verified own-site case study

Reducing this website’s mobile lab LCP.

We optimized our own public homepage, preserving its design, theme switcher, opening sequence, analytics, attribution, and checkout behavior, then published every measured sample.

Written and measured by the web developers company engineering team · August 9, 2026
Evidence boundary. This is a two-run before-and-after Lighthouse 13.4.1 mobile lab comparison on our own website. It is not Chrome UX Report field data and makes no ranking, traffic, conversion, revenue, or client-result claim.

answer first

The measured median changed from 74.5 to 93.5.

Median Largest Contentful Paint fell from 4.872 seconds to 2.625 seconds. Median First Contentful Paint and Speed Index also fell. Total Blocking Time and Cumulative Layout Shift increased, so they are shown alongside the improvement rather than omitted.

Performance score74.5 → 93.5out of 100 · improved
Largest Contentful Paint4.872s → 2.625slower · improved
First Contentful Paint3.045s → 2.481slower · improved
Speed Index4.360s → 2.481slower · improved
Total Blocking Time1ms → 14.5mshigher · regressed in median
Cumulative Layout Shift0.000153 → 0.004822higher · regressed in median

methodology

Comparable live mobile lab samples.

We ran two consecutive Lighthouse samples on the live canonical homepage before the change and two after it. All four used Lighthouse 13.4.1 mobile defaults on August 9, 2026. The reported comparison is the median of each two-sample set. Best Practices and SEO scored 100 in every sample.

Two runs expose some variance but do not establish a field distribution. Network, host, CPU, cache, and third-party timing can change a synthetic result. The downloadable evidence retains the raw numeric values and fetch times used here.

SamplePerformanceFCPLCPSpeed IndexTBTCLS
Before 1743.083s4.884s4.813s1ms0.000153095
Before 2753.007s4.859s3.907s1ms0.000153095
After 1952.396s2.430s2.396s29ms0.004821658
After 2922.566s2.821s2.566s0ms0.004821658
Download the exact machine-readable evidence

implementation

Four bounded changes to the loading path.

01

Removed a render-blocking font stylesheet request

The same Google-hosted base font files were declared directly in the existing CSS. This removed the stylesheet dependency without changing the chosen typefaces.

02

Preloaded the heading font used by the LCP element

The exact Archivo font resource already used by the page was made discoverable in the document head.

03

Deferred scripts in their existing dependency order

The app, marks, themes, and intro assets kept their order and behavior while no longer blocking HTML parsing.

04

Made versioned static assets immutable

Requests carrying the site’s explicit asset version receive long-lived immutable caching. Unversioned pages and assets remain revalidatable.

verification

Performance did not replace behavior testing.

Automated checks confirmed the script order, font preload, absence of the render-blocking font stylesheet, version consistency, and distinct cache policies for versioned assets and HTML. Existing SEO, attribution, client-event, Stripe webhook, and server tests also passed before deployment.

The release preserved the public interface, themes, opening sequence, GA4 configuration, first-touch attribution, checkout, and verified Stripe webhook behavior. The source change is recorded in commit 0b59c97.

What this result does not establish

  • It does not establish real-user Core Web Vitals or a Chrome UX Report assessment.
  • It does not cover every route, device, network, geography, or returning-visitor cache state.
  • It does not establish a search-ranking, traffic, conversion, or revenue effect.
  • It does not erase the measured TBT and CLS increases shown above.

continue reading

The reusable engineering method.

start a project

Need measured website performance work?

Tell us what the business needs the site or application to do. We’ll reply with an honest first direction and the questions needed to scope it.

Talk to an engineer