Your navigation is losing buyers before they see a product.

Buyers can't find what they came for.

Across the 1,284 pages Levri has analysed, the median store makes people work too hard to find the thing they already want. Category trees are bloated, the search bar is hidden behind a magnifying-glass icon, and filters reset the moment you turn the page. The visitor who arrived ready to buy clicks back to Google instead.

This isn't a fringe problem. Baymard Institute's 2025 navigation benchmark - 16,000+ UX elements across 180 leading US and European stores - found 67% have "mediocre" to "poor" navigation UX. Not small sites. Leading ones.

And the cost is measurable. Visitors who use internal search convert at 4.63% versus 2.77% for those who browse - a 1.8x gap that holds across fashion, electronics, and grocery alike. When you bury the tool that converts best, you're taxing your highest-intent buyers. Here are the six patterns we see most, and what to ship.

The six patterns that move discovery.

1. The buried search bar - your highest-intent visitors can't find the tool that converts them 1.8x better.
2. Zero-result dead ends - 10-15% of searches return nothing, and most stores just shrug.
3. Bloated category trees - long flat menus that nobody can scan.
4. No location cues - 95% of sites never show people where they are.
5. Filters that forget - selections reset on pagination, taking the user's work with them.
6. Slow autocomplete - search that lags past a second loses the room.

Most stores we scan are running several of these at once. The fixes are mostly configuration and markup, not a rebuild.

#1 - The buried search bar.

Between 30% and 50% of ecommerce visitors will use internal search if they can find it, and those searchers convert at 1.8x the rate of browsers - on Amazon, internal search converts at ~12%, six times the sitewide average. Yet most stores hide search behind an icon, shrink it into a corner, or let a promo banner steal its space. Once they find the product, the next job is the page itself converting - that's product page conversion.

What we see on scans

• A search icon with no visible text field on desktop.
• The search bar below the fold on mobile.
• No search at all on category or product pages.
• Search competing with nav or promo bars for room.

What to ship

• Make search a full-width, always-visible field on desktop - not a collapsible icon.
• On mobile, put it in the sticky header so it follows the scroll.
• Use placeholder text that signals scope: "Search 2,400+ products...".
• Show search on every page, not just the homepage.

Typical lift: +5% to +12% conversion for search users. Impact: high - it surfaces your highest-intent buyers.

#2 - Zero-result dead ends.

The average store returns "no results found" for 10-15% of queries; well-tuned engines push that below 5%. Every zero-result page is an exit ramp - the visitor told you exactly what they wanted, and the store said "we don't have that," often when it does.

What we see on scans

• A generic "no results" page with no alternatives offered.
• No typo tolerance or fuzzy matching.
• No synonym handling ("trainers" vs "sneakers", "sofa" vs "couch").
• No popular-product or category suggestions on the dead-end page.

What to ship

• Add fuzzy matching and typo tolerance - "bleu dress" should still return blue dresses.
• Map synonyms for your catalogue's key terms.
• Show popular products or related categories on zero-result pages.
• Log zero-result queries weekly and map them to products or stock gaps.

Typical lift: +2% to +6% site-wide conversion. Impact: high - every recovered search is a buyer you nearly lost.

#3 - Bloated category trees.

A navigation menu with 40 ungrouped links isn't a menu, it's a wall. Around 60% of stores never chunk categories into scannable groups, so the visitor has to read every option instead of pattern-matching to the right cluster. The cognitive cost lands at the worst moment - the first few seconds, before they've committed to anything.

What we see on scans

• Flat top-level menus with 20+ ungrouped links.
• Categories ordered by internal logic, not by what sells.
• Mega-menus with no headings, so nothing is scannable.
• The same depth everywhere - no sense of what's primary.

What to ship

• Group categories under 4-7 scannable headings; lead with what people actually buy.
• Cap the top level - push the long tail one level down, not into the face of every visitor.
• Add headings and visual grouping to mega-menus so the eye can jump, not read.
• Order by demand (sales data), not alphabetically or by SKU count.

Typical lift: +2% to +5% on category-entry rate. Impact: medium - structural, but it compounds on every page that shows the nav.

