How Does Website Design Affect Ecommerce Conversion Rates?
Conversion is not lost evenly across a shop. It is lost in a small number of specific places, each with a known cause. This guide names the places, then takes them one at a time, starting with where the cheap wins are rather than where the funnel starts.
Where Conversion Is Actually Lost
Shops do not leak steadily. They leak at five identifiable points. Knowing which is which prevents the commonest waste in this work, which is redesigning the part that was already fine.
The homepage, where visitors arriving at the front door fail to get into a category. The mobile experience, which carries most of the traffic and receives the least design attention. The product page, where a decision either happens or does not. The trust moment, immediately before card details are entered. And checkout, which is last in the sequence and first in importance.
Those five are not equally weighted and they are not equally fixable, which matters more than their order.
Checkout is where the most money is lost and where the most is recoverable, because everything wrong with a checkout is within your control and needs nobody's permission. Mobile is where the largest gap usually sits between how much traffic arrives and how much attention the experience has received. The homepage, by contrast, receives more attention than its contribution warrants, because it is the page everybody in the business looks at.
So the order of investigation is not the order of the journey. Look at checkout and mobile first. Look at the homepage last, which is the opposite of how almost every redesign is scoped.
The Homepage's Real Job
A homepage has two jobs and neither is to impress. It has to tell a stranger what you sell, in words, then get them into a category quickly. Everything else is optional.
Why the carousel usually costs more than it earns. Most visitors see only the first panel, because the others move before anybody has finished reading. It occupies the most valuable space on the page while delivering a fraction of its content. It adds weight. And on a phone it frequently becomes something people swipe past by accident.
There is a further reason it persists. It explains why arguing against carousels rarely works. A carousel is usually an organisational compromise wearing a design costume. Three departments each wanted their thing featured, nobody would decide, so a rotating panel let everybody win. It is solving a meeting rather than serving a customer.
Why enormous hero images cost sales. They occupy exactly the space where routes into the catalogue should be. A visitor who must scroll before discovering what you sell has been asked to invest effort before receiving anything, which is a poor opening trade.
What a homepage should carry instead. A plain statement of what you sell. Clear routes into the main categories. A visible search box. A handful of products that are genuinely popular or genuinely seasonal. And enough evidence of a real business that a first time visitor relaxes slightly.
One distinction worth holding. For returning customers the homepage is a navigation tool. For new visitors it is an orientation tool. It has to do both. Designs that serve only one are usually serving the returning visitor, because that is who the business imagines when it pictures its own site.
Mobile
Most shopping happens on a phone. Most shops are designed on a large screen by somebody sitting still with a fast connection. The gap between those two sentences is where a great deal of money goes.
The specific failures. Tap targets too small or too close together, so the wrong thing gets pressed. Important actions placed where a thumb cannot comfortably reach on a tall screen. Forms that summon the wrong keyboard, do not accept autofill or use fields too small to check what was typed. Images sent at full size to somebody on mobile data. Fixed banners and cookie notices consuming a third of an already small screen. And interactions that depend on hovering, which does not exist on a touch screen.
Underneath all of that is one point that matters more than the rest. A desktop layout squeezed down is not a mobile design.
The reason is that mobile removes an option desktop takes for granted. On a wide screen, price, images, variant choices and delivery information can all sit beside each other, so a visitor takes them in together and no ordering decision is required. In one narrow column they arrive in sequence, so somebody has to decide what comes first, second and third.
That is the hardest mobile decision and it is a commercial one rather than a visual one. Which single thing does a shopper need first? For a considered purchase it is frequently the image, for a repeat purchase the price, for anything urgent the delivery date. Squeezing a desktop layout answers that question by accident.
How to check. On your own phone, on mobile data rather than office wifi, going all the way to a completed order. A resized browser window on a desktop reproduces the layout and none of the experience.
Trust Before Payment
A first time visitor about to pay is assessing one risk: what happens if this goes wrong. Every trust signal is an answer to that, which is why the effective ones are mundane and specific rather than impressive.
What reassures. A real postal address, findable rather than buried. A telephone number that looks answered. A returns position written in plain terms rather than defensively. Delivery cost and timing stated on the product page. Reviews with variation, detail and dates, including some that are less than perfect. A payment step that looks like part of the same shop. And small signs of an active business, such as a current copyright line and stock information that turns out to be accurate.
The timing point that most shops get wrong. Trust signals have to appear before the moment of doubt, not after it. A returns policy linked only from the footer of the checkout page arrives too late, because the hesitation happened on the product page while somebody was deciding whether to risk it. Put the answer where the question forms.
On reviews specifically. Uniformity is what makes them read as fabricated. Twenty short five star reviews all phrased similarly do more harm than six detailed ones including a mild complaint answered well. And a rule that is not negotiable: never mark up ratings that were not genuinely given. Structured data is an assertion about reality. A false assertion is a serious matter rather than an optimisation.
The pattern across all of it is that specificity reassures and vagueness reads as risk. A named town, a stated number of days and a real person in a photograph all cost nothing and do more than any badge.
