What is conversion rate optimization?

The short answer

Conversion rate optimization is the practice of getting more of your existing visitors to do the thing you want (enquire, buy, book) by changing the experience and measuring whether it worked.

The important word is measuring. Redesigning a page because it feels better is not CRO. It is decorating. CRO means you had a hypothesis, you tested it against the current version, and the data told you which one won.

The appeal is straightforward: you have already paid for the traffic. Lifting the conversion rate is the only growth lever that does not cost more per visitor.

How the process actually works

1. 1Find where people leave. Analytics tells you which step loses them. This is where most of the value is, and it costs nothing but attention.
2. 2Work out why. Session recordings, surveys, talking to your sales team. The data says where; the humans say why.
3. 3Form a hypothesis. "Visitors do not enquire because the price is not mentioned anywhere and they assume it is expensive."
4. 4Test it properly. Against a control, for long enough, with enough traffic to mean something.
5. 5Keep the winner. Learn from the loser. Both are results. Most tests fail, and that is normal.

That step four is where most in-house CRO falls apart. A test called early because it "looks like it is winning" is not a test. It is confirmation bias with a stopwatch.

What CRO can honestly deliver

Compounding, unspectacular gains. Anyone promising to double your conversion rate is selling you a story about their last lucky test.

For e-commerce, the big opportunity is usually the checkout: Baymard Institute's research puts average cart abandonment around 70%, and a meaningful share of that is fixable friction rather than genuine hesitation. For lead generation, it is usually the form and the clarity of the offer.

The honest framing: CRO does not usually find one enormous win. It finds a series of small ones that stack, and it stops you making expensive changes that would have hurt you, which is a benefit nobody puts on a slide.

When you should not run CRO at all

This is the part CRO agencies rarely volunteer. Testing has a minimum viable traffic level, and below it the whole discipline stops working.

• You do not have enough traffic. With a handful of conversions a month, a test will take a year to reach significance, and you will call it early and fool yourself. Fix traffic first with search or SEO.
• Your tracking is broken. Then you are not testing. You are guessing with extra steps.
• The page is fundamentally unclear. Do not A/B test a button color on a page that never explains what you sell. Rewrite it, then test.
• The offer is the problem. No headline test fixes a proposition nobody wants.

For low-traffic sites, the honest recommendation is to stop testing and start fixing the obvious things, clarity, speed, one clear next step, fewer form fields. Best practice beats bad testing when the sample size is small. When you do have the traffic, conversion work becomes the cheapest growth you can buy.

Where the wins usually hide

If you are new to this, you do not need a testing tool to start. You need to look in the four places where conversion is most often lost, because they are the same four places in almost every business.

1. 1The form. Every field you add costs you completions. Baymard's checkout research found most checkouts ask for far more information than they need. The same is true of enquiry forms, ask for what you need to have a conversation, not what would be nice to know later.
2. 2The price question. If you do not address cost anywhere, a large share of visitors assume the worst and leave. You do not have to publish a price list. You do have to acknowledge the question exists.
3. 3The mobile experience. Not "does it work on mobile", is it genuinely usable on a phone, on mobile data, one-handed, outdoors. Test it that way. It is sobering.
4. 4The next step. One clear action per page. Five equally weighted buttons is the same as none, because it hands the decision back to a visitor who was hoping you would make it for them.

Those four are not tests. They are known problems with known fixes, and on a low-traffic site they are where you should spend your entire effort. Best practice beats bad testing when the sample size is too small to tell you anything.

Testing earns its place once you have volume, when a real experiment can reach a conclusion in weeks and not months. Until then, fix the obvious things, and put the money you would have spent on a testing program into getting more people to the site in the first place.

The mistakes that make CRO useless

Most failed CRO programs fail for the same handful of reasons, and none of them are about the tooling.

• Calling tests early. A test that "looks like it is winning" after three days is noise. Stopping when you like the number is not testing, it is confirmation bias with a stopwatch.
• Testing trivia. Button colors on a page that does not explain the offer. You are optimizing a rounding error.
• Ignoring the mobile experience. Most of your traffic, and often the worst experience. Check the page against Core Web Vitals on a phone before you test anything on it.
• Forgetting accessibility. A page that fails WCAG is excluding real customers with real money, which is a conversion problem before it is a compliance one.
• Testing on broken tracking. If the metric is wrong, the winner is arbitrary.

Do the boring things first: clear offer, fast page, short form, one next step, accessible markup. On most sites that produces more improvement than a year of experiments, and it costs nothing but the willingness to look honestly at what you have.

Frequently asked questions