What Is Conversion Rate Optimization (CRO)? A Starter Guide

Conversion rate optimization, defined

A conversion is a visitor completing an action that matters to your business. Google Analytics calls these key events and defines one as “an event that measures an action that’s particularly important to the success of your business” (Google Analytics Help, About key events). For a store, the main conversion is a purchase. For a service business, it might be a contact form submission or a booked call.

Conversion rate is the share of visitors (or sessions) who convert:

conversion rate = conversions ÷ visitors × 100

Example (made-up numbers):
  4,000 visitors this month, 60 orders
  60 ÷ 4,000 × 100 = 1.5%

Conversion rate optimization is the systematic work of raising that number by removing the reasons people don’t convert. Because it works on visitors you already have, it often costs less than buying more traffic — and any improvement makes every future visitor, from every channel, a little more valuable.

What CRO is not

1. Not copying someone else’s winning test. A change that helped another site worked for their visitors, offer and page. It’s a hypothesis for you, not a result.
2. Not tricking people. Fake scarcity, hidden fees revealed at checkout, or pre-ticked boxes can lift a number for a while and cost you trust, refunds and repeat customers.
3. Not only A/B testing. Testing is one step. Most of the value comes earlier: finding where the problem is and why.
4. Not a one-time project. Products, prices, traffic sources and seasons change. The loop keeps running.

Macro- and micro-conversions

The purchase or lead is the macro-conversion. On a small site, it happens too rarely to tell you much week to week. That’s where micro-conversions help: smaller actions that tend to come before the main one, such as viewing a product page, adding to cart, reaching the shipping page, spending a few minutes on a pricing page, or coming back a second time.

Micro-conversions let you see movement sooner and pinpoint the step where people stall. For more on picking good ones, see Micro-Conversions: The Small Signals That Predict a Sale.

The CRO loop

Every CRO program, from a one-person store to a large team, runs some version of this cycle:

1. Measure — find where in the path visitors drop off.
2. Diagnose — look at behavior on that step to understand why.
3. Hypothesize — write down what you’ll change, why, and what you expect to happen.
4. Change — make one focused change.
5. Test or compare — A/B test it if you have the traffic; otherwise compare before and after carefully.
6. Record and repeat — keep a log of what you tried and what happened, then pick the next problem.

The rest of this guide walks through each step.

Step 1: Measure — find the leak

Start by mapping the path to your main conversion as a sequence of steps and counting how many visitors reach each one. GA4’s funnel exploration report, for example, lets you “visualize the steps your users take to complete a task” (Google Analytics Help, Funnel exploration). You can do the same with any analytics tool that reports page views per URL.

In this made-up example, the big leak is between product page and cart: only 15% of product viewers add something. That is where to look first — not the button color on the home page. Also split the funnel by device and traffic source. A leak that only appears on mobile, or only for visitors from one ad campaign, points to a very different cause than a leak everywhere.

Prioritize by size and closeness to money. A leak late in the path (checkout, form submission) affects people who were very close to converting. A leak early in the path affects more people but each is less likely to convert anyway. A simple scoring of “how many people, how close to the conversion, how easy to change” is enough for a small site.

Step 2: Diagnose — understand why

Numbers tell you where. To understand why, look at what visitors actually do on that step.

1. Heatmaps. Scroll depth shows whether people reach your shipping info, reviews or “Add to cart” button. Click maps show what they try to tap — including things that aren’t links. See Website Heatmaps Explained.
2. Individual visitor paths. Reading a handful of journeys that ended without a conversion often reveals patterns: back and forth between product and shipping pages, repeated visits to the returns policy, leaving from the size guide. See How to Read a Single Visitor’s Journey Through Your Site.
3. Form field drop-off. If the leak is a form, a field-by-field report shows which question makes people quit. See Form Abandonment: How to Find the Field Where People Give Up.
4. Customer questions. Support emails and chat logs are free qualitative research. If people keep asking about delivery times, your pages aren’t answering it.
5. Watch someone use the site. Ask a friend who matches your customer to buy something while thinking aloud. A few sessions often surface the obvious problems.

