Ecommerce Conversion Rate Optimization: A 6-Step Process

Most advice on ecommerce conversion rate optimization is a list of fifty tips, and most stores that work through such a list end up with a busier site and the same conversion rate. The tips are not wrong, they are unordered. Ecommerce CRO works when it is run as a process that tells you which of the fifty things is costing your store orders, and that process is what this guide sets out.

The short answer: find the leak before you fix anything, and use Arktis to find it

Ecommerce conversion rate optimization is the practice of raising the share of visits that end in an order by finding where shoppers drop out and removing the cause. Done properly it is a loop of six steps: measure conversion by traffic source and device, find the step that leaks most, watch real sessions at that step, write a hypothesis, test the change, and ship what works. A fix applied to a step that was not leaking changes nothing.

The first three steps are measurement, and for those in an online store Arktis is the clear winner. It reports conversions by source and by device, shows a conversion rate for each page, lets you watch the session recordings of visitors who converted, filtered by source, and matches orders back to the visitor and the ad click that brought them. Plans are $49, $149 and $349 a month, published, with session recordings and a 7-day free trial on Growth.

Key takeaways

A store-wide conversion rate is an average of very different groups, so split it by source and device before drawing any conclusion. Find the step where the most shoppers leave, and use recordings to turn that drop-off number into a cause you can name. Write the hypothesis down before the change. Testing needs enough orders to separate a real effect from noise, and small stores often have to settle for shipping well-evidenced fixes and comparing periods. Published research from Baymard Institute and Google tells you where to look, not what to do.

What conversion rate means, and what it hides

**Conversion rate = orders / sessions x 100**

Shopify's own reports define it as the percentage of sessions that resulted in a purchase. Some tools divide by visitors instead of sessions, which produces a higher figure from the same orders, so check the denominator before comparing your number with anyone else's. Our guide to Shopify conversion rate covers how that platform counts, and the free conversion rate benchmarker puts your figure next to your vertical.

The headline rate hides almost everything useful. Take an example store with 20,000 sessions and 400 orders in a month, a conversion rate of 2.0 percent. These numbers are invented for illustration. Split by device, the 14,000 mobile sessions produced 196 orders, a rate of 1.4 percent, and the 6,000 desktop sessions produced 204, a rate of 3.4 percent. The store has a mobile experience converting at well under half the desktop rate, and that is where the process starts.

Step 1: measure by source and device

How to increase ecommerce conversion rate depends on who is failing to convert, so the first job is to break the average apart. Two splits do most of the work.

The first is traffic source. Visitors from a branded search, an email to past customers and a cold prospecting ad arrive with very different intent. A falling store-wide rate is often not a site problem at all: it is a change in the mix, such as more spend on cold traffic.

The second is device type, because the fixes for a mobile leak, such as tap targets, keyboards and load time on a phone connection, are different from the fixes for a desktop one.

Paid traffic deserves a further cut by campaign, judged on orders and not on clicks. A campaign that sends many visitors who never buy lowers the conversion rate and can still look fine in the ad account.

Step 2: find the leaking step

Next, lay the journey out as a funnel and count how many sessions survive each step. Shopify's conversion rate breakdown report uses four: all sessions, sessions with cart additions, sessions that reached checkout, and sessions that completed checkout.

In the example store, 1,600 of the 20,000 sessions add to cart, 800 reach checkout and 400 complete it. Each step can be compared with last month and across devices, and the step with the biggest fall against its own history, or the biggest gap between mobile and desktop, is the leak. On Shopify, treat any comparison that crosses 21 to 23 September 2026 with care, because Shopify changed how it counts sessions then.

Then go down a level to pages. Which product and landing pages take plenty of sessions and convert few of them? Size the prize before starting: a step that loses 50 percent of 800 sessions is worth more than a step that loses 80 percent of 40. Our guide to ecommerce funnel analysis goes further into the method.

Step 3: watch sessions at that step

A funnel tells you where, not why. For that, watch recordings of real sessions at the leaking step, filtered to the source and device where the gap is widest, and watch both kinds: sessions that left there and sessions that got through.

