Ecommerce Conversion Rate Optimization: A Practical Guide

What CRO is (and what it isn't)

Conversion rate optimization means improving the share of visitors who do what you want (usually: place an order) by understanding and removing what stops them.

In practice, it's a loop:

1. Find the problems: where do visitors drop off, and why?
2. Turn problems into ideas and decide which matter most.
3. Fix or test: fix the obvious, test the uncertain.
4. Learn and repeat.

What CRO is not:

• Not a list of "best practices". "Add trust badges" or "use urgency" may help one shop and hurt another. Your customers, products and prices are specific.
• Not only A/B testing. Testing is one tool in CRO, and for many small shops not the first one.
• Not a redesign. Big redesigns often reset what was working. CRO prefers targeted changes based on evidence.
• Not tricks that pressure people. Fake countdown timers and hidden costs may lift orders for a week and cost you returns, reviews and trust. In the EU, some are also illegal under consumer protection rules.
• Not only about conversion rate. A change that brings more orders with lower value, or more returns, isn't a win.

CRO is also the cheapest growth lever many shops ignore. If you spend money on ads to bring visitors in, every improvement in how many of them buy makes all of that spend more effective.

Know your numbers first: conversion rate, AOV, revenue per visitor

Before you change anything, write down where you are.

Conversion rate (CR) = orders ÷ sessions (or ÷ visitors, depending on your analytics). 400 orders from 20,000 sessions = 2.0%.

Average order value (AOV) = revenue ÷ orders. €24,000 from 400 orders = €60.

Revenue per visitor (RPV) = revenue ÷ sessions, or conversion rate × average order value. €24,000 ÷ 20,000 = €1.20 per session.

RPV is the most useful single number, because it captures both: a change that raises conversion but lowers order value shows up honestly.

A worked example for a hypothetical shop with 20,000 sessions a month:

A 10% improvement in conversion is worth €2,400 a month here, without spending a cent more on traffic.

Break the numbers down

A single site-wide conversion rate hides most of what's going on. Break it down by:

• Device: mobile usually converts much lower than desktop. Contentsquare's 2026 benchmark puts retail conversion at about 3.7% on desktop and 2% on mobile (as cited by Shopify).
• Traffic source: email and returning visitors usually convert far better than paid social.
• New vs returning visitors.
• Landing page type: homepage, category, product, blog.
• Funnel step: product page → add to cart → cart → checkout → order.

The funnel view is the most useful starting point. If 8% of product page visitors add to cart but only 30% of cart visitors start checkout, the cart is where to look first.

Take a snapshot of the last 8 full weeks (not including big sales) and save it. That's your baseline.

Check your tracking before you trust any number

Many CRO projects go wrong at this step: the numbers are off, and nobody notices. Before you use your baseline, check:

• Orders match. Compare the number of orders in your analytics with your shop's back office for the same week. A gap of a few percent is normal (consent, ad blockers, canceled orders). A gap of 30% means something is broken.
• Revenue is counted once, in one currency, with or without tax and shipping, consistently.
• Internal traffic is excluded. Your own team, your agency and your warehouse staff checking orders can make up a surprising share of visits on a small shop.
• Bots are filtered. A sudden jump in direct traffic with zero conversions is often a bot.
• Consent is understood. In the EU, analytics tools that need consent only see visitors who accept. Your conversion rate in those tools describes those visitors, not everyone.
• Funnel steps fire on every device. Add a product to the cart on your phone and check that it shows up.

Half a day spent here saves you from optimizing the wrong thing.

What "good" looks like (benchmarks and why to be careful)

Every shop owner wants to know: "Is my conversion rate good?" Here's what recent published benchmarks say:

So "average" is somewhere between 1.4% and 2.7%, depending on who measures, how, and which shops are included. Industry differences are larger than that whole range.

Why to be careful with benchmarks:

• Different definitions. Sessions or users? Including or excluding bots? With or without visitors who declined cookies?
• Different shops. A €15 pet-food reorder and a €1,500 sofa are different businesses.
• Different traffic mixes. A shop with lots of paid social traffic will convert lower than one with mostly email and direct traffic, even if its site is better.
• Averages hide the spread. Half of all shops are below average by definition.

The useful comparison is with yourself: this quarter vs last quarter, this month vs the same month last year, mobile vs desktop on your own shop. If your mobile conversion is a third of your desktop conversion, that's a stronger signal than any industry average.