Published research helps generate ideas too. In Baymard Institute’s survey of US online shoppers, the most common reasons for abandoning during checkout (setting aside “just browsing”) were extra costs such as shipping, tax and fees (40%), slow delivery (20%), not trusting the site with card information (19%), and being required to create an account (18%) (Baymard Institute, cart abandonment rate statistics, last updated September 22, 2025). These are candidates to check on your own site, not conclusions. Our cart abandonment guide goes through the on-site fixes.

Step 3: Write a hypothesis

A hypothesis turns a hunch into something you can check. It has three parts: what you observed, what you’ll change, and what you expect to happen. Writing it down keeps you honest when the results come in.

Because   [observation from data],
we will   [one specific change],
and expect [measurable result] on [metric].

Because only 30% of mobile visitors scroll far enough to see the shipping
information on product pages, and support often gets shipping questions,
we will move a one-line shipping summary next to the price,
and expect a higher add-to-cart rate on mobile product pages.

Good hypotheses share a few traits:

1. Grounded in something you saw, not “competitors do it.”
2. One change, so you know what caused any difference.
3. One primary metric chosen in advance, ideally the step right after the change (add-to-cart for a product page change).
4. A guardrail metric, such as completed orders or refund requests, so a “win” on clicks doesn’t hide a loss further down.

Step 4: Make the change

Common types of changes, roughly from cheapest to most involved:

Forms are a good place to start because the fixes are well understood. Web.dev’s sign-up form guidance puts it plainly: “On sign-up, ask for as little as possible,” and notes that “small changes can make a big difference to completion rates” (web.dev, Sign-up form best practices). Our Form UX Best Practices covers validation, progress bars and autofill.

Page stability counts too. Content that moves as it loads leads to mis-taps and frustration; web.dev recommends a Cumulative Layout Shift of 0.1 or less (web.dev, CLS). Check this whenever you add banners or widgets as part of an experiment.

For paid campaigns, relevance is often the biggest lever: a visitor who clicked an ad for one product shouldn’t land on a generic home page. See Campaign Landing Pages for Small Stores and UTM Parameters Explained.

Step 5: Test — or compare carefully

In an A/B test, visitors are split at random between the current version (A) and the changed version (B) at the same time, and you compare conversion rates. Because both groups see the site in the same period, seasonality and campaign changes affect them equally. That’s why a test gives a more trustworthy answer than “sales went up after we changed it.”

Three rules prevent most testing mistakes:

1. Decide the sample size or duration before you start, and don’t stop the moment B pulls ahead. Early leads often disappear.
2. Run full weeks, so weekday and weekend visitors are both represented.
3. Look at significance, not just the difference. A p-value tells you how surprising the gap would be if there were no real difference. The details are in A/B Testing Your Forms: Sample Size, p-Values and When to Stop.

If you test with separate URLs, follow Google’s guidance so the test doesn’t affect search: don’t show Googlebot different content than people see, use rel="canonical" on variant URLs pointing to the original, use 302 rather than 301 redirects, and “run the experiment only as long as necessary” (Google Search Central, Minimize A/B testing impact in Google Search).

When you don’t have enough traffic to test

Many small sites simply don’t get enough conversions for an A/B test to reach a clear answer in a reasonable time — especially on the purchase itself. That doesn’t mean CRO is off the table. It means adjusting the approach:

1. Test where the volume is. A step with more traffic, such as a sign-up or contact form, or a micro-conversion like add-to-cart, reaches a result much faster than the final order.
2. Make bigger changes. Small tweaks produce small differences that need huge samples to detect. Fixing a clear problem (a broken mobile layout, a confusing form) produces differences you can see.
3. Fix obvious problems without testing. A missing shipping policy, an error message that doesn’t say what’s wrong, or a form that rejects valid phone numbers is not a hypothesis — it’s a bug. Fix it.
4. Compare before and after fairly. Use the same number of full weeks before and after, avoid periods with sales or holidays, and note anything else that changed (ads, prices, stock).
5. Keep a change log. Date, page, what changed and why. Months later, it’s the only way to make sense of a trend line.

