Pagespeed Desktop Review 2026 06 16

PageSpeed desktop review — www.assistance.bg — 2026-06-16

Scope: prioritized no-code review of the provided PageSpeed Insights desktop snapshot for https://www.assistance.bg/, with source mapping to the repo surface that owns the live page.

Evidence baseline#

This review is based on the embedded Lighthouse JSON in the provided PageSpeed UI route, extracted with curl from https://pagespeed.web.dev/analysis/https-www-assistance-bg/trv6ytaqof?form_factor=desktop. It is not based on a fresh PageSpeed run.

Lab metrics:

The report is lab-only because CrUX field data is unavailable for both the page URL and origin. Key diagnostic totals: 800 KiB network payload, 1.0 s JavaScript execution, 1.6 s main-thread work, 5 long tasks, and 34 non-composited animated elements.

Source mapping#

The live/source target for this review is apps/www.

Evidence:

• README.md identifies apps/www as the marketing site.
• .mise.toml deploys apps/www/.next-build to Cloudflare Pages project assistance-www with cf:deploy:www.
• apps/www/wrangler.toml declares name = "assistance-www" and pages_build_output_dir = ".next-build".
• apps/www/next.config.mjs is a static Next export with output: 'export' and distDir set from NEXT_DIST_DIR.
• apps/www/public/_headers contains production rules for www.assistance.bg and immutable caching for /_next/static/*.
• The PageSpeed report and live HTML use Next /_next/static/* assets, matching apps/www rather than a Vite /assets/* app.

Earlier uncertainty around an apps/marketing production owner should be treated as stale or ambiguous for this URL. Do not use apps/marketing as the remediation target unless new deployment evidence contradicts the current apps/www mapping.

Prioritized findings#

P1 — Third-party analytics is the largest named desktop cost#

What PageSpeed reports:

• Unused JavaScript: 104 KiB / 160 ms.
• Google tag script: https://www.googletagmanager.com/gtag/js?id=G-3KQWLZJ69Q.
• GTM transfer/main-thread cost: 159,457 bytes / 136.95 ms.

Repo mapping:

• apps/www/app/providers.tsx defers telemetry until interaction, scroll, touch, keydown, or a 3s idle fallback.
• The deferred block mounts TelemetryTagManager, GoogleAnalytics, and PageTelemetry from the shared common package.

Recommended next actions:

1. Verify whether the 3s idle fallback causes analytics to load during Lighthouse collection without real user intent.
2. Audit the GTM container and remove unused tags/triggers.
3. If business requirements allow, keep analytics behind stronger user-intent or consent gates for the initial page load.
4. Prefer a minimal direct gtag setup over GTM if the container remains heavier than its value.

P1 — Landing page JavaScript has avoidable unused/hydration work#

What PageSpeed reports:

• Unused app JavaScript includes https://www.assistance.bg/_next/static/chunks/6258-ff03de593974c853.js.
• JavaScript execution time is 1.0 s.
• Main-thread work is 1.6 s.
• Lighthouse found 5 long tasks.

Repo mapping:

• apps/www/app/page.tsx renders the full homepage route: hero, products, use cases, code examples, features, capabilities, global reach, testimonials, trust indicators, and CTA.
• apps/www/app/providers.tsx globally mounts auth, feature flags, dev toolbar provider, theme, Framer Motion LazyMotion, tooltip provider, command provider/menu, form dialog provider, toaster, telemetry, and DevToolbar.
• apps/www/package.json includes heavy interactive libraries such as framer-motion, @react-three/fiber, @react-three/drei, three, cobe, Recharts, MDX/code tooling, command-menu dependencies, and form/dialog dependencies.
• apps/www/components/Global/GlobalReach.tsx already lazy-loads the 3D globe with ssr: false, but it remains part of the homepage interaction surface.

Recommended next actions:

1. Run the Next bundle analyzer for apps/www and identify the module composition of chunk 6258.
2. Use Chrome Coverage on the verified www.assistance.bg page to identify unused providers and below-the-fold client sections.
3. Move command menu, form dialog rendering, dev tooling, and other non-critical global UI behind explicit user intent where feasible.
4. Keep above-the-fold hero content static/server-rendered and split below-the-fold interactive sections independently.

