CRO · Research and A/B testing

Conversion rate optimization: fewer guesses, more tested changes.

Our conversion rate optimization program finds where your store loses buyers, ranks the fixes, and tests the ones that need proof with properly sized A/B tests. Because we also build, test variants don’t wait in a developer’s queue, and every result, win or loss, goes into your log.

Get my free audit Chat on WhatsApp

200+ clients served 500+ projects delivered

Free conversion review

Send your store URL and we will show where sales leak.

Three fields and you are done. We review the shopping journey for conversion rate optimization and send a PDF report of the leaks we find, ranked by impact.

  • Homepage to checkout reviewed for conversion rate optimization
  • Leaks ranked by likely revenue impact
  • A PDF report of the findings, and a video meeting if you want one
Chat on WhatsApp Prefer to talk first? Send us your store URL on WhatsApp.

Your PDF report comes by email. If you want to discuss going ahead after reading it, we’ll set up a video meeting. We use your details only for this audit.

Results

A/B test results from a client store.

One client A/B test from our conversion rate optimization work: the hypothesis, the single change, the sample and the timeframe, with the measurement design and an independent statistical check.

Client result

Mobile product pages: delivery & returns visibility

For a US home-goods and soft-furnishings store on Shopify, the delivery and returns policy was available on mobile product pages only inside a collapsed accordion. The hypothesis: shoppers hesitate when they cannot easily see the shipping threshold and returns terms at the purchase decision.

The test changed one component. The control kept the policy inside the accordion; the variant added the same existing policy as a compact row immediately beneath Add to cart: “Free US shipping $75+ · 30-day returns.” The shipping offer itself, return rights, prices, product images and checkout did not change, and the variant kept the original accordion — the summary row was the only addition.

Across 28 days (August 3–30, 2026), with eligible US mobile product-page visitors split 50/50, purchase conversion was 2.60% in the control (628 purchasers from 24,186 visitors) and 3.10% in the variant (752 from 24,234): a 19.51% relative lift, or 0.51 percentage points. Product revenue per visitor rose from $2.14 to $2.54.

  • 2.60%3.10% Purchase conversion rate
  • $2.14$2.54 Product revenue per visitor
  • 628752 Unique purchasing visitors
Independent analysis of a client A/B test for an anonymous US home-goods Shopify store: mobile product-page delivery and returns visibility, August 3–30, 2026. Control conversion 2.60% (628 purchasers from 24,186 visitors), variant 3.10% (752 from 24,234), relative conversion lift +19.51%, product revenue per visitor $2.14 vs $2.54
Aug 3–30, 2026 · 28 days · 50/50 visitor split · eligible US mobile product-page visitors · USD · anonymous store.

The conversion rate optimization test: one component, one hypothesis

  1. Hypothesis

    The design: Shoppers may hesitate if they cannot easily see the shipping threshold and returns policy at the purchase decision.

    The rule: The test surfaces the existing terms beneath the primary buying button. It does not change the actual shipping offer, return rights, price, product images or checkout.

  2. Control

    The design: The delivery and returns information stays inside a collapsed “Shipping & returns” accordion: available, but hidden until the shopper opens it.

    The rule: Everything else on the product page stays exactly as it was.

  3. Variant

    The design: The same existing policy appears as one compact row immediately beneath Add to cart: “Free US shipping $75+ · 30-day returns.”

    The rule: The variant retains the original accordion. Only the new summary row is introduced: no offer change, no new discount.

  4. Scope

    The design: One policy-visibility component test on mobile product pages, for eligible US visitors who enter on a product page.

    The rule: Not a full product-page redesign and not a checkout test. Only eligible mobile product-page traffic is measured.

The product shown in the design illustrations is an illustration and does not identify a real store or SKU.

Conversion rate optimization A/B test results by group, August 3–30, 2026
MetricControlVariant
Eligible experiment visitors24,18624,234
Unique purchasing visitors628752
Purchase conversion rate2.60%3.10%
Qualifying orders628752
Product revenue$51,791.16$61,611.36
Product revenue per visitor$2.14$2.54
Average qualifying order value$82.47$81.93

The relative purchase-rate lift is 19.51%, calculated from unrounded rates, not from the displayed 2.60% and 3.10%. The absolute difference is 0.51 percentage points. Revenue per visitor increases 18.73%, while average order value is slightly lower. Those are different measures and should not be combined into one “conversion lift.”

