What Is CRO? First Steps From Traffic to Orders

Conversion rate optimization, or CRO, is the habit of getting more of your existing visitors to complete an action that matters to your business, such as placing an order, requesting a quote or calling. The short version: decide what counts as a conversion, measure it reliably, find where people drop out, then change one thing at a time and check whether it helped. CRO does not bring more visitors. It makes the visitors you already have more valuable, which is why it often pays off before more traffic spending does.

E-commerce teams, local businesses and B2B marketers need a sane starting point, and a benchmark percentage is not it: a good rate depends on your sector, product, traffic source and the intent behind each page. If you are weighing landing page SEO against conversion, landing page SEO and conversion balance covers that trade-off; the process here runs from first measurement to first test.

What CRO is, and what it is not

People use the term for everything from button colors to site redesigns. A tighter definition helps you decide what to work on first.

A process, not a bag of tricks

CRO is a loop: measure, diagnose, hypothesize, test, learn, repeat. A colorful button might win once, but the loop is what produces gains you can explain and repeat. Teams that skip measurement and jump to tactics usually cannot tell afterward whether anything improved, because they never defined what "improved" means.

What it does not replace

CRO does not fix a weak offer, wrong audience or broken product. If visitors arrive with the wrong intent, no layout will convert them. It also does not replace customer service: a page can say the right things and still lose a sale if the delivery promise is not kept. Treat CRO as a way to remove friction and confusion from a decent experience, not to cover for a poor one.

The maths: rate, denominator and what counts

A conversion rate is the number of people (or sessions) that completed an action divided by the number that had the chance to complete it, shown as a percentage. The interesting decisions are in the numerator and the denominator.

Choosing the numerator

The numerator is the action you count. For a shop it might be completed orders; for a plumber, tapped phone links and submitted contact forms; for a software company, demo requests. Decide on the main action and write it down in a sentence. Counting everything that moves, from scroll depth to newsletter sign-ups, produces a number nobody can act on.

Choosing the denominator

You can divide by sessions or by users. Per-session rates treat repeat visits as separate attempts, which can make a long buying process look weaker than it is. Per-user rates group repeat visitors but need reliable user identification. Either choice is fine as long as you keep it fixed over time, since comparison only works with a consistent definition.

Google's Analytics help page on key events notes that reports such as Landing pages and User acquisition include a Key events column showing how often users triggered key events, and that explorations can use metrics such as Key events and Session key event rate. Menu names may change over time, so check the current key events help page (Opens in a new tab) for the exact wording.

Macro and micro conversions

A macro conversion is the outcome you ultimately want. Micro conversions are steps on the way: viewing a pricing page, adding to cart, starting a form. They are useful as diagnostics, because they show where the path breaks. They are poor as goals. A site with plenty of add-to-cart events and few orders has a checkout problem, and celebrating the first number hides the second.

Reading a low rate carefully

A low rate is not automatically a bad page. Informational traffic converts differently from comparison-shopping traffic, and a visitor reading a guide is rarely ready to buy on the first visit. Before judging, split the number by traffic source, device and page purpose. A weak mobile rate next to a healthy desktop rate points at the mobile experience, not at the offer.

Instrument before you optimize

You cannot improve what you cannot see, and you cannot trust what you never tested. The first working task in CRO is measurement.

What a key event is

Google's help page describes a key event as a measurement of an action that is particularly important to the success of your business. When users trigger it, Analytics records it and shows it in reports. The page says any collected event can become a key event: you create or identify an event for the action, then mark it as a key event. It also says you can create a Google Ads conversion from a key event, which helps keep conversion counts consistent between the two products. The page does not cover default key events and refers to a separate article for that topic.

The first event for a small site

Here is the decision rule: pick the event closest to money or real demand that happens often enough to measure. For a shop with plenty of orders, that is the order itself. For a site that sees only a handful of orders a month, begin with a more frequent upstream event, such as reaching checkout, and track orders alongside it. A local service business can start with call taps and form submissions; GA4 key events for click-to-call and chat shows how to set those up. A B2B site usually starts with demo or quote requests.

Measurement sanity checklist

• Is the main goal written as one sentence?
• Does the event fire exactly once per real action, and not again when the thank-you page reloads?
• Have you tested it yourself with a real order or form submission?
• Is the event marked as a key event?
• Are your own team's visits and test orders excluded?
• Does your consent setup respect your privacy policy and local rules?