P2 — Render-blocking and unused CSS are modest but clear wins#

What PageSpeed reports:

• Render-blocking Next CSS: 190 ms.
• Unused CSS: 39 KiB in https://www.assistance.bg/_next/static/css/cbd9d7e6ca22c268.css.

Repo mapping:

• apps/www/app/layout.tsx globally imports @code-hike/mdx/styles, config/code-hike.scss, and ../styles/index.css.
• apps/www/styles/index.css is the broad global style surface for the app.

Recommended next actions:

1. Confirm whether Code Hike CSS is needed on the homepage or only on MDX/code-heavy routes.
2. Keep design tokens and resets global, but route-scope heavy MDX/code/component CSS where possible.
3. Re-check LCP/FCP and render-blocking time after CSS scoping.

P2 — Legacy JavaScript adds secondary compatibility bytes#

What PageSpeed reports:

• Legacy JavaScript savings: 21 KiB.
• Affected chunks include 5908 and 6258.

Repo mapping:

• apps/www/next.config.mjs controls Next transpilation and explicitly transpiles several workspace/shared packages.
• apps/www/package.json pulls in a broad dependency set; legacy output should be attributed by bundle analyzer before changing browser targets.

Recommended next actions:

1. Identify the exact legacy transforms from Lighthouse and bundle output.
2. Confirm the production browser support policy.
3. Avoid transpiling modern dependencies unless required by the actual support matrix.

P2 — Forced reflow and non-composited animations create CPU risk#

What PageSpeed reports:

• Forced reflow: 82.45 ms.
• Non-composited animated elements: 34.

Repo mapping:

• apps/www/styles/index.css defines broad animation and transition behavior.
• apps/www/components/Effects/SpotlightCursor.tsx reads getBoundingClientRect() on every mousemove and updates React state.
• apps/www/components/Global/GlobalReach.tsx and multiple homepage sections use Framer Motion in-view animations.

Recommended next actions:

1. Use a Chrome Performance trace to identify exact forced-reflow initiators before rewriting animations.
2. Prefer transform and opacity for animation; avoid layout-affecting properties.
3. Throttle pointer-driven effects or move them to CSS variables/requestAnimationFrame instead of React state per mousemove.
4. Respect prefers-reduced-motion and disable decorative motion on low-power contexts.

P3 — Accessibility issues are small but concrete#

What PageSpeed reports:

• Accessibility score: 95.
• Failing audits: heading order and touch targets.

Likely repo mapping:

• Heading order: homepage/footer heading components under apps/www/components/*.
• Touch targets: testimonial dot buttons in apps/www/components/Testimonials/*.

Recommended next actions:

1. Fix heading levels semantically without changing visual hierarchy.
2. Increase testimonial dot hit areas/spacing while preserving the visual design.
3. Re-run Lighthouse Accessibility or axe against the verified deployment.

Recommended remediation sequence#

1. Treat apps/www as the remediation target unless new deployment evidence supersedes it.
2. Trim/defer GTM and analytics first; this is the largest named transfer/main-thread offender.
3. Identify chunk 6258 with bundle analyzer and reduce unused app JavaScript from global providers and below-the-fold sections.
4. Scope homepage-unused CSS, especially MDX/Code Hike styles if coverage confirms they are unused on /.
5. Fix forced reflow and non-composited animation sources from trace evidence.
6. Clear the heading-order and testimonial touch-target accessibility findings.

Follow-up success criteria#

• Performance remains at or above 90 while third-party transfer and main-thread time decrease.
• Unused JavaScript drops materially from 104 KiB, with GTM and app chunk contributions tracked separately.
• Render-blocking CSS stays below 190 ms and unused CSS drops from 39 KiB.
• Forced reflow drops below 82.45 ms and non-composited animation warnings decrease.
• Accessibility heading-order and touch-target failures are fixed or explicitly accepted with rationale.