Frame rate, mid-tier Android
Measured on device, not in a lab. The animation work is CSS scroll-driven, so this is a hardware question rather than a code one.
This site, measured.
We ask clients to hold their sites to budgets, so ours is published — and now checked against a real build rather than stated as an intention. Where we are over, it says so.
First-load JS, heaviest routeover target, under ceiling
149kBtarget 120kB · ceiling 160kB
Gzipped, every chunk that route pulls in.
First-load JS, median routeover target, under ceiling
146.9kBtarget 120kB · ceiling 160kB
The typical page rather than the worst one.
Shared baseline JSwithin budget
100kBtarget 100kB · ceiling 120kB
Loaded once, then cached across every route.
CSS, whole sitewithin budget
14kBtarget 20kB · ceiling 30kB
Gzipped. No framework, so this is all of it.
Self-hosted fontswithin budget
47.1kBtarget 60kB · ceiling 90kB
2 WOFF2 files, no third-party font request.
Measured 2026-08-12 from a production build of 29 routes, by scripts/measure.mjs reading the build output directly. 2 of 5 lines are over target. They are published anyway — a budget you only show when you pass it is marketing, not a budget.
96PerformanceMobile, throttled
96AccessibilityAutomated portion only
100Best practices
66SEOnoindex while on the staging host
2.0 sLargest contentful painttarget < 2.0 s
0Cumulative layout shifttarget < 0.1
30 msTotal blocking timetarget < 200 ms
Measured 2026-08-12 against futragency.vercel.app, mobile, using our own audit tool — the same one you can point at your site. SEO scores 66 because we are deliberately noindexed while the site is on a staging hostname whose canonical URLs point at the production domain. That resolves when DNS completes. We would rather publish a low score with its reason than wait until the number flatters us.
WCAG 2.2 AA, core journeys
Automated tooling catches roughly a third of the criteria, and the easy third. The remainder is a manual pass with a keyboard and a screen reader, and the 96 above should not be read as a conformance claim.
LCP on the production domain
The 2.0s above was measured on the staging host. It will be re-measured on futragency.com once DNS completes, and republished whether it improves or not.
Routes29
Images711.5 MB total
Fonts247.1 kB
CSS14kB gzipped
| Route | First-load JS |
|---|---|
/contact | 149 kB |
/work | 148.7 kB |
/audit | 147.8 kB |
/ | 147.5 kB |
/not-found | 146.9 kB |
FrameworkNext.js 15 · React 19
HostingVercel
StylingCSS custom properties · no framework
MotionCSS scroll-driven animations
SearchClient-side index over the content model
Display typeCCS Glyke, self-hosted WOFF2
page27
capability11
work20
person6
writing6
Rev 02
Site audit tool, six published technical articles, contact form wired to CRM with attribution, four-gate contact sheet, comparison pages, procurement pack, published legal pack. Domain attached with the legacy WordPress URLs redirected. Performance budget switched from targets to measured figures.
Rev 01
First issue. Search shell, six vertical destinations, capability pages with computed project mapping.
How revisions work
Behaviour on this site is measured monthly. Where the evidence suggests a change, it is proposed, approved, and issued as the next revision with a line in this log. The site improves in public.