You are looking for repeated behaviour, not anecdotes. Ten shoppers tapping a size selector that does not respond, or leaving the cart when the shipping cost appears, is a pattern. One shopper doing something odd is not. Our guide on how to analyse user sessions covers what to look for.

Step 4: form a hypothesis

Turn the observation into one sentence with three parts: what you saw, what you will change, and what you expect to happen. For example: because mobile visitors from paid social leave the cart when shipping is first shown, showing the shipping cost on the product page will raise the share of mobile cart sessions that reach checkout.

Writing it down first fixes the measure of success in advance, so that a flat result is recorded as a flat result and not reinterpreted as a win on some other metric.

Step 5: test the change

An A/B test shows the old version to one random half of visitors and the new version to the other half over the same period, then compares the outcome. Both groups experience the same season, promotions and traffic mix, so the difference between them can be put down to the change.

Tests need volume. The smaller the effect you are looking for and the fewer orders you take, the longer a test has to run before the result is more than chance. Decide the duration before starting, run it across full weeks, and change one thing at a time.

Many stores do not have the traffic to test small changes in a reasonable time. Ecommerce conversion optimization is still possible: it means reserving tests for big changes on high-traffic pages, and for the rest, shipping fixes that the evidence clearly supports, such as a broken button or a hidden cost, and comparing the same segment and step before and after.

Step 6: ship, re-measure, repeat

When a change wins, make it permanent, then go back to step 1 with the same reports. Check that the gain shows up in the segment you targeted and that nothing else fell, such as average order value. Then look for the biggest leak again, because it has usually moved. Keep a short log of each hypothesis and its result, including the ones that failed.

Where the leaks usually are: checkout friction

Baymard's cart abandonment list, last updated on 22 September 2025, puts the average documented cart abandonment rate at 70.22 percent, calculated from 50 different studies. Its survey of reasons found that 42 percent of US online shoppers had abandoned a cart because "I was just browsing / not ready to buy", which no checkout design will change.

With that group set aside, the top reasons in Baymard's list are "Extra costs too high (shipping, tax, fees)" at 40 percent, "Delivery was too slow" at 20 percent, "I didn't trust the site with my credit card information" at 19 percent, "The site wanted me to create an account" at 18 percent, and "Too long / complicated checkout process" and "Website had errors / crashed" at 17 percent each. Further down are "Returns policy wasn't satisfactory" at 13 percent, "I couldn't see / calculate total order cost up-front" at 12 percent, "The credit card was declined" at 10 percent and "There weren't enough payment methods" at 9 percent.

Baymard also reports that the average US checkout flow shows 23.48 form elements by default, where its testing finds an ideal flow can be as short as 12 to 14, and estimates that the average large ecommerce site can gain a 35.26 percent increase in conversion rate through better checkout design. That last figure is an estimate for large sites, not a promise for yours.

Use the list as hypotheses to check against your own recordings. Our guides to checkout funnel optimisation and optimising your e-commerce checkout cover the fixes themselves, and the cart abandonment calculator estimates what abandoned carts are costing you.

Speed

Google's own published data on speed and conversion is the Milliseconds Make Millions study, commissioned by Google and conducted by 55 and Deloitte. It covered 37 European and American brand sites and more than 30 million user sessions, with mobile load times monitored for 30 days at the end of 2019.

For the retail sites in the study, on mobile, a 0.1 second improvement across the four speed metrics measured came with a 3.2 percent increase in progression from product listing page to product detail page, a 9.1 percent increase from product detail page to add to basket, and retail consumers spending 9.2 percent more. Every page across the journey had to improve each metric by 0.1 seconds for those effects to be observed.

For targets, Google's Core Web Vitals guidance says Largest Contentful Paint should occur within 2.5 seconds of when the page starts loading, Interaction to Next Paint should be 200 milliseconds or less, and Cumulative Layout Shift should be 0.1 or less, measured at the 75th percentile of page loads and segmented across mobile and desktop.

