Most CRO audits hand you opinions dressed as findings. This is the checklist we run instead: thirty checks, three surfaces, and a bar you either clear or you don't.
Most CRO audits hand you a list of opinions with the word "findings" on the cover. Strengthen your value proposition. Improve trust signals. Optimize the mobile experience. You get to the end and still can't tell whether your store is fine or on fire, because nobody said what fine looks like.
A conversion rate optimization audit only works if every check has a bar you can clear or miss. "Is the add-to-cart button prominent" is not a bar. You can argue about that until somebody quits. "Is it at least 44 pixels tall, and does it stay on screen when you scroll past the buying block" is a bar. That one takes four seconds, and two people checking the same store get the same answer.
Below is the hand-run version of the checklist we use at Qosmic, in the order we run it. Thirty checks across three surfaces, ranked most expensive to least inside each one. By hand it's about an hour: fifteen minutes on the home page, twenty-five on the product page, twenty on the cart. None of it is clever. The work is in ranking what you find, not in finding it.
If you only have fifteen minutes
Five checks find most of the money.
- Is the add-to-cart button at least 44 pixels, and does it stay put when you scroll past the buying block?
- Does the cart show a shipping cost or an estimate before checkout, instead of "calculated at checkout"?
- Can you see the price without scrolling on a product page at 375 pixels wide?
- Does an email popup fire before you've scrolled?
- Can you reach a product page in two taps from the home page?
Any "no" there is a fix, not a hypothesis. Come back for the other twenty-five when you've got the hour.
CRO audit versus CRO analysis
People use these interchangeably. They're different jobs.
A CRO analysis starts with your data. You open analytics, build a conversion funnel, run a funnel analysis to see where people leave, cut it by device and traffic source, then watch session recordings to work out why. That's the quantitative and qualitative split everyone talks about. Numbers tell you where buyers drop off. Recordings and feedback tell you why. All of it needs traffic, a working analytics setup, and time you probably don't have this week.
A CRO audit starts with the storefront. You walk the site the way a buyer would and check it against a standard. No traffic required, no analytics required. It tells you what's broken.
Audit first, for a boring reason. Analysis will tell you that most of your mobile sessions die at the cart. Then you still have to go and look at the cart to find out why. If the checkout button is 30 pixels tall and the shipping line says "calculated at checkout," you didn't need a funnel report. You needed a phone.
The CRO framework below is the audit half. Run the analysis afterward, on the questions the audit couldn't settle.
Two rules before you start
Audit on mobile, at 375 pixels. Not desktop with the responsive toolbar open. An actual phone, or at minimum a 375-wide viewport on a throttled connection.
This isn't taste. Mobile is roughly 72% of ecommerce sessions and 55% of revenue. Desktop converts about 2.3 times better, 4.8% against 2.1%, per SaleCycle's 2026 numbers. Cart abandonment splits the same way, with Dynamic Yield putting mobile at 80.02% and desktop at 66.41%. The gap between where your traffic is and where your money is lives on a phone screen. Most audits get run on a 27-inch monitor, where everything looks fine. It usually does.
Fix in the order you find them. The checks are ranked inside each surface. A broken value proposition costs more than a missing cross-sell module. Get through the first five on each surface and you've found most of what matters.
The ecommerce CRO audit checklist
Surface 1: the home page (about 15 minutes)
1. Value proposition. Above the fold at 375 pixels, in selectable text rather than baked into the hero image, and specific enough that you couldn't paste it onto a competitor's site unchanged. Your brand name doesn't count.
2. Primary call to action. One dominant CTA in the first viewport. At least 44 by 44 pixels, high contrast against the hero, a verb-plus-object label, pointing at a collection or a product page. Not About. Not the blog.
3. Navigation. A product page is two taps away or fewer. Top-level labels name categories a stranger would recognize, not your internal vocabulary. No category carrying real revenue buried three levels down.
4. Product merchandising. Real shoppable products with an image, a name, a price and a working link, inside the first two scrolls. Lifestyle banners standing in for products don't count.