How this conversion rate optimization test was measured

Eligible US mobile visitors enter on a product page and keep the same assignment on return visits within the same browser identity. Returning sessions do not add to the sample denominator. Visitors are not sessions or proven cross-device people. A conversion is at least one qualifying paid order within seven days of the first eligible exposure; the test data contains exactly one order per purchasing visitor.

The observation window extends past the exposure end so late entrants receive the same seven days as early entrants, with the data frozen on September 7. The cumulative chart groups mature purchase outcomes by first-exposure cohort date, not order-placement date. These rates are not the entire store’s Shopify session conversion rate. Product revenue excludes tax and shipping; refunds, cancellations and product costs are not modelled. Revenue per visitor is therefore descriptive product revenue per eligible visitor, not contribution profit. No preregistration is claimed for this test.

Independent statistical cross-check: for the primary conversion metric, a fixed-horizon two-sided pooled two-proportion z-test gives p = 0.000810, and a 95% confidence interval for the absolute rate difference is +0.21 to +0.80 percentage points. This is an independent frequentist check on the test data; it is not a native Intelligems Bayesian “probability to beat control.” The 50/50 sample-ratio check gives p = 0.827 and does not flag the aggregate allocation at a 1% threshold; that alone does not establish correct randomization, instrumentation or absence of bias. Revenue per visitor is a secondary descriptive measure, with no separate significance claim.

The content test was designed for Intelligems Content Testing. The report and statistics on this page are an independent analysis of the test data, not a native testing-tool export, and the product pages shown are design illustrations of the two versions, not screenshots of the client store.

The conversion rate optimization test in full: the change, the statistics, and the two designs

Design illustration of the single change tested: the control keeps the delivery and returns policy inside an accordion on the product page; the variant shows the same policy in a row beneath Add to Cart: “Free US shipping $75+ · 30-day returns”
The control-versus-variant change: one component, one hypothesis. Design illustration — not a screenshot of the client store.
Independent statistical cross-check of the same A/B test: absolute purchase-rate difference +0.51 percentage points with a 95% interval of +0.21 to +0.80, primary two-sided p = 0.000810, allocation check p = 0.827, observed samples 24,186 and 24,234 against a planned threshold of 20,000 per arm
Independent statistical cross-check: the difference estimate, the sample-size plan and the analysis rules. Independently rendered — not a testing-tool export.
Design illustration of the control mobile product page: Textured Cotton Throw, $79.00, Add to Cart, with the delivery and returns policy inside the collapsed Shipping & returns accordion
Control, as designed: the policy sits inside the accordion. Design illustration.
Design illustration of the variant mobile product page: the same product page with the “Free US shipping $75+ · 30-day returns” row directly beneath Add to Cart
Variant, as designed: the same policy visible beneath Add to Cart. Design illustration.

Brands we’ve worked with

  • Eric Originals logo
  • E-Com Fashion logo
  • Amita Naithani logo
  • Eye on Security logo
  • History of Soccer logo
  • Sabulski Window Tinting logo
  • CleverBoss logo
  • AkosMD logo
  • Immigration Desk logo
  • ONE400 logo
  • Rinat Brodach logo
  • New York Estate Jewelry logo
  • Terese Sydonna logo
  • Dana Cooper logo
  • VMORA logo

Fit

Who it’s for

  • Stores with enough traffic and orders to run tests that reach a clear result
  • Brands paying for traffic that converts worse than they think it should
  • Teams that redesign on opinion and want evidence instead
  • Stores on Shopify, WooCommerce or BigCommerce, with GA4 working or ready to fix

Not the right fit for low-traffic stores, where tests can’t reach a result in a sensible time. For those, we fix the obvious problems directly and track the before and after, and say so in the audit.

Problems we fix

Why conversion rate optimization often goes nowhere.

  • 01

    Tests without research

    Button colors and headline swaps chosen at random rarely move revenue. The big problems show up in analytics, recordings and customer questions.

  • 02

    Calling tests too early

    Stopping a test the day it looks good, or before a full week of traffic, produces winners that disappear after launch.

  • 03

    Too little traffic per test

    Testing a page few people see means waiting months or acting on noise. Tests belong where the traffic and the money are.

  • 04

    Tracking that can’t measure the goal

    If purchases or add-to-carts aren’t tracked correctly in GA4 and the testing tool, no test result can be trusted.

  • 05

    Variants stuck in development

    Ideas wait weeks for a developer, so the program runs one test a quarter instead of a steady stream.