Consent and data protection requirements differ by country, so align your measurement with your own privacy notice and, if needed, with a compliance adviser.

Map the funnel on paper

List the path from entry to completion: landing page, product or service page, cart or form, payment or submission, confirmation. Attach one event to each step and confirm each fires. Then read the step-to-step transitions. The biggest relative drop is your first inspection target. When the counts are tiny, a swing of a few people can look like a large percentage change, so always show the raw numbers beside the rate.

Where visitors get stuck: three kinds of evidence

Numbers say where people leave. They rarely say why. Combine three sources to get from symptom to cause.

Quantitative data

Look at which pages people enter on, where they leave and how this differs by device and traffic source. If the problem is concentrated in one segment, a site-wide fix will likely miss it. Compare like with like and prefer trends over single days.

Behavior recordings and click maps

Session recordings and click maps show what visitors do: buttons that are tapped but do nothing, images mistaken for links, fields that are filled in wrongly. Look for patterns across many sessions rather than one memorable video. Mask form fields in recordings and check that the tool respects consent.

Direct feedback

Ask people. A one-question exit survey, a post-purchase question or a review of support chat transcripts reveals objections your analytics cannot. If several customers ask the same question, your page is missing that answer.

Example scenario: a bicycle repair shop

Example scenario: A local bike repair shop sees plenty of visits to its service page but few booking requests. Analytics shows most visitors are on phones and leave before reaching the form at the bottom. Recordings show people tapping the phone number, which is plain text rather than a link. Support emails keep asking what a basic service costs. Together the three sources suggest three hypotheses: move the booking form up, make the number tappable and explain how pricing works on the page.

Page-by-page first checks

Product pages should answer what it is, whether it suits me and whether I can trust the seller, with price, delivery and returns near the top. Category pages should help people reach the right item quickly, so check filters, sorting and empty-result handling. Landing pages should push one main action and match the headline to the query or ad that brought the visit. Forms and checkout should drop every field that is not truly needed and explain errors in plain language without erasing what the visitor typed.

From hypothesis to test

With a list of problems in hand, resist fixing everything at once. When ten things change together you cannot learn which one worked.

Write a proper hypothesis

A hypothesis links an observation, a single change and an expected result: "Mobile visitors leave before reaching the booking form, so moving it above the fold should raise submissions." If the sentence lacks an observation or a measurable outcome, it is a wish.

Prioritize with a simple score

Rate each hypothesis from one to five on impact, confidence in the evidence and ease of implementation (five meaning easiest), then add the scores. This is a rule of thumb of our own, not an industry standard or a Google method. Its job is to make the order of work explicit instead of favoring the most exciting idea.

The scores above only illustrate the method; use your own evidence when you rate.

Running tests without hurting search

Google's website testing guidance (Opens in a new tab) sets a few ground rules. Do not show Googlebot a different set of URLs from the ones humans see, because that is cloaking whether or not you are testing. Where variants have separate URLs, point them at the original with rel=canonical, which the page prefers over noindex. Use a temporary 302 redirect rather than a permanent 301. Run an experiment only as long as needed to reach a reliable conclusion, then remove the test elements, since overly long experiments can look like an attempt to mislead search engines. For a fuller walkthrough, split testing without cloaking goes through the setup. Google's guidance can change, so read the current page before you start.

Low-traffic sites

If you do not have enough visitors for a clean split test, use sequential comparison: make the change, measure the same event for a comparable period with a similar traffic mix, and write down anything that could distort the result, such as a promotion or a holiday. Report the outcome as a signal rather than proof.

Decision rule: fix or test

If something is plainly broken, such as a dead button or a form that rejects valid input, fix it now. If you genuinely do not know which option is better, such as two headlines, test it. Either way, record the change, the reason and the result in an experiment log.

Example scenario: a B2B software demo page

Example scenario: A project-management software company wants more demo requests. The demo page has a long form, and recordings show many visitors starting it and leaving at the phone number field. The team's hypothesis is that a required phone number feels intrusive for a first contact. They make the field optional, keep everything else unchanged and compare completed demo requests over the following weeks against the same weeks of a comparable earlier period, noting that no campaign ran in either. The point is not the exact lift, which depends on the site, but the discipline: one observation, one change, one event, one written conclusion.

