Performance report

What changed, measured the same way twice

Every figure here comes from a real Lighthouse test of this website.

Measured with Lighthouse

Lighthouse scores

Home page, mobile test

Performance

Before

After

Accessibility

100

Before

100

After

Best Practices

100

Before

100

After

SEO

100

Before

100

After

Core Web Vitals

Loading, responsiveness and stability

INP needs real user interactions, so a page-load test can't produce it. Total Blocking Time is shown as the lab stand-in.

Largest Contentful Paint

LCP

Good

When the largest visible element finishes rendering.

Before4.0 s

After926 ms

good ≤ 2.5 s

Cumulative Layout Shift

CLS

Good

How much visible content moves unexpectedly while loading.

Before0.09

After0.00

good ≤ 0.10

Total Blocking Time

TBT

Needs work

How long long tasks block the main thread. The lab stand-in for responsiveness.

Before2.5 s

After367 ms

good ≤ 200 ms

Interaction to Next Paint INP

Not measured

Supporting metrics

First paint and visual progress

First Contentful Paint

FCP

Good

When the first text or image appears.

Before1.5 s

After926 ms

good ≤ 1.8 s

Speed Index

Good

How quickly the visible content fills in during loading.

Before2.7 s

After1.0 s

good ≤ 3.4 s

Page weight

What the browser had to download

Before After

Total transfer: Before: 4.5 MB
After: 234 KB
JavaScript: Before: 315 KB
After: 147 KB
Images: Before: 4.1 MB
After: 0 KB
Fonts: Before: 52 KB
After: 52 KB

Requests: 22 before, 19 after

Fix by fix

What each change did

Home page on mobile, re-tested after each fix. Each row includes all the fixes above it.

Home page mobile results after each fix, applied in order
Cumulative state	Performance	LCP	CLS	TBT	Transfer
Seeded baseline	56	4.0 s	0.09	2.5 s	4.5 MB
Removed the oversized hero image	69	1.5 s	0.09	2.3 s	401 KB
Removed Chart.js, Moment and Lodash	68	1.5 s	0.09	3.5 s	234 KB
Removed hydration work and the late banner	100	1.5 s	0.00	2 ms	233 KB
Restored self-hosted fonts with next/font	93	938 ms	0.00	314 ms	232 KB

The score of 100 in the third row is misleading. Total Blocking Time only counts tasks after first paint, and the slow fonts most likely delayed first paint past the startup work. The last row is the better page, even though its score is lower. Across three five-run tests, essentially the same code scored 93, 97 and 91, which shows how much lab scores vary.

All results

Every page and device tested

Lighthouse results by page, device and stage
Test	Performance	Accessibility	Best Practices	SEO	LCP	FCP	CLS	TBT	Speed Index	Transfer	Saved report
Home, mobile · before	56	100	100	100	4.0 s	1.5 s	0.09	2.5 s	2.7 s	4.5 MB	Open
Home, mobile · after	91	100	100	100	926 ms	926 ms	0.00	367 ms	1.0 s	234 KB	Open
Home, desktop · before	86	100	100	100	923 ms	408 ms	0.04	310 ms	863 ms	4.5 MB	Open
Home, desktop · after	100	100	100	100	268 ms	268 ms	0.00	20 ms	358 ms	258 KB	Open
Performance report, mobile · before	92	100	100	100	820 ms	820 ms	0.00	357 ms	1.4 s	4.5 MB	Open
Performance report, mobile · after	91	100	100	100	947 ms	947 ms	0.00	373 ms	1.0 s	226 KB	Open
Performance report, desktop · before	100	100	100	100	373 ms	373 ms	0.000	0 ms	441 ms	4.5 MB	Open
Performance report, desktop · after	100	100	100	100	270 ms	270 ms	0.00	8 ms	378 ms	259 KB	Open
Services, mobile · before	97	100	100	100	789 ms	789 ms	0.00	205 ms	1.0 s	4.5 MB	Open
Services, mobile · after	96	100	100	100	779 ms	779 ms	0.00	244 ms	800 ms	222 KB	Open
Services, desktop · before	100	100	100	100	231 ms	231 ms	0.001	0 ms	467 ms	4.5 MB	Open
Services, desktop · after	100	100	100	100	221 ms	221 ms	0.00	6 ms	298 ms	257 KB	Open
Contact, mobile · before	93	100	100	100	776 ms	776 ms	0.00	321 ms	1.4 s	4.5 MB	Open
Contact, mobile · after	95	100	100	100	781 ms	781 ms	0.00	267 ms	812 ms	222 KB	Open
Contact, desktop · before	100	100	100	100	385 ms	385 ms	0.04	0 ms	568 ms	4.5 MB	Open
Contact, desktop · after	100	100	100	100	223 ms	223 ms	0.00	4 ms	396 ms	254 KB	Open

The seeded problems mostly affected the home page, so that is where scores improved most. Mobile scores on the other pages stayed within normal run-to-run variation, even though each page downloaded about 4.3 MB less. A score does not capture everything a visitor pays for.

Live site reference

A separate check of the deployed site

One PageSpeed Insights run of the live home page. It used Google's test machines and Vercel's network, so it isn't comparable with the local results above, and it has no before version.

Mobile

100

Performance

100

Accessibility

100

Best Practices

100

SEO

Desktop

100

Performance

100

Accessibility

100

Best Practices

100

SEO

Mobile FCP: 0.9 s
Mobile LCP: 0.9 s
Mobile TBT: 0 ms
Mobile CLS: 0
Mobile Speed Index: 1.6 s

PageSpeed Insights (Lighthouse 13.6.0), captured November 6, 2026. A single run can vary, and a perfect lab score doesn't prove real visitors have a fast experience. Google had no field data for this new site. Open the PageSpeed report

Method

How the tests were run

Tool: Lighthouse 13.6.0 CLI with its default mobile and desktop presets, including simulated throttling
Target: A local production build (next start) on one machine. This excludes real-world network and hosting, so values may differ on a live deployment. Lighthouse also warned that this machine's CPU is slower than it expects, which can lower scores in both versions.
Sampling: Median of 5 runs for each page and device. The home page's mobile runs scored between 40 and 63 before, and between 87 and 93 after, so small differences aren't meaningful.
Known differences: The before build showed placeholder numbers on its report, home and case study pages, while the after build shows real ones. The site's other content is the same. Later commits only update results, documents, page text and the share-image tags on interior pages.
Before: Commit 6fde805, tested November 6, 2026
After: Commit babd027, tested November 6, 2026

Read the case study