Step 1: Find the problems

Good CRO ideas come from evidence. Use four kinds of research, from "what" to "why".

Analytics: where do people drop off?

Your analytics tool (GA4, your shop platform's reports, or another tool) tells you what happens:

• Funnel drop-off: product page → cart → checkout → order. Where's the biggest leak?
• Device gaps: is mobile much worse than desktop at one specific step?
• Landing pages with high traffic and low conversion.
• Site search: what do people search for, and how often do they get zero results?
• Page speed by device: are mobile product pages slow?

Write down the three biggest leaks. That's where you'll look closer.

Heatmaps: what do people click and see?

Heatmaps show where visitors click, how far they scroll and what they ignore:

• Do mobile visitors reach the add-to-cart button?
• Do people click things that aren't links (images, labels)?
• Which menu items get used, and which never?
• How many people reach the reviews, size guide or delivery information?

Segment by device. A desktop and mobile heatmap of the same page often tell opposite stories. Our guide to how to analyze heatmaps shows what to look for on each page type.

Session recordings: what goes wrong?

Recordings show real visits, step by step. Don't watch at random; start with a question and a filter:

• Visits that added to cart but didn't order.
• Visits with rage clicks (repeated clicks on the same spot) or dead clicks (clicks on things that do nothing).
• Mobile visits to a product page that lasted over a minute without an add-to-cart.

Watch 10–20 per question and note patterns. Three people struggling with the same filter is a finding. See how to analyze session recordings.

Surveys: why don't they buy?

On-site surveys give you the visitor's own words. One well-placed question can be worth more than a month of analytics:

• On the product page: "Is anything stopping you from buying today?"
• In the cart: "What's holding you back from completing your order?"
• On exit (desktop): "What were you looking for that you didn't find?"
• After purchase: "What almost stopped you from ordering?"

Keep it to one question, open text, and aim for 50–100 answers before you draw conclusions. Then group the answers into themes and count them. See our list of on-site survey questions for online shops.

Other sources you already have

• Customer service emails and chats: the same questions over and over are content your product pages are missing.
• Product reviews, especially 3-star ones: they tell you what didn't match expectations.
• Return reasons: "too small", "different color than in the photo".
• Your own experience: buy something from your shop on your phone, on mobile data, as a new customer. Write down every annoyance.

A note on privacy: heatmaps, recordings and surveys collect data about visitors. In the EU, they usually need consent, and you should mask personal data and keep checkout out of recordings. See our guide to GDPR for heatmaps, session recordings and A/B testing.

Step 2: Turn findings into ideas and rank them

Research gives you findings: "31% of cart survey answers mention shipping costs", "mobile visitors don't scroll to the add-to-cart button". Turn each finding into an idea with a short hypothesis:

Because we saw [finding], we believe that [change] will [effect].

For example: "Because 31% of cart survey answers mention shipping costs, we believe that showing shipping costs on the product page will increase orders."

Our A/B test hypothesis template has a fuller version and 20 examples.

Then rank the ideas. A scoring framework helps you avoid testing the loudest person's favorite:

• ICE: Impact, Confidence, Ease (1–10 each). Quick, subjective.
• PIE: Potential, Importance, Ease. Leans on page traffic.
• PXL: yes/no questions about the change and the evidence. Slower, more consistent.

Add one question none of them asks: can this be tested with our traffic? See how to prioritize A/B test ideas for the same 10 ideas scored all three ways.

Step 3: Test or just fix?

Not every idea needs an A/B test. Use this rule:

• Just fix it if it's a bug, an obvious usability problem, a legal requirement, or something you'd keep even if a test showed "no difference". Examples: a broken filter, missing delivery information, a slow image, a confusing error message.
• Test it if you're genuinely unsure whether it will help, if it could hurt (a new layout, removing information, a different offer), or if it's expensive to build and maintain.
• Research more if you don't have evidence yet.

And then check whether a test is even possible. An A/B test needs enough visitors to tell a real effect apart from chance. On a page converting at 2%:

(95% confidence, 80% power, one variant. Use our sample size calculator with your own numbers.)

If you can only detect +30% or more, test bold changes and earlier steps (add-to-cart instead of orders), or fix and monitor. When you do test, our A/B testing guide for ecommerce walks through the whole process.

When you fix without testing, still measure. Note the date of the change, and compare the same metric for the same number of full weeks before and after, avoiding sales and campaigns. It's weaker evidence than a test, but much better than nothing.

