How a wooden-toy workshop fixed a crawling nightmare and doubled pre-Christmas sales

Faceted filters spawned 38,000 junk URLs, mobile LCP sat at 6.2s and Google was crawling filter combinations instead of new products.

Zarnitsa hand-turns beech and lime-wood toys in a Varna workshop and sells about 640 of them online. Customers loved the filters — shop by age, wood, colour, price. Googlebot loved them too, and spent its visits wandering every combination instead of the new Christmas range.

On phones, the page every parent landed on took over six seconds to show its hero picture. With roughly 70% of Q4 sessions on mobile, that cost sales in the weeks that matter most.

Six findings from a full crawl, 90 days of server logs and field Core Web Vitals data. Flip the tree to see what pruning does.

1. 01Faceted navigationHigh

   Four filters, infinite URLs

   Age × material × colour × price, plus ?sort= and ?view=, combined freely. Every combination was a crawlable, indexable URL with no canonical — 38,000 near-duplicates of 640 products.

2. 02Crawl budgetHigh

   Googlebot lost in the filters

   Server logs showed 86% of Googlebot hits landing on parameter URLs. New Christmas lines took 19 days on average to be discovered.

3. 03Core Web VitalsHigh

   A 6.2s hero on mobile

   The LCP element was a 2.4 MB PNG of the workshop, lazy-loaded by mistake and queued behind a render-blocking carousel script and two web-font files.

4. 04Image optimisationMedium

   Product photos straight off the camera

   3,900 product images averaged 1.1 MB, with no width/height attributes (CLS 0.31) and no responsive sizes, so phones downloaded desktop originals.

5. 05Crawl budgetMedium

   Dead toys still in the sitemap

   Discontinued products returned soft 404s; the XML sitemap listed 2,300 filter URLs and 180 retired products.

6. 06Faceted navigationMedium

   Internal links fed the problem

   Filter chips were plain links, so every category page linked out to dozens of parameter URLs and passed equity into them.

The work ran on our Signal plan: technical first, because no amount of new content helps if Google is too busy crawling ?colour=red&sort=price to find it.

Each tactic is a fold in the book: cut away what Google should never see, then make what is left load fast.