SPEED & CORE WEB VITALS · スピード最適化とCore Web Vitals

Japan Has Fast Infrastructure. Japanese Sites Are Often Still Slow.

High-speed fibre and 5G coverage make Japan one of the world's best-connected markets. Japanese websites are still frequently slow. Hero images are oversized. CJK web font files are unsubsetted and weigh hundreds of kilobytes. Legacy JavaScript accumulated over years of incremental development blocks the main thread on every page load. Fast infrastructure does not offset a slow page. We audit the actual causes and fix them.

Core Web Vitals audits completed for Japanese-market sites in both Japanese and English · CJK web font subsetting as a standard deliverable on every Japan-market project · Field data from Google Search Console Japan CrUX used throughout
This is for you if

Three types of businesses operating in Japan come to us with a speed and Core Web Vitals brief.

Your site was built for the domestic market and has grown incrementally for years. A large number of images, multiple font families including Japanese-character web fonts, accumulated JavaScript from tag managers, marketing tools, and legacy features, and a CMS that has been modified repeatedly. The site works, but the performance numbers in Google Search Console's Core Web Vitals report are poor. You have not prioritised it because the site generates traffic and leads. The question is how much more it would generate if the performance problems were fixed.

You have an international-facing English site that was built separately from the main Japanese site, often at a different time and with different technical decisions. The English site is slow on mobile and performing poorly in CrUX field data for target markets like the US, UK, and Australia. The Japanese site has a local CDN and more aggressive optimisation. The English site does not. You need the international site to perform to the same standard.

Your Japan-localised site was designed to Japanese conventions and includes Japanese copy. The performance numbers are poor. CJK web fonts are loading unsubsetted. The mobile experience on mid-range Android devices, which represent a significant share of the Japanese market, is slow. The content and design are correct. The technical implementation is not.

What's broken

Four performance problems that appear consistently in Japan-market site audits.

CJK Web Fonts Are Loading Unsubsetted

Japanese web fonts are large. A single Noto Sans JP or Hiragino-fallback web font file can weigh 2MB to 10MB uncompressed, covering the full Unicode CJK range. When a site loads a CJK web font without subsetting to the characters actually used on the page, it is downloading a font file orders of magnitude larger than necessary. Font subsetting, combined with the unicode-range CSS descriptor to load only the character ranges needed per page, is the single highest-impact performance intervention available on most Japanese sites. It directly improves both Largest Contentful Paint and First Contentful Paint.

Hero Images Are Large, Unoptimised, and Uncompressed

Japanese site designs frequently use large, high-quality photography, particularly in the hospitality, real estate, and consumer goods sectors. These images are often served as large JPEGs without conversion to WebP or AVIF, without proper compression, and without preload hints that would allow the browser to fetch them at high priority early in the page load. A 1.2MB hero image served without modern encoding is a reliable LCP failure, regardless of the speed of the underlying network.

Legacy JavaScript Has Accumulated Over Years of Development

Many Japanese sites that have been actively developed for five or more years carry JavaScript debt. Tag manager containers with outdated or unused tags, jQuery plus a more recent framework loaded simultaneously, tracking scripts for platforms that are no longer in use, and feature code from discontinued functionality. This accumulated JavaScript loads on every page, blocks the main thread, and directly degrades Interaction to Next Paint. The fix requires a structured audit, not a search for a plugin.

No CDN or a CDN Without Correct Configuration

Some smaller Japanese sites are still served directly from origin servers. Others have a CDN in place but with default cache settings that are not appropriate for the site's asset types. For international-facing Japanese sites, CDN edge node selection for the target market, such as US or European points of presence, is often not configured. The result is avoidable latency for every international visitor.

What we engineer

Six technical deliverables, with CJK font subsetting as a standard component on every Japan-market project.

Full Performance Audit and CrUX Field Data Analysis

We pull field data from Google Search Console's Core Web Vitals report and CrUX for your domain. For Japan-market sites, we look at field data segmented by device type. Japanese mobile users on mid-range Android devices experience performance differently from desktop users. The audit identifies which pages are failing, which metrics are the cause, and what the root causes are. The output is a prioritised fix list.

CJK Web Font Subsetting and Loading Strategy

We audit every web font loaded on the site. For CJK fonts, we implement subsetting using the unicode-range CSS descriptor to load only the character ranges required for each page. Where Google Fonts is in use for CJK fonts, we migrate to self-hosted subsetted files to eliminate the external request. Font-display strategy is reviewed and optimised to prevent invisible text during font swap. For most Japanese sites, this single intervention produces the largest measurable improvement in LCP and FCP.

