Your checkout is fine. Your catalogue is what loses the sale

Checkout abandonment is the most-optimised number in ecommerce, because it is the easiest to see. Every platform reports it, every consultant has an opinion on it, and there is a well-known list of fixes.

Meanwhile the larger loss happens earlier and is mostly unmeasured: people who never reach checkout because they could not find, compare or believe in the thing they came for.

Why the earlier stage is invisible

Checkout is a funnel with named steps, so drop-off between them is obvious. Browse is not a funnel. It is someone landing, searching, filtering, opening two products, going back, and leaving. Nothing in that sequence looks like a failure, so nothing gets flagged.

The result is that a catalogue problem shows up as “traffic is not converting”, which is a sentence that sends teams to the ad spend rather than the product pages.

Three structural causes, in the order they usually bite

Search that matches strings. A shopper types a phrase that describes what they want, not the words in your product titles. If search matches literally, everything not phrased your way returns nothing, and a no-results page is where sessions end. This is the single most common one, and it is invisible unless you are reading your own site-search logs — which most teams are not.

Filters that do not match how people choose. Filters usually mirror how the catalogue is organised, which mirrors how it was imported, which mirrors a supplier’s spreadsheet. Shoppers filter by the constraint they actually have. When those disagree, the shopper does the filtering by hand, opening products one at a time, and most of them stop before they find it.

Variant handling that hides availability. A product in six sizes where two are out of stock should say so before the shopper picks one. Systems that resolve availability only at add-to-cart convert interest into a dead end at the exact moment interest peaked.

Speed matters, but not where people spend it

Speed work concentrates on the homepage, because that is where the score is measured. The pages that decide the sale are the category page with forty products and the product page with a variant matrix, and both are heavier and less optimised.

A useful reframing: measure the time from landing to first product page, and from product page to add-to-cart, on a mid-range phone on a normal connection. Those two numbers describe the actual journey. A fast homepage in front of a slow category page is a well-scored site that still loses.

What to look at before rebuilding anything

Three things, all already available.

Site-search queries returning zero results, ranked by frequency. This is the highest-value list in ecommerce and it costs nothing to read. Every entry is someone who wanted something and was told you do not have it.

Category pages with high entry and high exit — shoppers arrived, looked, and left without opening a product. That is a browse problem, not a product problem.

Add-to-cart failures by reason. If out-of-stock is a meaningful share, availability is being surfaced too late.

If those three come back healthy, the catalogue is not your constraint and checkout work is the right call after all. Knowing which it is beats guessing, and none of it requires new instrumentation.

We built the ecommerce platform demo around the browse stage rather than the checkout stage, because that is where we kept finding the loss. The demo page shows the search and variant behaviour, with a live storefront to click through.

If the catalogue itself lives somewhere else — an ERP, a supplier feed, a marketplace — then keeping it accurate in both places is the real project, and that is an integrations scope rather than a storefront one.