Step 4: Learn and repeat

CRO is a loop, not a project. After each change or test:

• Write it down: what you changed, why, what happened, what you learned. One paragraph is enough.
• Update your backlog: a lost test might kill three related ideas, or suggest a better one.
• Go back to research: every month, look at your funnel, a few heatmaps, new survey answers and a handful of recordings.

Over a year, the learnings matter more than any single win. "Our customers care more about delivery speed than price" is worth more than "the green button won".

Expect many tests not to win. At companies like Booking.com, Google and Netflix, roughly 10% of ideas improve the metric they target; at Microsoft overall, about a third (Kohavi, Deng and Vermeer, KDD 2022). Shops that fix real, researched problems tend to do better than that at first, but a loss is a normal result, not a failure.

What to do at your traffic level

Your traffic decides which CRO methods make sense. These are rough bands for the main page or template you'd work on, assuming a conversion rate around 2–3%:

If you're in the first two rows, our guide on A/B testing with low traffic explains the options in detail.

Mobile deserves its own plan

On most shops, the majority of visits come from phones, and phones convert worse. Some of that gap is natural: people browse on the bus and buy later on a laptop. But much of it is fixable. When you review your shop, do it on a phone first:

• The first screen of the product page: can you see the product, the price, the delivery promise and the add-to-cart button without scrolling? If the button is three screens down, a sticky button or a shorter intro may help.
• Tap targets: size and color selectors that are too small, links too close together, pop-ups that are hard to close.
• Forms: the right keyboard for each field (numbers for phone and ZIP code), autofill that works, error messages that say what's wrong.
• Speed on mobile data, not on your office Wi-Fi.
• Pop-ups: a newsletter pop-up that covers the whole screen on the first page view costs more on mobile than on desktop.

Look at your mobile funnel separately. If mobile visitors add to cart at a similar rate to desktop but rarely finish checkout, the problem is in the cart or checkout. If they rarely add to cart, it's the product page.

The 10 most common problems on small shops

These come up again and again. Check your shop for each one before you plan anything clever.

1. Unclear delivery costs and times. Shoppers want to know what it costs and when it arrives before they add to cart. In Baymard Institute's survey of US shoppers who abandoned a purchase, 40% named extra costs (shipping, tax, fees) as a reason, and 20% slow delivery (Baymard Institute, updated October 2026). Show delivery cost and expected date on the product page.

2. Slow mobile pages. Large images, too many apps and scripts, and fonts that block the page. Test your top product page on a mid-range phone on mobile data. If the main image takes more than a couple of seconds, start there.

3. Missing product information. Sizes, materials, dimensions, compatibility, care instructions. If customer service gets the same question twice a week, the answer belongs on the product page.

4. Weak product photos. No zoom, no "in use" photos, no scale reference, colors that don't match reality. Return reasons will tell you.

5. Clumsy mobile filters and search. Filters that close after every tap, search that fails on typos, no results for obvious terms. Check your zero-result searches monthly.

6. Forced account creation. 18% of the shoppers in Baymard's survey named "the site wanted me to create an account" as a reason to abandon (Baymard Institute). Offer guest checkout, visibly.

7. Trust gaps for new visitors. New visitors from social ads don't know you. Reviews, a clear returns policy, real contact details and familiar payment methods matter more to them than to returning customers. In Baymard's survey, 19% didn't trust the site with their card details.

8. Surprise costs or steps late in checkout. Anything that appears for the first time in checkout (fees, delivery options, a required phone number) creates doubt. Baymard found 12% abandoned because they couldn't see the total cost up front.

9. A cluttered cart. A prominent discount code field sends people off to search for codes. Cross-sells that push the checkout button below the fold on mobile. Keep the cart focused on finishing the order.

10. Unclear returns policy. Especially for fashion and furniture, "Can I send it back?" is part of the buying decision. 13% of Baymard's respondents named an unsatisfactory returns policy. Say it near the price, in one sentence, with a link to the details.

Most of these are fixes, not tests. Fix them first, then test what's genuinely uncertain. For more ideas sorted by page and traffic needed, see ecommerce A/B test ideas.

A 90-day CRO plan

This plan assumes one person spending about 4–6 hours a week. Adjust to your situation.

Weeks 1–2: Baseline and quick fixes

• Week 1: Write down your baseline for the last 8 full weeks: conversion rate, AOV and RPV, by device and traffic source, and the funnel (product page → cart → checkout → order). Check that orders are tracked correctly.
• Week 2: Buy something from your own shop on your phone, as a new customer, on mobile data. Do the same on a competitor's shop. Fix everything that's clearly broken. Check the 10 common problems above.