Image Optimisation (WebP, AVIF, and Lazy Loading)

All significant images are converted to WebP with AVIF where browser support allows, with correct fallbacks. Hero and above-the-fold images receive preload hints. Images below the fold are lazy-loaded. Width and height attributes are declared to prevent layout shift from image loading. For sites with large image catalogues, this is implemented at the asset pipeline level.

JavaScript Bundle Audit and Legacy Script Removal

We audit the JavaScript loaded on critical pages. Unused scripts are removed. Tag manager containers are reviewed and outdated tags cleaned out. Duplicate framework loading, such as jQuery and a modern framework coexisting, is identified and resolved where possible. Third-party scripts are moved to async or deferred loading. Where a script is not providing measurable value, we flag it for removal.

CDN Implementation and Cache Configuration

For sites not on a CDN, we implement CDN delivery. For sites already using a CDN, we audit the cache configuration and edge node selection. For international-facing Japanese sites, we confirm that edge nodes in the target markets are being used correctly. Static asset cache headers are reviewed and corrected.

Before and After Scores and Monitoring Setup

Before and after scores are documented for both lab tools and field data. Monitoring is configured in Google Search Console. Where appropriate, a lightweight real user monitoring setup is implemented so ongoing performance is visible without requiring a new audit.

What changes

Four outcomes from a well-executed Japan-market performance project.

Before
After
Before Japanese web fonts are large. A single Noto Sans JP or Hiragino-fallback web font file can weigh 2MB to 10MB uncompressed, covering the full Unicode CJK range. When a site loads a CJK web font without subsetting to the characters actually used on the page, it is downloading a font file orders of magnitude larger than necessary. Font subsetting, combined with the unicode-range CSS descriptor to load only the character ranges needed per page, is the single highest-impact performance intervention available on most Japanese sites. It directly improves both Largest Contentful Paint and First Contentful Paint.
After CJK font subsetting is the highest-impact intervention available on most Japanese sites. When a 4MB unsubsetted CJK font file is replaced with a 40KB subsetted font loading only the characters on that page, the browser receives the font data far earlier in the load sequence. This directly reduces First Contentful Paint and Largest Contentful Paint in both lab and field data. For sites where CJK fonts were the primary LCP bottleneck, the score change is substantial.
Before Japanese site designs frequently use large, high-quality photography, particularly in the hospitality, real estate, and consumer goods sectors. These images are often served as large JPEGs without conversion to WebP or AVIF, without proper compression, and without preload hints that would allow the browser to fetch them at high priority early in the page load. A 1.2MB hero image served without modern encoding is a reliable LCP failure, regardless of the speed of the underlying network.
After The performance gap between desktop and mobile on Japanese sites is often wider than the raw network difference would suggest. Mobile CPUs process large JavaScript bundles more slowly than desktop CPUs. When JavaScript is audited and reduced, INP and Time to Interactive improve specifically on mobile devices. Japanese users on mid-range Android hardware, which represents a significant share of mobile browsing in Japan, experience this improvement directly.
Before Many Japanese sites that have been actively developed for five or more years carry JavaScript debt. Tag manager containers with outdated or unused tags, jQuery plus a more recent framework loaded simultaneously, tracking scripts for platforms that are no longer in use, and feature code from discontinued functionality. This accumulated JavaScript loads on every page, blocks the main thread, and directly degrades Interaction to Next Paint. The fix requires a structured audit, not a search for a plugin.
After For Japanese companies whose English-language site is performing poorly in CrUX field data for US, UK, or Australian users, fixing the performance issues removes a disadvantage in international search results. International visitors who previously arrived and left quickly because the page was slow now experience a site that loads correctly for their device and connection.
Before Some smaller Japanese sites are still served directly from origin servers. Others have a CDN in place but with default cache settings that are not appropriate for the site's asset types. For international-facing Japanese sites, CDN edge node selection for the target market, such as US or European points of presence, is often not configured. The result is avoidable latency for every international visitor.
After Without monitoring, performance degradation is discovered at the next audit, often months later. With monitoring in place, new images added without proper compression, new scripts introduced through a tag manager, or font loading regressions are visible in Google Search Console data as they occur. The monitoring configuration we implement gives ongoing visibility without requiring manual checks.
How it works

Four stages. CJK font analysis is part of the audit, not an afterthought.