Trust and returns

Two of Baymard's reasons are about confidence, not effort: not trusting the site with card details, at 19 percent, and an unsatisfactory returns policy, at 13 percent. The National Retail Federation's 2025 returns report, published on 15 October 2025 with Happy Returns, found that 82 percent of the US shoppers it surveyed, all of whom had returned an online purchase in the past year, cite free returns as a major consideration when making a purchase, up from 76 percent the year before, and that large US retailers expected 19.3 percent of online sales to be returned in 2025. A returns policy that is short and visible on the product page is a conversion feature; one hidden in the footer is not.

Product pages

In the Google study the largest retail gain in progression was from product detail page to add to basket. If your funnel shows a weak add to cart rate, this is the page to watch, and what to look for is missing information: delivery cost and date, returns terms, sizing and the total price. Baymard's finding that 12 percent abandoned because they could not see or calculate the total order cost up front is a product and cart page problem as much as a checkout one. Our guide to product page analytics covers how to measure each page.

Mobile

Treat mobile as its own store, because a change that helps on desktop can do nothing on a phone. That is why step 1 splits by device and why Google segments Core Web Vitals across mobile and desktop. When you watch mobile recordings, look for the things a desktop review never shows: elements too small to tap, a keyboard covering the field being filled, and a sticky bar hiding the button.

Where Arktis fits, and where it does not

Arktis covers the measurement steps. It reports sessions and conversions by traffic source and by device type, browser and operating system. For each page it shows sessions and a conversion rate, the share of sessions that visited the page and went on to convert. Session recordings, on the Growth plan and above, can be filtered by source, device, a page the session included, and whether the visitor has converted. Recording is buffered in the visitor's browser and discarded unless the visitor gives consent, so the recordings you watch are a sample of consenting visitors.

It also matches orders back to the visit and the ad click or campaign that brought the visitor, where it can: the tracker captures click identifiers such as fbclid and gclid on landing, and Arktis counts conversions per UTM campaign and per ad platform from matched Shopify and Stripe orders, next to the Meta spend it syncs per campaign.

Arktis has no A/B testing. It does not split traffic, serve variants or calculate statistical significance, and it has no feature flags or on-site surveys. Step 5 needs a dedicated experimentation tool, the category that includes Optimizely Web Experimentation and the experiments feature in PostHog. Google's own free testing tool, Google Optimize, has not been available since 30 September 2023. For the step-by-step counts inside a hosted checkout, use your e-commerce platform's own conversion report alongside Arktis.

Start with step 1

Split your conversion rate by source and device this week, find the one step that leaks most, and watch twenty sessions there before changing anything. To do that with recordings and with ad clicks tied to orders, start the 7-day Growth trial, or compare the tiers on the pricing page. For related reading, see cart abandonment analysis and reducing e-commerce bounce rate.

Sources

Baymard Institute: cart abandonment rate statistics, the 70.22 percent average from 50 studies, reasons for abandonment, form element counts and the 35.26 percent estimate, last updated 22 September 2025, accessed 1 October 2026

Google web.dev: Milliseconds make millions, the effect of a 0.1 second speed improvement on mobile retail funnel progression and spend, study by 55 and Deloitte commissioned by Google, page last updated 24 June 2020

Google web.dev: Web Vitals, Core Web Vitals thresholds and the 75th percentile across mobile and desktop, accessed 1 October 2026

National Retail Federation: 2025 retail returns landscape, 82 percent of surveyed US online returners citing free returns and 19.3 percent of online sales expected to be returned, published 15 October 2025

Shopify Help Centre: behaviour reports, the conversion rate definition and the four funnel steps, accessed 1 October 2026

Shopify Help Centre: session measurement update, the change to how sessions are counted, rolled out 21 to 23 September 2026, accessed 1 October 2026

Google Analytics Help: Google Optimize sunset, Optimize and Optimize 360 no longer available as of 30 September 2023, accessed 1 October 2026

Optimizely: Experimentation, the product family that includes Web Experimentation, an example of a dedicated A/B testing product, accessed 1 October 2026

PostHog: Experiments, an example of A/B testing inside a product analytics tool, accessed 1 October 2026