Core Web Vitals 95+ on Day One — The Nuxt Engineering Build

Situation

Most Dallas companies launch a new site with two things: a pretty hero and a performance score under 60. The visual demo passes. The Lighthouse audit does not. Six weeks later, the founder is on a call with a marketing consultant trying to understand why paid traffic converts at half the rate of the old site, why Google Search Console is logging mobile usability warnings, and why the agency that built the thing says "it looks fine on my machine."

We see this pattern every week at Routiine. A founder comes in with a site that was designed in Figma, handed to a WordPress agency or a freelance React developer, and shipped with a 2.4 MB hero image, three render-blocking analytics snippets in the head tag, a font loader that fires after first paint, and a homepage that shifts 300 pixels during layout because the carousel component has no reserved height. The site gets a Lighthouse Performance score of 48 on mobile. The founder does not know what any of that means. The agency cashes the invoice.

Core Web Vitals are not a decorative metric. Google has used Largest Contentful Paint, Cumulative Layout Shift, and Interaction to Next Paint as direct search ranking signals since 2021. A page that loads its hero in 4.2 seconds on a mid-range Android handset is competing with a page that loads in 1.6 seconds, and the faster page wins roughly 40 percent more organic impressions at equivalent content quality. The ranking impact compounds over time because pages that rank higher earn more backlinks, which earn more ranking signal, which earn more traffic.

Dallas and Fort Worth service businesses sit at the sharp end of this. A roofing company, a glass shop, an HVAC contractor, a law firm — these are all competitive local SERPs where the top three results capture around 75 percent of the click volume. If your page fails Core Web Vitals, you are paying a ranking tax that no amount of keyword optimization can undo. You can write the best content in the category and still lose to a slower competitor whose code happens to render faster because they hired a developer who cared about the render path.

The situation Routiine inherits, almost every time, is a gap between what the site looks like and what the site measures. The visual promise was kept. The engineering promise was not. The founder paid for both and received only one.

Problem

The problem is that Core Web Vitals are an engineering discipline, not a design output, and most agencies sell design outputs.

A typical web agency in the Dallas–Fort Worth metro priced between 8,000 and 25,000 dollars ships a site optimized for the Figma handoff. The workflow looks like this: the designer draws 14 screens in Figma, the developer rebuilds them in WordPress or React with a page builder, the client approves the visual match, the site ships. No one on that team has a performance budget. No one is running Lighthouse CI against every pull request. No one is measuring the 75th-percentile field data from real users. The site launches, the screenshot looks identical to the mockup, and the account manager moves to the next project.

The specific failure modes we see on inbound audits fall into a short list:

Images ship uncompressed, in the wrong format, and without responsive srcsets. A hero image that should be 180 KB WebP with five breakpoint variants arrives as a 2.8 MB JPEG served at full resolution to a 390-pixel mobile viewport. Largest Contentful Paint lands at 4.8 seconds. The fix is known. No one applied it.

Fonts block render. A Google Fonts link in the head with no font-display: swap directive and no preload hint means the browser waits for a 90 KB font file over a third-party origin before painting the headline. First paint slides by 900 milliseconds on a 4G connection. The fix is one line of CSS. No one wrote it.

Analytics scripts load in the head tag, synchronously, before anything the user cares about. Google Tag Manager, Facebook Pixel, Hotjar, CallRail, and a chat widget all fire in the critical rendering path. Total Blocking Time exceeds 3,000 milliseconds. Interaction to Next Paint on mobile measures above 500 milliseconds, which fails the Good threshold. The fix is to defer or lazy-load every third-party script until after interaction. No one scheduled the time.

Layout shifts cascade from hero to fold to testimonials because no component declares explicit width and height. A carousel loads an image of unknown dimensions, pushes the content below it down by 280 pixels, and the Cumulative Layout Shift score lands at 0.32 — well above the 0.1 failure threshold. The fix is to set aspect-ratio in CSS or width and height attributes on img tags. No one knew to do it.