Checkout
By this point the shopper has decided. They want the item, they believe you and they are attempting to pay. Nothing that happens now can win the sale. It can only lose it, which makes checkout the least forgiving part of a shop and the most rewarding to fix.
Forced account creation. Asking somebody to commit to a relationship before they have received anything. Offer it afterwards, when they have a reason to want one.
Form length. Every field is a request and every request needs a reason the shopper would accept. A delivery address is obviously reasonable. A question about how they heard of you, standing between them and their purchase, is not.
Costs appearing late. Delivery, handling or any other charge arriving at the final step reads as concealment whether or not it was intended that way.
Delivery options that say nothing. Standard and express are not answers. When will it arrive is the question.
Error handling that punishes. Rejecting an entry and then clearing what was typed, which is where a tired person on a phone stops trying. Explain what is wrong, keep everything else, then never make somebody start again.
Why this is the recoverable one: nobody else has to agree. No publisher, no supplier and no third party is involved. Each fault is discrete, identifiable and within your gift.
The single change we would make first on most shops is not in the checkout at all. It is showing delivery cost and expected date on the product page, which removes the surprise from checkout by answering it before anybody gets there.
Speed As A Conversion Issue
Speed belongs in this list because a shopper who is waiting has not yet reached any of the five points above. It is also the loss that hides most completely, since nobody who abandoned a slow page appears anywhere in your enquiries.
The part worth knowing here is that shops are slow because of decisions rather than misfortune. Images uploaded at full size. Themes carrying features nobody uses. Third party scripts. Apps, each contributing its own code, accumulating until nobody remembers what half of them do.
That makes it a conversion problem with an unusual property: most of the causes were introduced deliberately, one at a time, each for a reason that seemed good in isolation.
The whole subject, including what can be fixed without a rebuild and what cannot, sits in why ecommerce website speed affects sales and rankings.
Product Information As A Conversion Issue
The product page is where a decision either happens or does not. The usual reason it does not is that the page describes the product without answering the question standing between the shopper and the purchase.
Those questions are specific and knowable: will it fit, will it work with what I already own, what is actually in the box, how heavy is it, does it run small. Most shops publish the manufacturer's description, which was written to present a product to a retailer rather than to resolve a buyer's doubt.
Treated as a conversion issue, this is among the cheapest available improvements, because the questions do not have to be researched. Your customers are already asking them by email and telephone. Each one is a doubt that stopped somebody who did not bother to ask.
How to build those pages properly is set out in how to design ecommerce product pages that convert.
What Testing Can And Cannot Tell You
Split testing is presented as the responsible way to make these decisions. For most shops it is not available. Pretending otherwise wastes months.
Why. A test compares two versions and needs enough outcomes in each to distinguish a real difference from ordinary variation. Orders are the outcome that matters. A shop taking a modest number of orders each week will either wait many months for an answer or reach one that chance could have produced. Neither is useful. Stopping a test early because it looks promising is how businesses convince themselves of things that were never there.
What to do instead. Fix the known problems rather than testing them. A checkout that clears typed input when validation fails does not need a controlled experiment. It needs correcting. The same applies to hidden delivery costs, forced account creation and a missing address. These are not hypotheses. They are faults.
What works at small scale. Watching five real people attempt to buy something while you say nothing. Completing an order on your own phone, on mobile data, as though you had never seen the site. Looking at which pages sessions end on. And reading the questions customers actually send, since every one is a page that failed to answer something.
One thing this page will not do is quote a figure for what any of this is worth. Numbers circulate for exactly these changes and they describe somebody else's shop, at an unknown date, with an unknown baseline. If a supplier offers you a specific percentage for a specific change, ask who measured it and when. Treat a missing answer as the answer.
Common Mistakes
Five. The first two account for most of the wasted redesign budgets we see.
Redesigning on taste. Somebody senior dislikes the look, so the shop is rebuilt. The five points in block one are untouched, the new version performs the same or worse, then nobody can explain why because nothing measurable was addressed.
Changing several things at once. Understandable and it destroys the information. If a shop changes its homepage, its checkout and its product template together, the result cannot be attributed to any of them, so nothing is learned that would help next time.
Copying a large retailer. Their layout was built for a business with brand recognition, enormous traffic and customers who arrive already intending to buy. Copying the appearance imports none of that. Some of what they do works only because everybody already trusts them.
Chasing a design award. Judged by designers on appearance by people who are not trying to buy anything, which is a different test from the one your customers apply.
Adding urgency that is not true. Countdown timers that reset, stock warnings that never change and claims about other shoppers that are generated rather than counted. Customers recognise these more often than the businesses using them assume. The cost is the trust described in block four.
Checkout first.
Homepage last.
We look at where a shop actually leaks before proposing anything, which usually means the checkout and the phone experience rather than the page everybody in the business looks at. No invented uplift figures, on any proposal.
What a build covers:
Ongoing management afterwards sits inside our fixed monthly fee. Build work is quoted per project.
Twenty guides.
One subject.
This page covers where sales are lost. The rest of the series covers category and product pages, speed, what a build costs and why the search ceiling is set during the build.
