That last part trips people up, because most explanations of CRO skip straight to a promised percentage: some flat "average uplift" figure, quoted as if it applies to every business, every page, and every audience equally. Treat any number like that with suspicion. Below is what CRO actually is, how the loop works, what separates a real result from a lucky-looking coincidence, and why an honest answer to "how much will this improve my conversion rate" is always "it depends what the test finds."
What "conversion rate optimisation" actually means
Most marketing spend is aimed at getting more visitors: paid ads, SEO, social content, all pointed at growing the top of the funnel. CRO works on the traffic a site already has. Instead of trying to attract more people, it asks a narrower question: of the people already arriving, why aren't more of them buying, booking, or enquiring, and what specifically would change that.
The word that separates CRO from ordinary website tweaking is "tested." Anyone can have an opinion about which headline reads better or which button colour looks more clickable. CRO treats that opinion as a starting hypothesis, not an answer, and runs it against real visitors before deciding it's correct. A page can look objectively worse to a designer and still convert better, because visitor behaviour and aesthetic taste are not the same thing. The only way to know which one wins is to test it.
That's also what separates CRO from a redesign. A redesign changes how a site looks, usually on a schedule set by a rebrand, a new season, or simply because the old one felt dated. CRO changes what a test proves is actually costing conversions, which is sometimes a full redesign and just as often one form field, one headline, or one button moved higher up the page. Treating the two as interchangeable is how a redesign budget gets spent on a page that looks newer and converts at exactly the same rate it did before.
The loop: audit, hypothesis, test, ship
CRO isn't a single action, it's a repeating cycle, and skipping a step is usually where it stops working:
- Audit. Look at the whole funnel, not just one page, to find where visitors are actually dropping off. This is diagnostic, not decorative: the goal is finding the single highest-impact leak, not compiling a list of everything that could theoretically be improved.
- Hypothesis. Turn the leak into one specific, falsifiable statement: not "the page could be better," but "visitors are abandoning the form because it asks for a phone number before they've decided to trust us." A hypothesis that can't be proven wrong isn't a hypothesis, it's a hunch wearing a lab coat.
- Test. Change one variable at a time and run it against enough real traffic to produce a result that means something. Changing five things at once and watching the number move tells you nothing about which of the five did it.
- Ship the winner, kill the loser. If the change wins, it goes live. If it loses, it's dropped and the next hypothesis gets tested. Both outcomes count as progress, because either way, the funnel is now better understood than it was before the test ran.
Then the loop starts again, because a funnel that's been optimised once isn't optimised forever. Traffic sources shift, offers change, and a page that converted well against last quarter's visitors can quietly stop working against this quarter's without anyone noticing until someone tests it again.
What counts as a real result, not noise
The part most explanations of CRO leave out entirely: a test result isn't automatically real just because a number moved. Checking a test too early, and stopping the moment the numbers look good, is one of the most common ways a test produces a false win.
Statistician Evan Miller, in a 2010 analysis, showed what happens if you check a test's significance after every new visitor and stop the instant it crosses the standard 5% threshold. In his worst-case simulation, the false-positive rate reaches 26.1%, more than five times higher than the 5% most people assume they're working with. The fix isn't complicated: decide roughly how much traffic the test needs before it starts, and don't trust the number until the test has actually run its course.
This is why a serious CRO process moves slower on low-traffic pages than a founder checking a dashboard every morning would like. A page getting a few hundred visitors a month simply can't produce a trustworthy result in a week, no matter how good the change looks on day two. Patience isn't a nice-to-have here. It's the difference between a change that actually converts better and a change that just happened to look good during a short, noisy stretch of traffic.
Why nobody can promise an uplift before testing
This follows directly from the last section: if a result only becomes trustworthy once a test has run long enough against enough real traffic, then no one, however experienced, can know what that result will be before the test has started. A specific promised percentage, quoted before an audit has even looked at the actual page, isn't a forecast. It's a guess dressed up as one.
What determines the size of any fix is entirely specific to that page, that traffic, and that audience, at that moment. A form that's leaking half its submissions on mobile responds very differently to a fix than a headline mismatch between an ad and the page it lands on. We've already broken down the specific ways a page leaks visitors before anyone has tested anything: weak proof, unclear next steps, forms asking for too much. Which of those, if any, applies to a given site, and how much fixing it actually moves the number, is exactly what the audit and the test exist to find out. Anyone naming a percentage before that work has happened is skipping the only step that makes the number honest.
CRO vs a redesign vs a new landing page
These three get confused constantly, and they're not the same decision.
A redesign is a scheduled visual overhaul: new look, new layout, usually driven by a calendar rather than evidence about what's actually underperforming. A landing page is a specific tool: a standalone page built for one offer and one action, as opposed to a full website carrying navigation and multiple goals. We've written more on the difference between a landing page and a website, since picking the wrong page type for a given piece of traffic is itself a common, fixable leak.
CRO sits above both of those decisions rather than replacing either one. It's the ongoing testing layer that can be applied once a page, whether that's a full website or a dedicated landing page, already exists and already has traffic. Building the right kind of page in the first place is a separate call from testing and improving whichever page you've already built. Skipping straight to "let's redesign it" without first testing what's actually broken is how a business ends up with a newer-looking page that converts at the same rate as the old one.
Where The Social Target fits in
Nine years, 600+ clients, and 50+ active accounts have taught us that this discipline travels. The habit of isolating one variable, running it long enough to mean something, and reading the result honestly behaves the same way whether the thing under test is an advert or a checkout step.
Want to see what this looks like as an actual engagement? See how we run CRO as a service: what we actually audit, how a test gets built and read, and why we won't quote you an uplift percentage before we've looked at your specific funnel.
↳ Frequently asked
01What does CRO stand for?
Conversion rate optimisation. It's the practice of increasing the percentage of visitors to a site or page who complete a specific action, using testing against real traffic rather than opinion or a redesign.
02What is the difference between CRO and a redesign?
A redesign is scheduled: it changes how a page looks because the calendar or a new brand says so. CRO is evidence-led: it changes whatever a test shows is costing you visitors, which might be a full rebuild and might be one form field. Confusing the two is how a business pays for a redesign and never finds out whether the thing that was actually broken got fixed.
03How long does a CRO test need to run?
Long enough to gather enough real visitor traffic to produce a trustworthy result, which depends entirely on how much traffic the page already receives. Low-traffic pages need longer than high-traffic ones. Stopping a test early because the numbers look good is one of the most common ways a result turns out to be noise rather than a genuine win.
04Can conversion rate be improved without more traffic?
Yes, that's the entire point of CRO. It works on visitors a site already has rather than trying to attract new ones, which is why it's often the highest-return next move for a business that's already spending on paid media or has already invested in a proper site build.
05What's a good conversion rate?
There's no single universal number. It varies enormously by industry, traffic source, page type, and what action is being measured, buying, booking, or simply enquiring. Comparing a site's conversion rate against a generic published benchmark is far less useful than comparing it against that same site's own past performance and testing whether a specific change moves it.
06Do you need a developer to run CRO tests?
Not for most changes. Copy, layout order, form fields, and calls to action can usually be tested without custom development. A hypothesis that requires a genuinely new flow or feature is the exception where development work becomes part of the test itself.