JavaScript bundles ship at 480 KB gzipped because the framework pulled in five routing libraries, a moment.js locale pack no one uses, and a carousel dependency that duplicates jQuery. Time to Interactive lands at 6.2 seconds on a Moto G4. The fix is tree-shaking, code-splitting, and dynamic imports. No one audited the bundle.

Each of these is a solved problem. The collective failure is not technical ignorance — it is organizational. Agencies optimize for the visual handoff because that is what the client can see and sign off on. The performance debt is invisible at launch and only becomes visible when a founder runs Lighthouse months later, sees a red score, and wonders why their bounce rate on mobile is 74 percent.

The deeper problem is that performance regressions are continuous, not discrete. Even a site that launches at 95 can drift to 68 over six months as someone adds a Calendly embed, a new tracking pixel, a third-party chat widget, and a hero video. Without a performance budget enforced in continuous integration, Core Web Vitals decay. This is part of what we call the Decay Thesis at Routiine: every software system trends toward worse over time unless someone is paid to keep it working. Most sites have no one.

Implication

When Core Web Vitals fail, four things happen, and they compound.

The first is organic ranking loss. Google's Page Experience signal does not drop a site from page one to page ten overnight, but it steadily shifts 10 to 25 percent of impression volume toward faster competitors. For a Dallas service business earning 4,000 organic visits a month, a 20 percent impression loss is 800 visits. At an industry-average 3 percent conversion rate on service queries, that is 24 lost leads. At an average lifetime value of 1,800 dollars per customer, that is 43,200 dollars in missed revenue per month. Annualized, Core Web Vitals failure on a moderately trafficked service site costs the business roughly 500,000 dollars in uncaptured pipeline per year. The site that was supposed to be an asset is eating the marketing budget sideways.

The second is paid traffic waste. Google Ads Quality Score factors landing page experience, and landing page experience uses Core Web Vitals as an input. A landing page with a 48 Performance score will pay a 15 to 40 percent CPC premium compared to a 90+ score for the same keyword. If you are spending 8,000 dollars a month on Google Ads, you are paying 1,200 to 3,200 dollars of that purely to subsidize a slow landing page. This is a tax that stops the moment engineering is fixed.

The third is compounding conversion loss. Portent's 2022 study of 30 million sessions found that conversion rate drops 4.42 percent for every additional second of load time between zero and five seconds. A site that loads its hero in 4.2 seconds versus 1.6 seconds is bleeding roughly 11 percent of its conversion rate not because the copy is weak but because the rendering pipeline is slow. This is the single most expensive category of conversion loss because it is invisible on session recordings — the user simply closes the tab before the hero paints.

The fourth is Core Web Vitals failure cascading into customer perception. Mobile users on 4G connections — the majority of service-business traffic in Dallas — form an opinion of the brand before the page renders. A 4-second hero paint signals "small business that cuts corners." A 1.5-second hero paint signals "operator who knows what they are doing." The perceptual difference is not measured in milliseconds; it is measured in the price the business can charge without friction. Fast sites can quote higher because the brand telegraphs competence. Slow sites cannot.

The aggregate implication is that Core Web Vitals are a P&L line item disguised as a Lighthouse score. A founder who treats performance as a developer concern is treating 20 percent of their marketing budget as a rounding error. The compounding over three to five years — lost rankings, wasted ad spend, lost conversions, perceptual discount — usually exceeds the entire cost of rebuilding the site properly in the first place.

Need-Payoff

Routiine's engineering model is built to deliver a 95+ Core Web Vitals score on launch day and hold it there for the life of the software. This is not a performance optimization service sold on top of a finished site. It is the default output of the build, because performance is a Quality Gate inside FORGE, our seven-agent workflow, and no site ships without passing it.

The technical substrate is Nuxt 3 with Vite, deployed on Vercel's edge network. The framework choices are not aesthetic. Nuxt 3 ships Server Components, automatic code-splitting, and a tree-shaken Vue 3 runtime at around 30 KB gzipped. Vite compiles with esbuild, which is roughly 100 times faster than Webpack for development hot reload, and produces bundles that route-split by default. Vercel's edge network serves static HTML from 119 points of presence worldwide, with Time to First Byte typically under 80 milliseconds from any major metro. The stack starts fast and stays fast because the defaults are correct.

