Leaks at Browsing: Category and Listing Pages

A category page's job is to turn a browser into a shortlist: show the range's shape, let the visitor narrow it by the attributes they actually decide on, and present each product with the facts comparison needs (price, key variant, availability). Browsing leaks when narrowing fails - filters that miss the deciding attributes, sorts that serve the catalogue instead of the shopper, product cards that hide price or stock until clicked - and the visitor leaves owning a blur instead of a shortlist. Fix narrowing first; decoration later.

Watch recordings of category browsing and the behaviour is unmistakable: visitors alternate between scanning cards and reaching for a filter, and the session dies at the moment the filter set cannot express their decision ('waterproof', 'fits wide feet', 'under this total') or returns a count that feels like a verdict (zero results, or four hundred). The deciding attributes come from the products' real differentiation and from site search logs, where visitors type the narrowing they could not click. The wayfinding bench covers filters in depth; this guide's point is diagnostic: a browsing-stage drop with healthy product pages is usually a narrowing failure wearing a design problem's coat.

Each card is a tiny product page making one promise: click me and I might be the one. Cards leak when comparison requires clicking: price ranges hidden, the popular variant out of stock but not stated, ratings absent where the category runs on them, images inconsistent so size and quality cannot be judged side by side. Standardise the card's facts across the category (same image treatment, same fact set, same order) so the eye can compare at scanning speed. A card that requires a visit to answer its basic questions taxes the whole category's traffic, not just its own clicks.

The default sort decides what most browsers see, because most never change it. A default that fronts out-of-stock lines, discontinued colours or the catalogue's oldest stock leaks steadily and invisibly - the step rate just looks 'a bit low for the category'. Audit defaults quarterly against merchandising reality: best-fitting, available, current range first. Similarly, pagination that resets filters, or a filter state that vanishes on the back button, converts a productive browsing session into a repeated chore; browsers forgive it once.

Zero-result filter combinations, empty subcategories left in the menu, and search-within-category that cannot spell are browsing dead ends: each tells the visitor the shop does not contain their answer, usually falsely. Every dead end needs a productive exit - nearest matches, the filter relaxed with the relaxation stated, a route to the parent category with context kept. Instrument zero-result events; they are the browsing stage's purest leak signal, each one a stated requirement the shop failed on the record.

Judge browsing not by category pageviews but by what browsing produces: product views per session, shortlist behaviour (multiple product views, saves, comparisons), and the add rate of products reached from browsing versus from search. A redesign that lifts category clicks while lowering downstream adds has made browsing busier, not better. Read the stage's step rate together with the next stage's, always; stages pass visitors to each other, and a leak repair that merely moves the drop downstream is bookkeeping, not repair.

Category pages are often treated as plumbing and left to defaults, when browsing is where merchandising judgement earns its money: which products front the category and why, what 'new' and 'back soon' mean and how long they may say it, how ranges are grouped so a browser learns the shop's logic in one screen. A category arranged by the warehouse's logic (by supplier, by internal code order) leaks browsers who think in outcomes. The arrangement audit is cheap: for each main category, write the sentence a browser should conclude from the first screenful ('this shop has three price levels and the middle one is the range's heart') and check the page says it without words.