An online retailer moved its category pages from 'poor' to 'good' without a redesign

Representative examplesEvery case study in this library is an illustrative composite of the kind of engagement we deliver — written to show our method and standards, not to name clients.

Where they started

The client is a mid-size online retailer whose category pages carry most of its non-brand search traffic — and every one of them was failing Core Web Vitals on mobile. The storefront runs on a theme shared with checkout-adjacent templates, so careless changes anywhere risk the till. Merchandising apps add features on request, injecting scripts nobody tracks, and the trading calendar means fixes only get honestly measured during seasonal peaks, when the field data finally has enough users to speak.

What it was costing

CrUX showed category pages failing all three Core Web Vitals on mobile — the pages that carried 60% of non-brand entrances. A previous agency had proposed a full redesign; the retailer wanted to know what was actually slow first.

What they could see

• Field data marked every category page poor on all three Core Web Vitals for mobile users.
• A countdown-timer app the merchandising team loved slowed those pages on its own, visibly in the field data.
• Nobody could list which third-party tags ran on a category page, or who had asked for each.
• Lighthouse runs looked fine in the office while CrUX kept failing — the two scoreboards contradicted each other weekly.

The constraints we worked inside

• The theme was shared with checkout-adjacent templates — changes had to be surgical, per template.
• Merchandising apps injected scripts nobody owned; the fix had to include governance, not just cleanup.
• Seasonal traffic peaks made the field-data window the only honest scoreboard.

What had been tried before

The previous agency proposed a full storefront redesign with performance listed among the headline benefits.

It priced the problem without diagnosing it; the board would not fund a six-figure rebuild to fix what might be three templates and two scripts.

An in-house developer installed a one-click optimization app that minified and deferred everything sitewide.

The global deferral broke analytics on a checkout-adjacent template; it was reverted within a day, and speed work lost internal trust for a year.

What we proposed

We proposed diagnosing before touching anything: field data first, lab tools second. CrUX and real-user monitoring showed the largest category image and two third-party scripts as the actual causes of the failing vitals — layout shift, the lab's favorite complaint, was secondary. The fixes were surgical and per-template: responsive image variants with priority hints for the hero slot, and every third-party tag moved under a consent-gated loader with a named owner in a register. Each fix waited for the rolling 28-day CrUX window to confirm it against what real phones experienced.

Just as important is what we ruled out, and why:

• The full redesign the previous agency proposed — The slowness traced to one image path and two scripts, not the design; a rebuild would re-buy the same failures with new templates nobody had tested.
• Deferring every script and image aggressively sitewide — Deferred heroes are why the largest category image painted late; pushing delay further down the page moves the failure rather than removing it.

How the work ran

Delivered by the growth pod — technical SEO specialist + frontend engineer over 6 weeks, with working increments reviewed with the client every week.

The stack, and the reasoning

CrUX / Search Console

Seasonal peaks make field data the only honest scoreboard; CrUX reports what real phones experienced on the shared theme, free, without lab-run optimism.

RUM monitoring

CrUX samples too coarsely to name culprits; RUM tied the slow experiences to the specific tags and image requests firing on category templates.

Tag governance

Cleanup without ownership re-accumulates; a consent-gated loader plus a register naming every tag's requester made future additions a decision instead of drift.

Image pipeline

The LCP failure was one oversized original served to every viewport; responsive variants with priority hints fixed it without template surgery on shared code.

What went wrong

How we worked together

Cadence

A Monday 30-minute call with the ecommerce manager and the theme's developer, plus a fortnightly CrUX readout against the previous window.

Client side

The ecommerce manager owned merchandising app contracts and tag decisions; one developer owned every template change; nobody else touched the storefront during the engagement.

Decisions

Anything touching checkout-adjacent templates needed the developer's sign-off in staging; removing a tag required the ecommerce manager to check it against the register.

They provided

Analytics and RUM access, a staging environment with the shared theme, the app vendor contact for the timer script, and two developer hours weekly.

What changed

The headline: all three core web vitals on category pages (p75), 28-day crux window after final fix — Poor → Good, read from CrUX field data. A second check: category-page entrances from non-brand organic queries, same period at +9%.

Speed stopped being a mystery the team argued about. New app requests now pass a simple question — which tag, which owner, what does it cost the page — before anything is installed, and the register made the answer checkable in a minute. The merchandising team kept its countdown timer, correctly, and the developer stopped dreading deploys to the shared theme. Peak season arrived with pages that had already been measured by the users who mattered.

The result was read from CrUX field data against the pre-engagement baseline over the stated window, with a guardrail check on category-page entrances from non-brand organic queries, same period. Where platform-reported numbers and business outcomes differ, this record says which layer it is quoting.

What they own now

• The tag register with owners, purposes, and page scope for every script
• The consent-gated loader configuration and rules for adding future tags
• A CrUX monitoring sheet with thresholds and a monthly check routine
• The RUM dashboard views separating category templates from the rest
• Image pipeline documentation covering variants, hints, and naming conventions