5. Hero media. It's your LCP element, so it can't be lazy-loaded. Modern format, sized for mobile, explicit dimensions so nothing jumps after paint, under about 2.5 seconds on a throttled connection.
6. Overlays. The email popup doesn't fire on load. The dismiss control is visible and at least 44 pixels. Never two overlays at once on mobile. Nothing covers the hero in the first three seconds. Nobody has ever been won over by a popup that arrived before they scrolled.
7. Header. Sticks through scroll. Cart icon carries a live item count. Cart is one tap from anywhere on the page.
8. Site search. In the header once you pass about 50 SKUs. Returns something relevant for a plain product term and survives one deliberate typo.
9. Brand story. At least one specific claim someone could check. Sourcing, formulation, warranty, how it's made. "Quality you can trust" isn't a claim, it's wallpaper.
10. Footer. Shipping, returns, contact and FAQ all linked, all resolving to real pages. No leftover Shopify boilerplate. A contact method that actually reaches someone.
Surface 2: the product page (about 25 minutes)
This is where most stores lose the sale. It's also the page people audit least carefully, because they've looked at it a thousand times and stopped seeing it.
1. Add to cart. Stays on screen on mobile once the inline button scrolls away. At least 44 pixels. High contrast. One primary action. The label says what happens. Never disabled without a reason sitting right next to it.
2. Buying block layout. Title, price, rating and add-to-cart all reachable inside about one and a half scrolls at 375 pixels. If the image eats the whole first viewport while the price sits below the fold, that's not a style choice. That's a leak.
3. Image gallery. Multiple angles, a scale reference, an in-use shot, a detail macro. Swipeable, zoomable, dimensioned so nothing shifts. The bar: nobody should have to wonder what the thing actually looks like.
4. Reviews. Rating and count near the title, tapping through to the full set. Show the distribution, not just the average. A 4.2 made of fives and ones is a different product than a 4.2 made of fours. Individual reviews carry substance, dates, and photos where the category calls for it.
5. Shipping and delivery. A delivery estimate and the shipping cost or threshold, both visible without opening an accordion. Don't save it for checkout.
6. Price. Next to the title. Currency unambiguous. Per-unit price where it matters. Any strikethrough reflects a real previous price rather than permanent decoration.
7. Returns, guarantee, trust marks. Grouped within a scroll of the add-to-cart button and stated in specifics. "30-day free returns" works. A wall of badges doesn't.
8. Product description. Benefit-led opening, bulleted key points, specs in a table. Scannable in ten seconds. Not manufacturer boilerplate, not one long paragraph.
9. Objection handling. Every category has one question that blocks the sale. Size, ingredients, compatibility, care, dimensions. It gets answered on the page, not in a support email.
10. Express checkout. Present, with enough visual hierarchy against add-to-cart that it doesn't steal buyers who meant to keep shopping.
Surface 3: the cart (about 20 minutes)
Audit the drawer and /cart together. Same surface, and you edit them together.
Worth knowing what you're up against here. Cart abandonment averages 70.22% across Baymard's meta-analysis of 50 published studies, and they put around $260 billion of it as recoverable in the US and EU through checkout design alone. This surface has the best ratio of cheap fixes to lost revenue anywhere on your store.
1. Checkout button. One dominant button visible without scrolling, in both the drawer and /cart. At least 44 pixels, clearly labeled, still reachable when the cart holds a lot of items. "Continue shopping" sits below it in both position and weight, and takes the buyer back to where they left off.
2. Order summary. Subtotal, shipping status, and tax or duty treatment all stated before the buyer commits. "Calculated at checkout" with no estimate is the single biggest abandonment driver on this surface. An estimate or a threshold statement clears the bar.
3. Add-to-cart confirmation. Immediate visible confirmation through a slide-out drawer that keeps the buyer where they were. Not a silent add. Not a hard redirect to /cart that loses their place.
4. Cart-to-checkout path. Two taps maximum from drawer to checkout. No interstitial upsell page. No forced account creation.