#4 - No location cues.

This one is nearly universal: 95% of sites don't highlight the visitor's current position in the main nav. Someone clicks Women's > Dresses > Midi and the nav looks identical to the homepage - context gone. On mobile it's worse, where small screens strip the cues and the back button becomes the only navigation tool, often sending people to Google instead of up your hierarchy.

What we see on scans

• No breadcrumbs on product or category pages.
• Main nav that doesn't mark the active section.
• Mobile category pages with no visible path back to the parent.
• "Back" that exits the site rather than moving up a level.

What to ship

• Add breadcrumbs to every product and category page - essential, not optional.
• Style the active top-level category distinctly (bold, underline, colour).
• On mobile, show a sticky breadcrumb or a "Back to [Parent]" link.
• Make in-page "back" move up the hierarchy, not to the browser's last page.

Typical lift: +2% to +5% pages-per-session, with lower bounce. Impact: medium - cheap to add, cumulative on session depth.

#5 - Filters that forget.

The visitor applies size M, blue, under $50, scrolls, clicks page two - and every filter resets. All their work, gone. They start over or, more often, leave. Filter persistence is one of the most commonly broken patterns in ecommerce, and it breaks exactly where intent is highest.

What we see on scans

• Filters reset on pagination, sorting, or back-navigation.
• Applied filters not shown above the grid.
• No "clear all" or per-filter removal.
• No result counts, so you can't tell if a filter returns 2 products or 200.
• Full-page reloads instead of instant updates.

What to ship

• Persist filters across pagination, sorting, and back-navigation.
• Show applied filters as removable chips above the grid.
• Display the result count next to each option before it's selected.
• Use instant (AJAX) filtering - no full-page reloads.
• Remember selections when people return to the category in the same session.

Typical lift: +4% to +9% category-to-product conversion. Impact: high - filters are the main tool purchase-ready visitors use to narrow down.

#6 - Slow or missing autocomplete.

Search that takes more than a second to respond raises bounce by 47% and cuts conversion probability by 73%; the benchmark is under 500ms, and the best stores answer in under 200ms. Most either offer no autocomplete or serve generic suggestions that ignore real inventory - out-of-stock items next to bestsellers, no category context. By the time results appear, trust is gone.

What we see on scans

• No autocomplete on the search bar.
• Suggestions that take 1-2 seconds to appear.
• Generic suggestions unrelated to the query.
• No thumbnails, prices, or stock status in results.
• Out-of-stock items surfaced as suggestions.

What to ship

• Make autocomplete respond under 500ms - ideally under 200ms.
• Show thumbnails, prices, and stock status in the dropdown.
• Include category suggestions alongside products ("Women's > Knitwear?").
• Suppress out-of-stock items unless you offer back-in-stock alerts.
• Edge-cache the 100 most common queries so speed is guaranteed.

Typical lift: +3% to +7% search-to-purchase. Impact: high - it's the first interaction with your search; if it's slow, nobody trusts the rest.

How Levri spots all six in 60 seconds.

Levri reads your navigation structure, search, category hierarchy, and filter behaviour in a single pass, then ranks each issue by revenue impact so you fix the highest-value leak first.

You paste the URL, and you get your fixes - a ranked list, each issue priced in $/mo, with a written hypothesis, a variant-B suggestion, and an expected lift range. No install, no tracking script, no "connect your analytics."

Fix these first.

In the order we'd ship - each is hours, not a rebuild:

1. Make the search bar visible (surfaces your highest-intent buyers).
2. Fix zero-result pages (stop losing people to dead ends).
3. Add breadcrumbs and active-nav highlighting (low effort, immediate depth).
4. Make autocomplete fast and relevant (under 200ms).
5. Persist filters (protect the work your most engaged visitors did).
6. Restructure categories (biggest change, compounds everywhere).

Once buyers reach a product, make sure the page closes the sale - see product page conversion - and that the homepage points them the right way to begin with, covered in why your hero section is killing your CTR. Ship three of these and measure by Friday; if discovery doesn't move, you're fixing the wrong layer, which is exactly what Levri is built to diagnose.