  • 06

    No record of what was learned

    Without a log of every test, the same ideas get tested twice and nobody remembers why a page looks the way it does.

In 1:10 · animated

Before we change anything, we check five things.

A short animated walk-through of the first checks in our audit. No sound needed: everything is on screen, and the transcript is below.

What we check first before a CRO test, an animated explainer of 1:10 with on-screen captions and controls to pause.

Read the transcript
  1. What we check first before a CRO test Five checks, research before any test.
  2. 1. Can the goal be measured? If purchases aren’t tracked correctly in GA4 and the testing tool, no result can be trusted. So we compare tracked purchases with store orders first. (Example store: the tool misses 16% of orders.)
  3. 2. Where does the funnel leak? GA4 funnels by device show where buyers drop out. Baymard Institute’s collected studies put average documented cart abandonment at 70.22%. (Example store, mobile. Benchmark source: Baymard Institute.)
  4. 3. Which fixes need no test? Broken buttons, surprise costs and a forced account are bugs, not hypotheses. They get fixed straight away. (Example store.)
  5. 4. Is there enough traffic to test? A page few people see takes months to test or produces noise. We size each test before it starts and run it for full weeks. (Example: a collection page with low traffic, sized before the test.)
  6. 5. Are decision rules written down? Stopping the day a test looks good produces winners that vanish. The metric, run time and rule are agreed before launch.
  7. Then we fix them, in that order. Want this checked on your store? The free growth audit covers all five, in writing.

Our approach

Conversion rate optimization as a program, not a project.

Research comes first. We look at GA4 funnels by device and traffic source, session recordings and heatmaps, on-site search, customer service questions and short surveys, then walk the store as a buyer would. Baymard Institute’s collected studies put the average documented cart abandonment rate at 70.22%, and its research lists many causes that are fixable without a test, such as surprise costs or a forced account.

Every finding becomes a hypothesis in a ranked backlog: what we think is wrong, what we’ll change, what we expect to happen, and how many orders the page sees. Some fixes are obvious bugs and simply get fixed. The rest become A/B tests, prioritized by expected impact, confidence and effort.

We size each test before it starts. Using your traffic and conversion rate, we work out how long it needs to detect a meaningful difference, run it for full weeks to cover weekday patterns, and decide in advance what result counts as a win. Nielsen Norman Group’s A/B testing 101 covers why these steps matter.

Because we code, we build the variants ourselves in your theme or the testing tool, and check them on real devices before launch. Speed is part of conversion too, so slow pages go to our site speed work, measured against Google’s Core Web Vitals.

Results go into a test log with screenshots, numbers and what we learned. Winners get built properly into the store; losers are just as useful, because they stop the same idea from being argued over again. When paid traffic is involved, we share the findings with the people running Meta and Google campaigns, so ads and pages tell the same story.

Not every improvement needs a test. Broken filters, missing size guides, slow images and confusing shipping costs are fixed straight away, then measured before and after. A/B tests are saved for changes where opinions differ and the page sees enough orders, so testing time goes where the uncertainty is.

Conversion rate optimization research sources: analytics, recordings, surveys, support questions and a store walkthrough
Research finds the problems before any test is planned.

How we use AI here

How we use AI in conversion rate optimization.

AI makes CRO research faster. The A/B tests still decide what works.

  • Research summaries

    AI helps summarize session recordings, heatmap notes, on-site surveys and support emails into a list of problems shoppers actually hit.

  • Voice of the customer

    Reviews and survey answers are clustered by theme, so test ideas start from what buyers say, not what we guess.

  • Variant drafts

    AI drafts copy and layout options for each hypothesis; we build the strongest ones and test them properly.

Where it doesn’t help: “AI test ideas” are a brainstorm, not evidence. A change ships when an A/B test with enough traffic says it wins, not when a tool predicts it will.

What’s included

Deliverables, not promises.

One real result, on our Shopify CRO page: a store’s conversion rate went from 1.59% to 2.86%, comparing 30 days before the changes with 30 days after, not a split test.

  • Conversion research

    GA4 funnels, recordings, heatmaps, on-site search, support questions and a short survey, summarized as findings with evidence.

  • Tracking check

    Purchases, add-to-carts and checkout steps verified in GA4 and the testing tool before any test runs.

  • Ranked hypothesis backlog

    Every idea with its evidence, expected impact, confidence, effort and the traffic the page gets.

  • Test design and sizing

    Sample size and duration worked out before launch, with the decision rule written down.