5. Express wallets. Shop Pay, Apple Pay, PayPal or Google Pay present, at least one reachable without scrolling on mobile, positioned to sit alongside the standard checkout button rather than bury it.
6. Line items. Thumbnail, product name, selected variant, per-item price, quantity stepper, remove control. Quantity changes update the total without reloading the page. Controls at least 44 pixels.
7. Pre-checkout gates. Terms checkboxes, age verification, order notes, gift options, delivery-date pickers. Nothing blocks the checkout button without a visible, specific error, and optional fields stay collapsed by default.
8. Delivery expectation. An arrival window at the cart stage, matching what the product page promised. Buyers shouldn't have to enter checkout to find out when the order lands.
9. Returns and trust access. Refund policy, shipping policy and a contact method all reachable from the cart, plus visible secure-payment indicators, so a nervous buyer can check the terms without leaving.
10. Cross-sell. Below the checkout CTA, limited to genuinely related items, addable in one tap without leaving the cart or resetting the scroll position.
The scorecard
Copy this. Tally as you go.
Surface · Checks · Failed
Home page · 10
Product page · 10
Cart · 10
Total · 30
Write down the failing condition, not the category. "ATC is 32 pixels, below the 44 bar, and doesn't stay put past the buying block" is a work item. "Improve mobile CTA" is a note you'll ignore.
What a CRO report should actually say
Most conversion rate audits end up as a slide deck of screenshots with red circles on them. That's not a report. That's a mood.
A useful one has three things.
A pass/fail count by surface, not a score out of 100. Those scores are invented, and nobody has ever looked at a 72 and felt any urgency. Thirty checks, three surfaces, how many failed in each. The spread tells you more than the total.
Failures bunched on the product page mean a merchandising problem. Buyers arrive interested and leave unconvinced. Failures on the cart mean friction, which is usually the cheapest thing to fix and the fastest to show up in revenue. Failures on the home page mean a positioning problem, which is expensive to fix properly and very commonly fixed badly.
The specific failing condition, with its bar. Someone else should be able to re-check any finding in under a minute. If they can't, it isn't a finding.
One ordered fix list, not three. Take the worst surface and work down it. Spread three fixes across three surfaces and you won't be able to attribute the result to any of them, which means you learn nothing and guess again next quarter.
How often to run it again
An audit is a snapshot, and storefronts don't hold still. Run the full thirty again when:
- A theme updates. These quietly revert custom work. Sticky add-to-cart is the usual casualty.
- You install or remove an app. Apps inject overlays, shift layout, and pile weight onto the LCP element.
- You ship a seasonal or campaign redesign. Banners are the most common reason a hero stops passing.
- A quarter goes by. Things drift even when nobody touches anything.
The interval matters more than the instrument. A rough audit every month beats a beautiful one every year, because the failures that cost you money are usually recent. Something changed, nobody noticed, and the number moved three weeks later.
That gap is the reason Qosmic exists. It re-runs these and five hundred more against your store continuously, so a theme update that quietly kills your sticky add-to-cart shows up the same week instead of next quarter. You can get the same result with a spreadsheet and a calendar reminder, and plenty of good operators do. The thing that matters is that somebody is looking.
Where this stops working
You can't A/B test your way through this list. Most stores don't have the traffic to reach significance on thirty tests, and running them one after another would take years. That's not an argument against testing. It's an argument for fixing what's plainly broken first and saving tests for the genuine coin flips.
The checks above are the kind you don't need a test to settle. If the price is below the fold on mobile, that's not a hypothesis. If the popup fires on load, that's not a hypothesis. Test what reasonable people could disagree about. Fix what they couldn't.
One more limit worth being honest about. This audits your storefront, not your traffic. A store that passes all thirty can still convert badly because it's buying the wrong visitors. The audit tells you the shop is in order. It doesn't tell you the right people are walking in.
Ecommerce CRO is just getting more out of the visitors you already have instead of paying for more of them. This is where it starts, before the analytics and the testing and the roadmap somebody built on a hunch.
