Most ecommerce CRO advice is a pile of tactics with no order to it. The order is the part that matters, and testing belongs third, not first.
Most ecommerce CRO advice is a pile of tactics with no order to it. Add urgency. Reduce friction. Use social proof. Test your headline. Every item on the list is defensible, and the list as a whole is useless, because it never tells you what to do on Monday.
The order is the part that matters. Get the sequence wrong and you'll spend six months testing button colors on a store that's losing people for reasons no test was ever going to surface.
So: what it is, the sequence that works, and the parts you can safely ignore until later.
What ecommerce CRO actually is
Conversion rate optimization is getting more out of the visitors you already have instead of paying for more of them. That's the whole idea. Everything else is method.
The conversion rate formula is as simple as it looks:
Conversion rate = (orders ÷ sessions) × 100
Two things about that formula are worth knowing before you build a quarter of work on top of it.
It counts sessions, not people. Someone browses on Tuesday, comes back Thursday, buys on Sunday. That's three sessions and one order, which reads as 33%. Count the same behavior per user and it's 100%. Session-based is the Shopify default and it's what almost every published benchmark uses, so it's the right one to work with. Just know that a store that looks like it's underperforming by half is sometimes only being counted differently.
It's an outcome, not a diagnosis. Your conversion rate tells you something is wrong. It never tells you where. Two stores at 1.2% can have completely unrelated problems, and the number looks identical in both dashboards.
As for what counts as good, the median Shopify store sits somewhere between 1.4% and 2.3%, and the average is the least interesting thing in that data.
That gap is the actual subject of this article. Not the average, which is an average of stores that aren't you.
One correction to the framing before we go further. Conversion rate is a bad single target. Discount everything by 30% and your conversion rate goes up while your business gets worse. The metric that doesn't lie to you is revenue per session, and over a longer horizon, customer lifetime value. Optimize conversion rate as a proxy for those, not instead of them. Plenty of stores have optimized themselves into a worse business by treating one number as the goal.
The sequence: fix, then measure, then test
Almost every guide to ecommerce conversion optimization opens with testing. Testing is third.
One: fix what's plainly broken. Walk your storefront on a phone and check it against a standard. No traffic required, no analytics required, no hypothesis required. If your price sits below the fold at 375 pixels, that isn't a hypothesis, it's a bug. This is the thirty-check audit, and by hand it takes about an hour.
Two: measure what you can't see by looking. Now open analytics. Where in the funnel are people leaving, on which device, from which traffic source. Watch session recordings for the parts the numbers can't explain. This step needs real traffic and a working analytics setup, which is why it's second and not first.
Three: test what reasonable people could disagree about. Two headlines, both plausible. Two page layouts, both defensible. That's what a test is for.
The reason for this order is arithmetic, and it's worth doing the arithmetic once because it settles a lot of arguments.
Take a store converting at 2% and suppose you want to detect a 10% relative improvement, so 2% against 2.2%. At 95% confidence and 80% power, the standard two-proportion sample size calculation puts that at roughly 80,000 sessions per variant. Two variants, so 160,000 sessions, and that's for one test.
A store doing 20,000 sessions a month would need about eight months to finish it. If it came back flat, which most tests do, you'd have learned nothing and burned the year.
Meanwhile the audit takes an hour and doesn't need a single session.
This is why testing-first advice quietly assumes you're an enterprise. Fix first because it's free and certain. Measure second because it's cheap and specific. Test last, and only on things that genuinely could go either way.
Where the money actually leaks
Three surfaces, and they fail in different ways.
The home page is a positioning problem when it fails. People land, don't understand what you sell or why it's for them, and leave. Expensive to fix properly and very commonly fixed badly, because everyone has an opinion about the home page and most of those opinions are about taste.
The product page 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. Failures cluster here when buyers arrive interested and leave unconvinced: no scale reference in the photos, the one blocking question unanswered, shipping cost withheld until checkout.
The cart is friction, and it's almost always the cheapest thing to fix. Cart abandonment averages 70.22% across Baymard's meta-analysis of fifty published studies. A large share of that is recoverable through checkout design alone.
If you want a fast read on which surface is yours, use the two steps underneath conversion rate. Add-to-cart rate below your sector median means the problem is upstream on the product pages. A healthy add-to-cart rate with a weak cart-to-purchase rate means it's the cart. Two numbers and you know which half of the store to work on.
And check the device split before anything else. Mobile is the majority of sessions for most stores and the minority of revenue, and desktop converts about 1.6 times better at the median. Your blended conversion rate is mostly a report on how your mobile experience is doing.
Conversion rate optimization best practices, with actual bars
Most best-practice lists are unfalsifiable. "Make your CTA prominent" can't be checked, so two people looking at the same store will disagree forever. These have bars you either clear or you don't.
