STACK_01
Next.js 14, App Router
Server-side rendered by default. Edge-cached at Vercel. Image optimisation. Streaming routes.
( _web-infrastructure )
Websites engineered for paid traffic, AI engines and human buyers, in that order.
Next.js 14 builds with conversion architecture, schema markup, and Core Web Vitals under target. Built to convert ad clicks at 8%+ and rank in AI answers from day one.
Web infrastructure · live work
What we typically find
The failure modes we audit out first.
4 patterns we see almost every time we look at an existing account. The audit pulls them apart before we touch anything.
Failure mode 01
Templated builders that load in 6 seconds
Wix, Squarespace, GoDaddy. Fine for a hobby site. Death for paid traffic. LCP over 4s halves conversion rate before anyone reads a headline.
Failure mode 02
No schema, no llms.txt, no chance of AI citation
Most service business sites have zero structured data. AI engines can't extract what to cite. You're invisible in the answer layer.
Failure mode 03
Hero copy that doesn't answer the buyer's question
'Your trusted partner since 1987' is not a hook. The buyer needs to see the outcome inside 4 seconds.
Failure mode 04
Conversion paths that take 8 clicks
Most service sites bury the booking flow. Sticky nav CTA, exit-intent capture, form abandonment recovery: standard, not optional.
Visualised
Core Web Vitals on our default build, versus what we see in audits.
All three vitals well below Google's target on our shipped builds. Most service business sites we audit fail at least two of three.
LCPpassing
0.0s
target ≤ 2.5s
INPpassing
0.0ms
target ≤ 200ms
CLSpassing
0.0
target ≤ 0.1
Time-to-LCP, our builds vs typical service-business sites
PYKSL Next.js builds1.2s
WordPress + page builder4.6s
Squarespace template3.8s
Wix template5.1s
↳ LCP over 2.5s halves conversion rate on paid traffic. The math doesn't move.
The stack
What we ship, role by role.
Every layer named, scoped, and owned. No black boxes.
STACK_02
Vercel + Cloudflare
Hosting on Vercel, WAF + extra cache layer in front via Cloudflare. Five-nines uptime in practice.
STACK_03
shadcn/ui + Tailwind
Component primitives we control. No design-system rot, no opinionated framework lock-in.
STACK_04
Sanity / MDX content
Editor-friendly CMS where it earns its keep. Otherwise MDX for engineering-led teams.
STACK_05
Resend + Cal.com
Transactional email and calendar booking that work out of the box. Drop-in integrations.
STACK_06
Schema + llms.txt
JSON-LD on every page, llms.txt at root, structured data tested against the Rich Results validator pre-launch.
How we ship it
4 steps. Each one auditable.
Step 01, audit
Current site teardown
Lighthouse, schema validator, conversion-path walk-through, content scoreboard. We deliver a written thesis.
Step 02, architecture
Sitemap + IA
Information architecture, URL conventions, schema plan, conversion-path map. Approved before we touch code.
Step 03, build
30-day shipping sprint
Foundation, core pages, industry pages, content, conversion + tracking. Weekly demo every Friday.
Step 04, launch + measure
Ship and monitor
Soft launch, redirect map deployed, Search Console + IndexNow pings, AI engine submissions. Performance dashboard live.
Average client outcome
Average LCP across active client sites on mobile networks.
Vercel Analytics, p75 across active sites
1.2s
FAQ, web infrastructure
The honest answers.
- 01
Why Next.js 14 specifically?
Server-side rendering by default, edge-cached static pages, image optimisation built-in, and an App Router that makes schema injection trivial. The combination is currently the cleanest path to CWV under target and AI-engine-readable HTML.
- 02
Do you build on our existing CMS?
Sometimes. If you're on WordPress and married to it, we can keep the back-office and rebuild the public layer in Next. For a clean slate, we usually pair Next + Sanity or MDX.
- 03
What's your typical Core Web Vitals target?
LCP under 2.0s on mobile, INP under 200ms, CLS under 0.05. We monitor with Vercel Analytics and ship a weekly delta.
- 04
How does AI-engine optimisation actually work?
llms.txt at root, JSON-LD schema on every page, server-rendered HTML so engines don't need to execute JS, definition/FAQ/comparison content patterns that engines extract reliably. Plus tracking your share-of-voice in ChatGPT, Claude, Gemini, Perplexity.
- 05
Can you migrate us from our current site without losing rankings?
Yes. 301 redirect map per URL, schema parity preserved, content audit before launch. Worst-case dip is usually under 10% for 2 weeks. Most projects come out ahead within 90 days.
Web infrastructure · taking clients
Built for paid traffic. Built to be cited.
If your current site loads in over 3 seconds or has zero schema, the audit will surface it in 24 hours.