Ecommerce conversion optimisation attracts a great deal of attention to the wrong places. Button colours and hero images get tested; checkout friction and site search go unexamined, and that is where most of the revenue is lost.
Ordered by typical impact.
1. Checkout
Checkout abandonment runs high across the industry, and the reasons are consistent and fixable.
Unexpected costs. The most cited reason by a wide margin. Shipping, taxes and fees appearing at the final step. Show the full cost early — on the product page or in the basket — even if the number is less attractive. People abandon over surprise more than over price.
Forced account creation. Guest checkout should always be available. Offer account creation after the purchase, when the customer has a reason to want order tracking.
Too many steps and fields. Every field costs completions. Address lookup rather than manual entry, no separate billing address unless requested, no phone number unless you genuinely need it.
Limited payment options. Digital wallets in particular reduce friction dramatically on mobile, where typing card details is the main barrier.
No progress indication. People abandon when they cannot see how much further it goes.
2. Product pages
The page where the decision is made, and usually the page with the least attention per visit.
Images. Multiple angles, scale reference, zoom, and the product in use. For anything where fit or size matters, this is the single highest-impact element.
Information completeness. Dimensions, materials, compatibility, what is included. Every unanswered question is either a support contact or an abandonment, and most are abandonments.
Delivery and returns, visible on the page. When will it arrive, what does it cost, what happens if it is wrong. Buyers check this before adding to basket, and making them hunt for it loses them.
Reviews. Genuine reviews, including critical ones. A product with only five-star reviews reads as filtered, and buyers specifically seek out the negative ones.
Stock and urgency, honestly. Real stock information helps. Manufactured countdown timers are recognised and damage trust.
3. Site speed
The relationship between load time and conversion is well established and the effect is large on mobile.
The usual causes in ecommerce are specific: unoptimised product images, a large number of third-party tags, review and personalisation widgets loading synchronously, and heavy theme JavaScript.
Images are almost always the biggest item. Modern formats, correct sizing per viewport, lazy loading below the fold and never on the main product image. See Core Web Vitals.
4. Site search
Visitors who use site search convert at several times the rate of those who do not, and search is frequently the worst-performing part of an ecommerce site.
Check three things. Does it handle misspellings and synonyms. Does it return sensible results for your top hundred search terms — read the search logs, which are the most direct statement of intent you will ever get from customers. And does it return anything useful for zero-result searches, which are both a conversion loss and a product gap signal.
Improving search is unglamorous and routinely produces larger gains than redesigning the homepage.
5. Trust
Particularly significant for unfamiliar brands.
A visible physical address and phone number. Clear returns policy stated in plain language. Real reviews. Secure payment indicators. An about page describing actual people. None of these are conversion tactics exactly; collectively they determine whether a first-time visitor is willing to enter a card number.
6. Mobile specifically
Mobile usually carries most of the traffic and converts at roughly half the desktop rate. Some of that gap is behavioural — people browse on mobile and buy on desktop — and a meaningful part is fixable.
The common fixable issues: tap targets too small, forms requiring the wrong keyboard type, horizontal scrolling, sticky elements covering the add-to-basket button, and digital wallets not offered.
Test on an actual mid-range phone on a normal connection, not in a desktop browser resized.
Testing, with the traffic reality
A/B testing needs volume. Detecting a 10% relative improvement on a 2% baseline conversion rate requires tens of thousands of visitors per variant.
Below roughly 1,000 transactions a month, most tests will not resolve. Use qualitative methods instead — session recordings, checkout funnel analysis, exit surveys and user testing — and make changes based on evidence of friction rather than on statistical proof.
Where you do test, test big changes, run for at least two full weeks, and do not stop early because the result looks good.
Measuring the right thing
Conversion rate alone can mislead. A change that raises conversion and lowers average order value may reduce revenue.
Track revenue per visitor as the primary measure, with conversion rate and average order value underneath it. And watch returns: a change that increases orders and increases returns proportionately has achieved nothing except more handling cost.
Segment before you optimise
A single site-wide conversion rate hides almost everything useful. Four splits change what you work on.
By device. Mobile and desktop frequently differ by a factor of two. If mobile carries 70% of traffic and converts at half the rate, the mobile experience is the whole project.
By traffic source. Branded search, paid acquisition, email and social arrive with completely different intent. A poor blended rate is often one channel delivering unqualified visitors rather than a site problem.
By new versus returning. Returning visitors convert far better. If new-visitor conversion is very low, the issue is trust and clarity rather than checkout.
By landing page. Visitors landing on a product page behave differently from those landing on a category page or the homepage.
The exercise usually relocates the problem. A site that appears to have a conversion issue frequently has a mobile issue, or a paid traffic quality issue, and optimising the desktop checkout would not have touched it.
Post-purchase, which is part of conversion
Treating the transaction as the end point leaves the most profitable part of ecommerce unattended.
The order confirmation is the most-opened email you will ever send and is usually a bare receipt. Delivery expectations stated clearly reduce support contacts substantially. A straightforward returns process, counterintuitively, increases purchases, because it removes the risk that stopped people buying.
And the second purchase costs a fraction of the first. A simple sequence after delivery — did it arrive, is it what you expected, here is something related — produces repeat revenue at a cost no acquisition channel can match. Most sites spend everything on getting the first order and nothing on the second.
A first-week checklist
For a store that has never had a conversion review, the following finds most of the obvious losses in a few days.
Complete a purchase yourself on a mid-range phone, on a normal connection, as a new customer with no saved details. Note every point of hesitation. Check the full cost is visible before the final step. Confirm guest checkout exists and digital wallets are offered. Read the last hundred site searches and see how many return nothing useful. Look at the checkout funnel and find the step with the largest drop.
None of this requires a testing tool or a developer, and it usually produces a list of five concrete fixes worth more than a quarter of button tests.
What to do about the ones who leave
Even a well-optimised store loses most visitors and a majority of filled baskets. Recovering a share of that is a separate discipline from preventing it, and both are worth running.
The prevention work above reduces the volume; a cart abandonment sequence recovers a portion of what remains. Prevention is worth more per hour invested, which is why it belongs first, but the recovery sequence is largely automated once built.
Frequently asked questions
What is a good ecommerce conversion rate?
It varies widely by sector, price point and traffic source. Comparing against your own trend and segmenting by device and channel is more useful than any published benchmark.
Should checkout be one page or several?
Either works. What matters is the number of fields and the clarity of progress, not the number of pages. A single page with thirty fields is worse than three pages with five each.
Do trust badges increase conversion?
Recognised payment and security indicators help modestly. Generic badges with no recognised meaning add clutter and sometimes reduce trust by drawing attention to the question.
How much does site speed affect sales?
Substantially on mobile, where the effect of each additional second is well documented. It is also among the cheapest improvements available on most sites.
