AI Briefing
KO

SEO-Driven Development in Practice: The Journey to a 99% Google-Certified 'Good URL' Rate

·2025.11.04 17:24

Key point

Through server and client optimization, the ratio of good URLs in GSC was pushed up to 99%.

Details

With SEO-driven development set as a development goal, the team continuously monitored Google Search Console (GSC) and Core Web Vitals, removing slow URLs one by one to achieve 99% Good URLs.

At first, translation key data accounted for 90% of HTML size, causing a major bottleneck, but removing it led to a 66% reduction in HTML size, a 16% improvement in server latency, a 10% improvement in FCP, and a 21% improvement in LCP.

Afterward, bottlenecks continued to be cleaned up.

  • Slowness on the first request for logged-in users was resolved by adding an index to the session table, improving TTFB by 70%
  • API calls in SSR were parallelized using Promise.all and similar methods, improving TTFB by 20%
  • Public API calls that had been going outward were switched to Internal API, reducing API latency by 32%
  • CloudFront was introduced at the front, and HTTP/3 was applied, further reducing LCP by 10% and FCP by 7%

In late January 2025, moving some page logic from Client Side to SSR converted about 100,000 URLs into 'Good URLs' on mobile alone, achieving 99% on mobile in February 2025. The remaining task was the roughly 40,000+ URLs needing improvement on desktop, and applying 3rd-party library bundle optimization and CLS improvements to this brought _app.js down from 759KB to 489KB (35% reduction), and achieved CLS under 0.1 on desktop detail pages, wrapping things up.

Ultimately, the core of the achievement lies not in the numbers themselves, but in the process of using data to find the causes of slowness, persuading and collaborating with colleagues, and improving both the server and the browser at the same time.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.