  • Variant build and QA

    Variants built in your theme or testing tool and checked on real devices and browsers.

  • Analysis and rollout

    Results analyzed by device and traffic source, winners built properly into the store.

  • Test log

    Every test with screenshots, numbers and the lesson, kept in a document you own.

How findings become a ranked A/B testing backlog: evidence, impact, confidence, effort and traffic
Each idea is scored before it earns a test.

How we run it

Four steps, in this order.

Research before tests, and tests sized before they start. Typical timing: research in the first week, the first tests live within 2 weeks, and each test running at least 2 weeks before we judge its conversion rate.

  1. Step 1

    Research

    Analytics, recordings, surveys and a full walk through the store, with tracking checked first.

  2. Step 2

    Backlog

    Findings turned into ranked hypotheses; obvious bugs fixed right away.

  3. Step 3

    Test

    Variants built and checked, tests sized and run for full weeks, decision rules set in advance.

  4. Step 4

    Learn and roll out

    Winners built into the store, every result logged, and the next tests chosen from what we learned.

The four steps of our conversion rate optimization program
Research, backlog, test, then learn and roll out.

Pricing

Starting prices, in US dollars.

The audit is free. Paid work starts at these prices, and your exact cost is fixed in writing before any work begins.

  • Free audit

    Free no obligation

    Send your store URL and we will show where sales leak.

    Get my free audit

  • Fixed scope

    CRO fix sprint

    From $500 one-off

    The quick wins from the conversion rate optimization review, fixed in one sprint.

    • Top issues fixed
    • Tested before going live
    • Done within two weeks
    Start with the free audit

  • Monthly CRO

    From $600 monthly

    Ongoing conversion rate optimization tests and fixes for stores that want steady gains.

    • Tests run each month
    • Winning versions made live
    • A monthly report
    Ask about this option

All prices are starting prices in US dollars. Custom work beyond these options is quoted after the audit.

By industry

What changes for your category.

  • Jewelry

    Jewelry pages need detail and trust: zoomable images, size guides, materials, certificates and returns. Those are where we look first, and often where the tests are.

  • Fashion & apparel

    Fashion conversion hangs on fit and returns. Size guides, fit notes from reviews and clear return terms are frequent test areas, along with collection filters on mobile.

  • Luxury & high-ticket

    High-ticket buyers need reassurance and contact. We test trust elements, financing visibility and ways to ask a question before buying, and measure inquiries as well as orders.

Sources

The documentation we work from.

Platform rules change. These are the official pages we check the work against, so you can check it too.

FAQ

Straight answers.

Can’t find yours? Ask on WhatsApp or by email and we’ll answer it.

How much traffic do we need for A/B testing?

Enough orders on the page to detect a meaningful difference in a few weeks. We calculate it from your traffic and conversion rate before proposing tests. Below that, we fix clear problems directly and measure before and after.

Which testing tools do you use?

We use a testing tool that fits your platform and budget, or theme templates and server-side splits where that’s cleaner. We recommend one after the research, based on your traffic and stack.

How long does a test run?

At least one full week, usually two or more, so weekday and weekend behavior are both included. The exact length comes from the sample size calculation, decided before launch.

Will tests slow down the store?

They can, if the tool flickers or loads heavy scripts. We check speed and layout shift with the variant live, and prefer approaches that render the variant without a visible swap.

Do you redesign the whole site?

Not as part of this program. We change the pages that lose the most buyers, one tested change at a time. A full redesign is a separate project.

What conversion rate should a store have?

There’s no single benchmark worth chasing, because conversion rate depends on traffic sources, price, category and device mix. We compare your store against its own past and also track revenue per visitor, which counts order value too.

Do you analyze mobile and desktop separately?

Yes. Every result is broken down by device, and when phones and desktops behave differently, we run separate tests or roll out the winner only where it won.

What happens to tests that lose?

They go into the log with what we learned. A losing test still tells you something about your buyers, and stops the same idea from being retried without a reason.

Latest guides

Read before you decide.

All guides

Free growth audit

Find out what’s holding your store back.

Tell us about your store. We’ll review your tracking, campaigns and SEO and send you a written audit with the top fixes, ranked by impact.

  • Written audit as a PDF, delivered within 2 workdays (see a sample)
  • A video meeting to go through it, if you want to discuss next steps
  • Free and no obligation
  • Read-only access or screenshots are enough
Chat on WhatsApp [email protected]

Free audit
Chat on WhatsApp