  1. 01

    Field Data Audit and Font Analysis

    Week 1

    We pull CrUX field data and Google Search Console Core Web Vitals data for the domain. For Japan-market sites, the audit includes a specific CJK font inventory: which fonts are loaded, what size they are, whether they are subsetted, and what loading strategy is in place. We also audit the JavaScript payload and image sizes for critical pages. The output is a prioritised fix list.

  2. 02

    Root Cause Confirmation and Fix Plan

    Week 2

    We confirm root causes in the site code and asset pipeline. For CJK font fixes, we define the subsetting strategy: which unicode ranges to target, whether to self-host or use a font service, and what the font-display setting should be. For JavaScript fixes, we map which scripts can be deferred and which legacy scripts can be removed. The fix plan is agreed before implementation begins.

  3. 03

    Implementation

    Weeks 2–5

    CJK font subsetting is typically the first implementation step for Japanese sites because it produces the largest improvement with the least structural risk. Image optimisation follows. JavaScript audit and reduction is next. CDN configuration and cache headers are reviewed and corrected. Implementation is on a staging environment before production deployment.

  4. 04

    Verification, Documentation, and Monitoring Setup

    Week 5–6

    Lab score improvements are documented immediately after implementation. Field data improvements are tracked over the following four to six weeks as the CrUX 28-day rolling window updates. Before and after scores are provided in a written summary. Monitoring is configured in Google Search Console.

Common questions

FAQ

Why are Japanese web fonts such a significant performance problem?

Japanese websites require fonts that cover thousands of kanji, hiragino, katakana, and kana characters. A full CJK Unicode font file covers over 20,000 characters and weighs several megabytes. English fonts typically cover a few hundred characters and weigh a few dozen kilobytes. When a Japanese site loads an unsubsetted CJK font, it is loading a file that is 50 to 100 times larger than it needs to be for the characters actually displayed on the page. Font subsetting reduces this to only the characters present in the page content, cutting font payload by 95 percent or more in many cases.

What is font subsetting and how does it work?

Font subsetting creates a smaller font file containing only the characters needed for a specific page or page type. Using the CSS unicode-range descriptor, the browser is instructed to load only the font subset that covers the characters present in the page content. A product description page that uses common kanji needs a much smaller subset than a page that covers the full character range. Subsetting is implemented at the font file level, with multiple small files replacing a single large one. The browser only downloads the subsets it needs.

Does Google use field data from Japanese users specifically?

Google's Core Web Vitals field data in Search Console and CrUX is based on real user visits to your site, from whatever country those users are in. For sites primarily serving Japanese users, the field data reflects Japanese user experiences. For international-facing Japanese sites, the field data includes visitors from all markets. Segmenting the field data by device type is important because mobile performance on Japanese mid-range Android devices differs from desktop performance.

Our site is on fast Japanese hosting. Why is it still slow?

Fast server response time reduces Time to First Byte, which is one component of page load performance. It does not offset slow-loading assets. A 4MB unsubsetted CJK font, a 1.5MB uncompressed hero image, and 300KB of synchronous JavaScript will produce slow Core Web Vitals scores on fast hosting and on a fast network. The bottleneck is the asset payload and loading strategy, not the server. We audit at the asset and code level, not the infrastructure level.

Can you fix Core Web Vitals on Japanese CMS platforms like Movable Type or EC-CUBE?

We have worked on Movable Type, EC-CUBE, WordPress with Japanese-specific theme setups, and custom-built Japanese CMS platforms. The underlying performance fixes, font subsetting, image conversion, JavaScript auditing, and CDN configuration, apply across platforms. Some platforms impose constraints on templating and asset management that affect implementation approach. We identify those constraints during the audit stage and scope implementation accordingly.

Our team

The people behind the work

Not a black box. Real specialists you can call, with their names on the work.

Niraj Raut

Niraj Raut

Founder — Ecommerce SEO
Keshab Joshi

Keshab Joshi

PPC Expert
Hawrry Bhattarai

Hawrry Bhattarai

Google Ads Expert
Arogya Rijal

Arogya Rijal

SaaS SEO Expert
Start here

Fast Fibre Does Not Fix a Slow Page

Japan's infrastructure is fast. Many Japanese websites are not. The causes are consistent: unsubsetted CJK fonts, uncompressed hero images, years of accumulated JavaScript, and loading strategies that have not been reviewed since the site was first built. Ignited Nepal starts with a field data audit that identifies what is actually failing for real users before recommending any implementation work.

Ignited Nepal is a Growth Engineering Company based in Nepal, working with businesses that need technical performance to match their growth goals.