Weeks 3–6: Research

• Week 3: Install or check heatmaps and recordings on your homepage, top category pages, product page template and cart. Set up masking of personal data and make sure checkout isn't recorded. Make sure your cookie banner and the tools work together.
• Week 4: Launch one survey question on the product page or cart. Read customer service emails and 3-star reviews from the last three months; note recurring themes.
• Week 5: Analyze heatmaps per page type, segmented by mobile and desktop. Watch 20 recordings of visits that added to cart but didn't order. Note patterns.
• Week 6: Group survey answers into themes and count them. Write a one-page summary: the 5–10 biggest problems, each with evidence.

Weeks 7–8: Ideas and priorities

• Week 7: Turn each problem into one or more ideas with a short hypothesis. Score them with ICE, PIE or PXL.
• Week 8: For each top idea, decide: fix, test, or research more. For test candidates, check the sample size and duration. Pick one test and two or three fixes.

Weeks 9–12: Fix, test, measure

• Week 9: Ship the fixes. Note the dates. Build the test variant and QA it on real devices.
• Week 10: Launch the test (a dry run first if your tool supports it). Check that both versions get visitors and that orders are recorded. Then leave it alone.
• Weeks 11–12: Keep the test running for its planned full weeks. Meanwhile: compare your fixed pages before/after, run a second survey question, and prepare the next ideas.

Week 13: Review

• Read the test result after its planned duration. Write a one-paragraph report for every fix and test: what, why, result, learning.
• Compare your conversion rate, AOV and RPV with the week-1 baseline (same number of weeks, no big sales).
• Update the backlog and plan the next 90 days.

Tools you need (and don't need)

You need fewer tools than vendors suggest.

You need:

• Analytics that tracks orders correctly: your shop platform's reports, GA4 or a privacy-friendly alternative.
• Heatmaps and session recordings. Free options exist (Microsoft Clarity is the best known); paid tools add things like longer data retention, revenue per click, surveys or testing in the same place.
• An on-site survey tool, or a survey feature in your heatmap tool.
• A spreadsheet or backlog tool for ideas, scores and results.

You need later (once traffic allows):

• An A/B testing tool. Client-side tools are quickest to start with; server-side or platform-native testing is better for prices and big template changes. See our A/B testing guide for the differences.

You probably don't need:

• Several overlapping tools that each add a script to every page. Every script costs loading time.
• AI "optimization" tools that promise automatic wins without telling you what they changed or how they measured it.
• Enterprise personalization suites before you've fixed the basics.

Platform notes: Shopify, WooCommerce and Magento each have their own quirks for tracking orders, checkout and caching. See our guides to CRO on Shopify, WooCommerce and Magento / Adobe Commerce.

Varly brings heatmaps, recordings, surveys, an ideas backlog with prioritization, and A/B tests together in one script, with privacy defaults such as masking on the visitor's device and checkout never tracked by default. It's not the right choice for every shop: it has no mobile-app SDK, it's a young product, and it's currently in private beta, so if you need a tool today, start with what's available to you.

FAQ

What is a good conversion rate for an online shop? Published averages for 2026 range from about 1.4% to 2.7%, and industries differ far more than that. The better question is whether your rate is improving, and how mobile compares with desktop on your own shop.

How long does CRO take to show results? Fixes for clear problems can show up in your numbers within weeks. A testing program takes months to pay off, because each test needs several weeks and many won't win.

Should I focus on conversion rate or average order value? On revenue per visitor, which combines both. A change that lifts conversion but lowers order value (or raises returns) isn't a win.

Do I need A/B testing to do CRO? No. Research and fixing obvious problems are CRO too, and for small shops often the most valuable part. A/B testing becomes useful when you have enough traffic and genuinely uncertain ideas.

What should I fix first? The biggest leak in your funnel, on mobile, with the clearest evidence. For most small shops, that's delivery information, mobile product pages or the cart.

How much does CRO cost? For a small shop doing it in-house, mostly time: a few hours a week, plus tools. Many tools have free plans to start with.

---

Start with research, not guesses. Varly combines heatmaps, recordings, surveys and A/B testing in one script, links every idea to the evidence behind it, and tells you whether your traffic is enough to test it. Varly is in private beta, with planned pricing at launch of $99 per website per month. Join the waitlist.