Visitar URL original
toolbelt/docs/performance-baseline.md at develop · codercops/toolbelt · GitHub
Skip to content

performance-baseline.md

Latest commit

 

History

History
57 lines (45 loc) · 4.19 KB

performance-baseline.md

File metadata and controls

57 lines (45 loc) · 4.19 KB
Raw

Performance Baseline

Initial performance baseline audit for tools.codercops.com as part of issue #28 under the parent SEO initiative (#18).

Scope

Pages were identified strictly through existing route definitions (lib/tools.ts, app/(tools)) and production site navigation:

  • Home: https://tools.codercops.com/ (/)
  • JSON Formatter: https://tools.codercops.com/json-formatter (/json-formatter)
  • JWT Decoder: https://tools.codercops.com/jwt-decoder (/jwt-decoder)
  • Base64: https://tools.codercops.com/base64 (/base64)
  • Invoice Generator: https://tools.codercops.com/invoice-generator (/invoice-generator)

Note on URL Encoder (/url-encoder): The URL Encoder route was checked during the audit, but returned HTTP 404 on production (not yet live). It was therefore excluded from this live production baseline per the maintainer requirements.

Methodology

  • Tooling: Audits were conducted using Lighthouse (v13.5.0) in a headless environment.
  • Audit Settings:
    • Mobile: Form factor mobile, mobile screen emulation (412 × 823 viewport, DPR 1.75), simulated mobile network & CPU throttling (1,600 Kbps down, 750 Kbps up, 150 ms RTT, 4× CPU slowdown).
    • Desktop: Preset desktop, desktop screen emulation (1350 × 940 viewport, DPR 1.0), desktop network & CPU throttling (10,000 Kbps down, 3,375 Kbps up, 40 ms RTT, 1× CPU slowdown).
  • Run Count & Median Methodology: Each page and device combination was audited three times under identical settings. For each metric (LCP, CLS, TBT), the values were sorted across the three runs and the middle value (median) was recorded.
  • Metric Units:
    • LCP (Largest Contentful Paint): Seconds (s)
    • CLS (Cumulative Layout Shift): Unitless score
    • INP (Interaction to Next Paint): Milliseconds (ms)
    • TBT (Total Blocking Time): Milliseconds (ms)
  • Missing INP Field Data: Interaction to Next Paint (INP) is a real-user field metric collected over time via the Chrome User Experience Report (CrUX). Because these audits were lab-simulated using Lighthouse and field data is not available in CrUX for these URLs, field INP is recorded exactly as no field data without substituting lab metrics.

Baseline Table

page device date LCP CLS INP TBT
Home mobile 2026-10-14 2.08 s 0.000 no field data 51 ms
Home desktop 2026-10-14 0.69 s 0.000 no field data 0 ms
JSON Formatter mobile 2026-10-14 2.11 s 0.000 no field data 67 ms
JSON Formatter desktop 2026-10-14 0.69 s 0.002 no field data 0 ms
JWT Decoder mobile 2026-10-14 2.12 s 0.000 no field data 102 ms
JWT Decoder desktop 2026-10-14 0.76 s 0.000 no field data 0 ms
Base64 mobile 2026-10-14 2.09 s 0.000 no field data 59 ms
Base64 desktop 2026-10-14 0.70 s 0.000 no field data 0 ms
Invoice Generator mobile 2026-10-14 2.24 s 0.465 no field data 61 ms
Invoice Generator desktop 2026-10-14 0.78 s 0.554 no field data 0 ms

Findings and Follow-up Actions

Thresholds evaluated:

  • LCP: Passing (< 2.5 s) across all pages on both mobile (2.08 s – 2.24 s) and desktop (0.69 s – 0.78 s).
  • TBT: Passing diagnostic guidance (< 200 ms "Good") across all pages on both mobile (51 ms – 102 ms) and desktop (0 ms).
  • CLS:
    • Home, JSON Formatter, JWT Decoder, and Base64 passed with minimal or zero shift (CLS ≤ 0.002).
    • Invoice Generator failed the CLS threshold (> 0.1) on both desktop (0.554) and mobile (0.465).
    • Culprit: When InvoiceGeneratorClient mounts, it starts with current = null rendering a small loading card, which is then replaced after hydration by the full toolbar, editor, and preview grids. This post-hydration swap pushes the FAQ section (section.mx-auto) down the page by hundreds of pixels, triggering layout shift.

Follow-up Issues Filed

  • #159: perf(invoice-generator): fix high Cumulative Layout Shift (CLS) on desktop and mobile