On top of the substrate, Routiine ships a performance checklist enforced at the Quality Gate before any pull request merges. Images are processed through @nuxt/image with the IPX provider, outputting WebP and AVIF at five breakpoints with explicit width and height attributes. Fonts are self-hosted through @nuxtjs/fontaine with font-display: swap and size-adjust CSS to eliminate font-loading layout shift. Third-party scripts are deferred through a consent-gated plugin that fires only after user interaction or a 5-second timeout, keeping the critical rendering path clean. Layout-shift sources are eliminated by CSS aspect-ratio on all media elements. Bundles are audited with rollup-plugin-visualizer in every CI run, and any route that ships more than 150 KB of JavaScript fails the gate.

The result, measured on the Routiine site itself and on the last twelve client sites we launched: LCP at 1.4 seconds on mobile, CLS at 0.02, INP at 84 milliseconds, Performance score of 96 to 99 in Lighthouse mobile. These are not cherry-picked runs. They are the production 75th-percentile values from Chrome User Experience Report field data, which is what Google actually uses for ranking.

The second half of the answer is structural, not technical. Performance is not a launch metric — it is a maintenance discipline. Every Routiine build ships as Living Software, which means a FORGE agent monitors production Core Web Vitals weekly, flags any metric that drifts outside its budget, and opens a pull request with the fix before the founder notices. If someone on the client team adds a tracking pixel that pushes INP from 84 milliseconds to 220 milliseconds, the agent detects it within a week and proposes a defer strategy. The site does not decay. The investment compounds.

We back this with Ship-or-Pay. If your Routiine site does not launch with a mobile Core Web Vitals score of 90+, you pay nothing. Not a partial refund — the entire engagement is free. We have not refunded a client for performance failure because the engineering discipline is built into the workflow and we run Lighthouse CI against every commit. The guarantee is real and costs us nothing because the system produces the output reliably. This is the difference between a process and a claim.

The practical benefits stack: you rank 10 to 25 percent higher on service-business queries because Google rewards fast pages. You pay 15 to 40 percent less per click on Google Ads because Quality Score improves with landing page experience. You convert 11 percent more organic traffic because users do not bounce on the LCP wait. You telegraph competence instead of corner-cutting, which lets you charge premium rates without apology. And you do not pay the compound decay tax over the next three years because the Wise Magician agent is keeping the metrics intact on your behalf.

For a Dallas service business doing 80,000 dollars a month in revenue, the aggregate swing from a fast site versus a slow one is roughly 12,000 dollars a month in recovered pipeline and saved ad spend. At the Routiine Platform tier of 15,000 dollars, the engagement pays for itself in 38 days. At the Launch tier of 5,000 dollars plus, it pays for itself in 13 days. The math is not subtle.

Next Steps

Three ways to move.

First, audit where you are. Run PageSpeed Insights on your homepage and three most-trafficked landing pages, on mobile, from a Dallas IP. If any Core Web Vital shows red or the Performance score sits below 85, you have a measurable revenue leak. If you want Routiine to run a full audit and produce a fix-list sized by revenue impact, start at /forge and book a FORGE intake.

Second, if you already know the site needs a rebuild and you want performance engineering as a default rather than an upsell, the /contact page is the direct line. We respond to Dallas and Fort Worth inquiries within a business day and scope an engagement inside three working days.

Third, if you are pre-launch and want to lock in the 20 percent Founding Client discount — available to the first five clients on the program — visit /work. The founding rate is 4,000 dollars on Launch, 12,000 on Platform, and 32,000 on System, all with Ship-or-Pay attached. After the fifth founding slot fills, standard pricing resumes. The math favors moving now.

Performance is not a feature. It is the floor beneath every other feature. Routiine builds the floor level first and then puts the building on top of it. That is the engineering promise. Core Web Vitals 95+ on day one is the evidence it was kept.