Put the price above the fold on mobile. At 375 pixels, title and price and add-to-cart should all be reachable inside about one and a half scrolls. If the hero image eats the entire first viewport, that's a leak wearing a design decision.
Make tap targets at least 44 pixels. Add to cart, checkout, quantity steppers, popup dismiss buttons. This one is measurable in four seconds and fails constantly.
Keep the add-to-cart button on screen. Once the inline button scrolls away on mobile, a sticky one takes its place. Theme updates kill this quietly and often.
Show shipping cost before checkout. "Calculated at checkout" with no estimate is the single biggest abandonment driver on the cart. A threshold statement clears the bar. A real estimate is better.
Don't fire the email popup on load. Nobody has ever been won over by a popup that arrived before they scrolled. Give it a scroll depth or an exit intent, and never run two overlays at once on mobile.
Show the review distribution, not just the average. A 4.2 made of fives and ones is a different product than a 4.2 made of fours, and buyers know it.
Answer the one blocking question on the page. Every category has one: sizing, ingredients, compatibility, care, dimensions. If the answer lives in a support email, you're losing the people who won't send one.
Give buyers a scale reference. One photo where the product sits next to something whose size everyone already knows. Returns data is full of people who guessed wrong.
Cut apps you don't need before you add speed tooling. Weight is the most common cause of slow Shopify stores, and you can't fix a bloated stack by adding to it.
Pick one thing at a time. Ship three changes across three surfaces and you won't be able to attribute the result to any of them. You'll learn nothing and guess again next quarter.
None of that is clever. Most CRO tips aren't. The work is in ranking what you find and then actually doing it, which is a scheduling problem more than an insight problem.
Doing it yourself, hiring an agency, or buying software
All three work. They fail in different ways, and the honest version of this comparison isn't that one wins.
Doing it yourself is free and it's how you learn what your store actually does. The failure mode is that it never happens twice. The audit is a snapshot, storefronts drift, and the version of you that had a spare hour in March does not reappear in June. If you go this route, the calendar reminder matters more than the checklist.
An agency brings judgment you don't have and a deadline you'll respect, which is worth more than people admit. It's the right call for a redesign, a replatform, or a positioning problem, because those need taste and argument, not just detection. The failure modes are cost, a reporting cadence measured in months, and the fact that the work stops when the retainer does.
Software is continuous, which is the thing humans are worst at. It's also narrower: a tool can tell you your sticky add-to-cart broke three days after a theme update, and it can't tell you your brand positioning is confusing.
We build Qosmic, so read this knowing that. It sits in the third category with one difference worth naming: most tools in this space stop at telling you. Qosmic runs five hundred checks continuously, reads GA4 and Hotjar and Klaviyo alongside Shopify so the storefront findings sit next to the numbers, and then builds the fix inside your own theme using your own components and stages it as a draft you approve. Detection and repair in the same place, because the gap between "we found it" and "somebody shipped it" is where most CRO programs die.
You can get a version of that with a spreadsheet, a calendar reminder and a developer who owes you a favor. Plenty of good operators do exactly that. The thing that matters is that somebody is looking on a regular schedule, because the failures that cost you money are usually recent. Something changed, nobody noticed, and the number moved three weeks later.
What to do first, depending on where you are
Under about 10,000 sessions a month. Run the audit, fix what fails, run it again after every theme update or app install. Don't test anything. You don't have the traffic and you won't for a while, and that's fine. The fix list will outrun what testing could tell you anyway.
Around 10,000 to 50,000. Audit first, then get analytics honest: server-side events if you can, a clean funnel, and the device split visible. Start testing only the genuine coin flips, one at a time, and expect each one to take weeks.
Above 50,000. Now testing earns its keep and a real program makes sense. Keep auditing anyway. Every store we look at above this threshold has at least one broken thing that no test would have caught, because nobody thought to look.
At any size, if you've just replatformed or redesigned. Audit immediately. Redesigns break more than they fix about as often as not, and the damage shows up in revenue about a month after anyone would think to check.
The short version
Ecommerce CRO is getting more from the traffic you have. Conversion rate measures it badly on its own, so watch revenue per session too.
Fix, then measure, then test. In that order, because fixing is free and certain, and a single test at 2% baseline needs roughly 80,000 sessions per variant to detect a 10% lift.
Three surfaces. Home page failures are positioning, product page failures are persuasion, cart failures are friction and they're the cheapest to fix.
Best practices only count if they have a bar. If two people can look at the same store and disagree about whether it passes, it isn't a check, it's a preference.
And run it again on a schedule. Storefronts don't hold still, and the expensive problems are almost always the recent ones.
