks for information before establishing value, so shortening it to three fields will lift completion." A hypothesis this specific tells you exactly what to change and exactly what result would prove or disprove it. A vague goal like "improve the homepage" cannot be tested at all.

Test. Change the one variable the hypothesis names, and only that variable, so that whatever happens can be attributed to the change rather than to something else moving at the same time. On lower-traffic pages, a sequential before-and-after comparison over a matched time window is often more practical than a true split test, which needs enough simultaneous traffic on both versions to reach a reliable result quickly.

Measure and iterate. Let the test run a full cycle, review the result against the original hypothesis, and either keep the change and move to the next drop-off point, or discard it and try a different hypothesis for the same problem. Either outcome is useful. A disproved hypothesis still tells you something true about your visitors that a redesign never would.

For the underlying metric this loop is trying to move, and how to calculate it correctly by segment rather than as one blended figure, see how to calculate conversion rate.

A worked example. A founder's first ninety days running the loop.

Take a hypothetical, but representative, founder-led B2B services business. Website traffic is steady, but enquiries have flattened for two quarters despite growing spend on paid social. Rather than commissioning a redesign, the founder runs the loop against the contact page first, since it carries the most traffic of any conversion point on the site.

Weeks one and two: audit. The funnel report shows visitors reaching the contact page at a healthy rate, but only a small fraction ever submit the form. Session recordings show a pattern: visitors open the form, look at the ten fields requested, including budget range and preferred start date, and leave without typing anything.

Weeks three and four: hypothesis and test. The hypothesis is specific: the form asks for information too early, before trust is established, so cutting it to name, email and a single open question about what they need will lift completion, even if it means slightly less qualification data upfront. The team builds a second version with the shortened form and runs it against the original for a full four-week window, splitting traffic evenly rather than changing the live page outright.

Weeks five to eight: measure. The shortened form completes at close to three times the rate of the original. Lead quality dips slightly, some enquiries now need a qualifying call the old form would have filtered out earlier, but the net increase in booked calls comfortably outweighs the small rise in unqualified conversations. The team keeps the shortened form and writes the result down.

Weeks nine to twelve: iterate. With the form fixed, the next-largest drop-off is now visible: a chunk of visitors leave the page entirely before scrolling to the form at all. The loop starts again with a new hypothesis, this time about what sits above the fold, using the same audit, hypothesis, test and measure sequence as before. Nothing about the process changes between rounds; only the page element under test does.

Three months in, nothing about the website looks dramatically different to a visitor scrolling past. What has changed is a form that converts roughly three times better and a documented reason why, which is a far more durable result than a full relaunch that nobody can attribute a specific lift to.

What good looks like. A programme, not a sprint.

A healthy CRO programme has a visible cadence: a running list of hypotheses ranked by expected impact and effort, a test in progress at any given time, and a written record of what was tried and what happened, win or lose. That record is the actual asset. It stops the same idea being re-tested eighteen months later by someone who has forgotten the first result, and it builds a genuine picture of what this specific audience responds to, which rarely matches generic best-practice advice.

It also treats qualitative and quantitative evidence as equally important. The analytics tell you where the drop-off is; they rarely tell you why. A short user testing session, a handful of sales call recordings, or even reading the objections prospects raise before they convert usually explains a data point that the funnel report alone leaves you guessing at.

What good does not look like is a redesign every eighteen months justified by CRO language after the fact. If a business can point to a documented sequence of small, evidenced wins rather than one large unmeasured relaunch, that is the clearest sign the process, not just the vocabulary, is actually in place.

Pitfalls to avoid. Where founder-led CRO efforts lose their nerve or their evidence.

The most common pitfall is calling a test too early. A result that looks strong after three days is frequently noise, not signal, particularly on pages without heavy daily traffic. Committing to a minimum test window before starting, and holding to it regardless of how the early numbers look, removes the temptation to stop the moment a number moves in the right direction.

The second is changing several things at once and calling it a test. A new headline, a new image and a shortened form launched together might lift conversion, but you learn nothing repeatable about why, and the next test starts from the same lack of evidence as the first.

The third, particularly for ecommerce funnels, is treating checkout as fixed rather than testable. The Baymard Institute's research, drawn from more than 48 separate studies, puts the average cart abandonment rate at just over 70% as of 2025, and finds that fixing documented checkout usability issues, the average site carries around 39 of them, can lift conversion by as much as 35.26%. The most cited fixable cause, reported by 39% of abandoners, is being surprised by extra costs at checkout: shipping, taxes or fees not shown earlier in the flow. That single, well-evidenced fix is often worth more than a full checkout redesign.

The fourth is letting a founder or senior stakeholder's personal preference override a test result, sometimes called the loudest-voice problem. A CRO loop only works if the evidence wins the argument, even when the evidence contradicts the person paying for the work. Businesses that struggle with this are usually better served bringing in outside conversion rate optimisation services to run the loop with some distance from internal opinion, at least until the discipline of trusting the data is established.

A fifth, quieter pitfall is stopping the loop the moment it produces one good result. A single successful test proves the process works; it does not mean the funnel is now finished. The businesses that get the most out of conversion rate optimisation, explained plainly, are the ones that keep the loop running past the first win, working down the list of drop-off points one at a time, quarter after quarter, rather than treating one good quarter as the finish line.

Once the process itself is running, the specific terminology, macro conversions versus the smaller micro conversions along the way, becomes worth learning properly. See macro versus micro conversions explained for how to set up the right goals in the first place, and what is CRO in marketing? for where this work sits relative to acquisition spend.