Notice what the team did not do. It did not redesign the whole page, rewrite the headline and swap the hero image in the same week. Any of those might help, but changing them together would have made the result impossible to interpret.

SEO and CRO: where they help and where they clash

The same page often serves both disciplines, and the two can pull in opposite directions if nobody coordinates them.

How search supports conversion

When search brings people whose intent matches the page, conversion rates tend to follow, because the visitor finds what they came for. Poorly matched queries bring traffic that rarely converts. So read your conversion rate together with the queries and page types behind it. To see how several channels share credit, SEO-assisted conversions and attribution models is a useful companion.

How conversion work can hurt search

Stripping a page down to buttons, deleting explanatory text or hiding important content can reduce what the page offers to searchers. Google's helpful content guidance (Opens in a new tab) asks whether content provides original information, research or analysis and was made primarily to help people. A page built to convert should still pass that test, because genuinely helpful pages tend to serve both goals.

A 30-day starter plan

You do not need a large team. A modest, repeatable monthly rhythm beats an ambitious plan that stalls.

Week by week

1. Week one: write the main goal and primary event, test the event, record the current rate.
2. Week two: map the funnel, find the largest drop, review recordings and customer questions for your three busiest pages.
3. Week three: write three hypotheses, score them, and implement the top one as a single change.
4. Week four: measure, log the outcome and update the hypothesis list.

What to report

Keep the report short: main event count and rate, before and after, any confounders and the next step. Do not present small-sample swings as certainty. A change that did nothing is still a result worth logging.

Common mistakes

• Redesigning before measuring.
• Changing many things at once.
• Judging your rate against someone else's benchmark.
• Treating micro conversions as the goal.
• Leaving test scripts or losing variants live after the test ends.

Guardrails: honesty and privacy

Pressure to hit a number can push teams toward tactics that raise conversions briefly and erode trust.

No fake urgency

Countdown timers that reset, invented stock warnings and "people are viewing this now" messages that are not real may lift orders today and cost you refunds, complaints and reviews tomorrow. Show scarcity or deadlines only when they are true.

Consent and data minimization

Behavior tools and analytics must respect your consent set-up. Rules differ by country, so check with a qualified adviser. Mask form fields in recordings, avoid collecting personal data you do not need and keep access limited.

Dark patterns

Hiding the cancel route, pre-selecting paid extras or revealing the full price only at the last step can bump conversions in the short run, but the cost returns as chargebacks, bad reviews and lost repeat business. Good CRO helps the right person reach the right decision with less effort.

Frequently Asked Questions

What is CRO in simple terms?

It is the process of measuring and improving the share of visitors who complete a valuable action on your site. It works on the visitors you already have rather than on bringing in more.

How do I calculate conversion rate?

Divide the number of completed target actions by the number of sessions or users who could have completed them, then express it as a percentage. Pick one denominator and use it consistently.

What is a good conversion rate?

There is no universal answer. It varies by sector, product, traffic source and page intent. The most reliable comparison is your own site's past performance in the same segment.

Which event should a small site measure first?

The action closest to revenue or real demand that happens often enough to measure: orders for shops, calls or forms for local services, quote or demo requests for B2B. If it is too rare, track a more frequent upstream event alongside it.

Can A/B testing hurt SEO?

Done as Google's guidance describes, it should not. The page advises against cloaking, recommends rel=canonical for variant URLs, prefers temporary redirects and asks you to end tests promptly. Guidance can change, so check the current page.

Is CRO worth it with low traffic?

Yes, with a different method. Fix obvious problems, read recordings and customer questions, and compare before and after periods instead of running formal split tests. Treat the results as signals.

Related guides: Lead and Contact Form Design for Conversion, How to Write CTA Button Text, Social Proof and Reviews on Landing Pages, GA4 Key Events: What to Mark and What to Skip.

Conclusion

The first steps of CRO are plain: define the conversion, measure one event correctly, find where people drop out, learn why through recordings and feedback, then score hypotheses and test them one at a time. Keep SEO and CRO working for the same visitor, follow Google's testing rules and avoid tactics that trade trust for a short-term bump. No method guarantees a result, but disciplined measurement and a simple log make every month's work build on the last.

Once your measurement is in place and you want fresh visitors to test against, you can look at ad spaces on Akua Adforce; scope and terms are described on the product page.