Ecommerce is unusually measurable, which makes it one of the few areas where you can find the leak rather than guess at it.
The discipline is to look before changing anything.
Start in the analytics
Before redesigning anything, build the funnel: sessions → product views → add to cart → checkout started → checkout completed.
The drop between each pair tells you where the problem is. A store losing people between product view and add to cart has a product page problem. One losing them between checkout started and completed has a checkout problem. These need completely different fixes, and working on the wrong one is how redesigns produce no improvement.
Then segment by device. Mobile and desktop frequently behave so differently that the aggregate figure describes neither.
Leak one: product page speed on mobile
The most commonly underestimated factor.
Your product pages are probably fast on your laptop on office wifi. Your customers are on a mid-range phone on mobile data, and a page taking six seconds there loses a meaningful share before it renders.
The usual causes, in order:
Images. Uncompressed, served at desktop dimensions to phones, in outdated formats. Serve WebP or AVIF, size responsively, set explicit dimensions, lazy-load below the fold — but never lazy-load the main product image, which is the thing you are trying to show.
Third-party scripts. Analytics, chat widgets, review widgets, heatmaps, ad pixels, each loading more scripts. Audit them annually; there is always something installed for a campaign two years ago.
Apps and plugins. On Shopify and WooCommerce, each installed app typically injects scripts and styles on every page, whether or not that page uses it.
Test with a throttled connection on a mid-range device, not on your own machine.
Leak two: checkout friction
The place with the clearest arithmetic, because everyone here has already decided to buy.
Show the total early. Unexpected delivery cost at the final step is consistently the top cited reason for abandonment. Show it on the product page or in the cart, not after they have entered their address.
Allow guest checkout. Forced account creation loses real revenue for a benefit — a customer record — you can obtain afterwards by offering to save details post-purchase.
Cut the fields. Every field loses a percentage. Do you need a company name? A second address line as a required field? A phone number?
Use address autocomplete. Faster, and it reduces failed deliveries.
Offer the payment methods people actually use locally. This varies enormously by market, and a store offering only international cards in a market where everyone uses a local wallet is leaving money on the table.
Make errors recoverable. A declined card should not clear the form. A validation error should say precisely what to fix.
Test on a real phone. Including the awkward cases: a long address, an international card, a failed payment, going back a step.
Leak three: missing information
People do not abandon because your product is wrong. They abandon because they were unsure and leaving was easier than asking.
The hesitation questions are predictable:
- Will it fit, or work with what I have?
- When will it arrive?
- What if it is wrong — can I return it, and who pays?
- Is this the right variant?
- What are the materials, dimensions, specifications?
Answer these on the product page, not on a policy page three clicks away. Delivery estimate near the price. Returns policy summarised near the buy button. Sizing guidance inline rather than in a modal nobody opens.
An AI assistant that answers these instantly, from your real product data, catches a share of the people who would otherwise leave — which is the main commercial argument for ecommerce chat.
Leak four: trust gaps
First-time visitors who have never heard of you are assessing risk.
What helps, in rough order:
- Real reviews, including some that are not five stars. An unbroken wall of perfect reviews reads as fake.
- Clear returns policy, stated plainly rather than buried in legal text.
- Visible contact details — a phone number and address signal a real business.
- Recognisable payment logos at the point of payment.
- Professional product photography. Supplier stock images shared with fifty other stores undermine credibility.
- Honest delivery estimates. Over-promising creates a worse problem later.
Leak five: thin product content
Relevant to both conversion and search.
Manufacturer-supplied descriptions appear on every store selling the item, which means they rank for nothing and differentiate nothing. Products with specific, original descriptions covering real attributes and real use cases convert better and rank better.
At catalogue scale this is where AI genuinely helps — generating descriptions from real product attributes, reviewed for accuracy. The output needs checking; the alternative is thousands of SKUs with identical text.
What matters less than people think
Visual redesign. A prettier store that is equally slow and has the same checkout converts roughly the same. Redesigns aimed at a measured problem work; redesigns aimed at looking modern often do not.
Adding more apps. Each one adds load time. The net effect of a conversion app that slows every page is frequently negative.
Aggressive popups. They capture some emails and cost some sales. Measure the net rather than assuming.
Cart recovery emails. Useful, and a patch over a checkout problem rather than a fix for one. Fix the checkout first.
A sensible order of work
- Build the funnel from analytics and find the biggest drop, split by device
- Measure product page speed on a mid-range phone with throttled data
- Fix images and audit third-party scripts — usually the largest single gain
- Walk your own checkout on a phone, buying something real with a real card
- Remove every field and step that is not necessary
- Show total cost and delivery estimate as early as possible
- Answer the hesitation questions on the product page
- Re-measure and repeat on the next biggest leak
One leak at a time, measured each time. Changing five things at once means learning nothing about which worked.
If you want your funnel analysed to find where the money is actually going, get in touch.