Prioritizing ideas when you have too many

After a round of diagnosis you’ll usually have more ideas than time. A quick score keeps the choice honest. Rate each idea from 1 to 3 on three questions and add them up:

Work from the highest total down. The score isn’t science, but it stops the loudest opinion in the room — or the most fun idea — from jumping the queue.

A worked example (hypothetical)

To show how the steps connect, here is an illustrative example. The store and numbers are invented.

1. Measure. A small tea shop sees that most visitors reach product pages, but few reach the cart, and the gap is wider on mobile than on desktop.
2. Diagnose. The scroll heatmap of mobile product pages shows most visitors never reach the section with shipping costs and delivery times. Several visitor journeys show people opening the shipping policy page and leaving from it. Support has received repeated questions about delivery to other states.
3. Hypothesize. “Because mobile visitors don’t see shipping details and go looking for them, we will add a one-line shipping summary under the price, and expect a higher add-to-cart rate on mobile product pages. Guardrail: completed orders.”
4. Change. The shop adds the line and checks that it doesn’t push the “Add to cart” button out of view on small phones.
5. Compare. Traffic is too low for a quick A/B test on add-to-cart, so the shop compares four full weeks before and after, with no sales in either period, and notes that ad spend didn’t change.
6. Record. Whatever the result, it goes in the log with the dates, so the next idea builds on this one.

A starter CRO checklist for a small site

1. Define your main conversion and two or three micro-conversions, and write down how conversion rate is calculated.
2. Map the path to the conversion and count visitors at each step, split by device and traffic source.
3. Pick the single biggest leak close to the conversion.
4. Check heatmaps and a dozen visitor journeys on that step; read recent customer questions.
5. Write one hypothesis with a primary metric and a guardrail metric.
6. Fix obvious bugs right away; test the rest if traffic allows.
7. Record the result, whether it won, lost or was inconclusive.
8. Pick the next leak and repeat — monthly is a reasonable rhythm for a small team.

Common CRO mistakes

Running the loop in marketingLABO

marketingLABO puts the measurement and the changes for this loop in one dashboard, installed with a single JavaScript tag in the <head> of your pages. Here is how each step maps to what it does:

The Free plan covers analytics, heatmaps and visitor history for one site, plus one form and one banner. See pricing for what Light and Standard add, or create a free account to start measuring.

FAQ

What is a good conversion rate?

There is no single benchmark worth chasing. Rates vary widely by industry, price, traffic source and how you count. The useful comparison is your own site against itself over time, and one segment against another (mobile vs. desktop, ads vs. organic).

Is CRO the same as UX design?

They overlap. UX design aims to make the site easy and pleasant to use; CRO focuses that work on the steps that lead to a conversion and measures the effect. Most good CRO changes are UX improvements.

How long until I see results?

Fixing clear bugs can show up within weeks. Tests take as long as your traffic requires to reach a decision. Plan in months, not days, and judge the program by the log of lessons learned as much as by any single win.

Does CRO conflict with SEO?

Not if you follow Google’s testing guidance and avoid intrusive overlays. Clearer pages, faster loading and stable layouts tend to help both.

Do I need special tools to start?

No. Page-view counts per URL, a way to see where people scroll and click, and a spreadsheet for your hypotheses and change log are enough for the first few rounds. Testing tools become useful once you know which step to work on and have enough traffic there to reach a result.

Sources

1. Google Analytics Help — About key events
2. Google Analytics Help — Funnel exploration
3. Baymard Institute — Cart abandonment rate statistics
4. web.dev — Sign-up form best practices
5. web.dev — Cumulative Layout Shift (CLS)
6. Google Search Central — Minimize A/B testing impact in Google Search