Core Web Vitals

Overview

Core Web Vitals (CWV) is Google's public, field-measured framework for evaluating user-perceived page experience. It comprises three metrics — Largest Contentful Paint (LCP), Interaction to Next Paint (INP), and Cumulative Layout Shift (CLS) — assessed at the 75th percentile of real-user (Chrome User Experience Report, "CrUX") data over a rolling 28-day window. The framework matters for two distinct reasons: it is a documented (small) Google ranking signal, and it is the empirical handle on loading, interactivity, and visual-stability problems that consistently correlate with bounce rate and conversion in independent and vendor-published case studies.

This page consolidates the canonical 2026 thresholds, the historical INP-replaces-FID transition, the catalogue of common metric "killers" with remediation patterns, the published revenue/bounce evidence (Walmart, BBC, Pinterest, Vodafone, Renault, Rakuten 24, redBus, Deloitte/Google, Economic Times, Akamai/SOASTA, Google neural-net), the platform pass-rate landscape as of November 2025, and a register of widely repeated statistics that fail primary-source scrutiny. See Agency methodology for small-business website projects and Conversion rate — realistic expectations for small-business websites for the full upstream research briefs from which this synthesis was drawn.

CWV thresholds (2026, confirmed unchanged)

Google's official "Good" Core Web Vitals thresholds remain unchanged through 2026:

Metric	Good (≤)	Needs Improvement	Poor (>)
Largest Contentful Paint (LCP)	2.5s	2.5–4.0s	4.0s
Interaction to Next Paint (INP)	200ms	200–500ms	500ms
Cumulative Layout Shift (CLS)	0.1	0.1–0.25	0.25

All measured at the 75th percentile of real-user field data (CrUX, 28-day rolling).

Source: https://developers.google.com/search/docs/appearance/core-web-vitals (last updated 2025-12-10 UTC); web.dev/articles/lcp; web.dev/articles/defining-core-web-vitals-thresholds. Confidence: Verified (primary Google documentation).

A site "passes" CWV only when all three metrics are in the Good band at p75 — a single poor metric fails the whole assessment for the page-experience signal.

INP replaced FID on March 12, 2024

Interaction to Next Paint became the responsiveness Core Web Vital on March 12, 2024, replacing First Input Delay (FID). FID measured only the delay of the first interaction; INP measures the latency of essentially every interaction across the page lifecycle and reports a near-worst-case value, which is materially harder to pass and far more diagnostic of real-world responsiveness problems.

Source: web.dev. Confidence: Verified.

The transition is corroborated in Addy Osmani, "History of Core Web Vitals" (https://addyosmani.com/blog/core-web-vitals/), which is the standalone reference for the date and the FID-to-INP rationale.

INP 150ms target (200ms is the failing floor)

A common practitioner pattern is to set internal performance budgets above Google's "Good" thresholds rather than at them, so that small regressions are caught before the public CrUX measurement crosses the failing line. Documented practitioner targets are:

• LCP ≤ 2.0s (2.5s is the failing threshold, not the target)
• INP ≤ 150ms (200ms is "Good", so aim 50ms inside)
• CLS ≤ 0.05

Per the HTTP Archive 2025 Web Almanac: "In June 2025, 77% of pages achieved good Interaction to Next Paint (INP) scores" — meaning 23% still fail. INP remains the metric most local-market competitors fail, in part because most providers do not yet measure it.

Source: HTTP Archive Web Almanac 2025 (Performance chapter) — almanac.httparchive.org/en/2025/performance. Confidence: Verified.

Real-user monitoring via the web-vitals library can report INP at the 75th percentile on deployed pages, providing the field counterpart to lab-based Lighthouse measurements during development.

CWV is a tiebreaker-class ranking factor, not a heavy lever

Independent studies consistently find CWV correlates with rank but is not a heavy weight.

• Google's own framing (John Mueller, Search Central documentation): CWV is a "page-experience signal" that "contributes to" ranking — tiebreaker class.
• Google Chromium research (cited via web.dev, Google News Initiative training): "When a site meets Core Web Vitals, research showed that people were 24% less likely to abandon the page."
• Perficient (17-week study, June–September 2021, n=1,188 keywords across 6 industries): "The data confirms CWV scores are not a large ranking factor… It shows a general correlation between pages with higher CWV scores and rank, but the rollout of the Page Experience update did not change the shape or scope of that correlation to any noticeable degree." Confidence: high — independent, methodology disclosed.
• Advanced Web Ranking (March 2022, n=3M pages, top 20 results): LCP showed the clearest visual trend with rank position 1-3 vs 8-10. But many top-10 pages had no CrUX data at all, meaning Google ranked them on other signals entirely.
• Backlinko 208,085-page study (2021): no significant correlation between CLS and bounce rate, time on site, or pageviews. Useful skeptic check against vendor hype.

Bottom line: CWV is a real but small direct ranking factor. The conversion and UX gains are far larger and more defensible than the SEO gains.

Implication: CWV is best positioned as a conversion / bounce-rate lever, not an SEO ranking lever — the evidence on the conversion side is much stronger (Measurement frameworks for small-business websites; see also the Rakuten 24 and Deloitte/Google case studies catalogued below).

CWV as AI-citation gate (2026)

The largest publicly disclosed direct analysis of Core Web Vitals as an input to AI-Overview / AI-search citation rates — Dan Taylor, Search Engine Land, January 13, 2026, n=107,352 pages — found that CWV functions as a gate, not a positive signal: failing CWV reduces citation eligibility, but passing CWV does not confer a citation boost. Measured correlations were small (LCP r = −0.12 to −0.18; CLS r = −0.05 to −0.09); severe CWV failure suppresses AI Overview citation, while going from "Good" to "great" CWV does not lift it. Taylor's framing:

"Core Web Vitals are therefore best understood as a gate, not a signal of excellence. In an AI-led search landscape, this clarity matters."

Source: Dan Taylor, "Core Web Vitals are a gate, not a signal" — Search Engine Land, January 13, 2026. Confidence: Verified (largest publicly disclosed AI-citation × CWV correlation study to date).

This inverts the assumption marketers tend to make about "faster site = better AI Overview citation rate" (catalogued in the unverifiable-stats register below).

2026 performance budget for SMB sites

A widely cited 2026 performance budget for small-business marketing sites targets total page weight ≤ 1.0 MB (versus the Web Almanac 2025 median of 2,559 KB on mobile), JavaScript execution time ≤ 200ms on a mid-range Android device (Moto G Power class), and the LCP image preloaded with fetchpriority="high" and served in WebP or AVIF.

Source: HTTP Archive Web Almanac 2025 — Page Weight chapter (almanac.httparchive.org/en/2025/page-weight) for the 2,559 KB mobile median; Chrome Aurora team guidance for the JS execution-time ceiling. Confidence: Verified (page-weight median); Industry-consensus (budget targets).

Contractor-website CWV is load-bearing in the buyer journey

A contractor's own website CWV is a load-bearing trust signal. It is the second or third surface in the buyer journey (Lead-generation directories for trades and home services) — after discovery (referral, directory, ad) and before contact.

The structural reality (Lead-generation directories for trades and home services):

• 48% of mobile sites pass CWV (mid-2025)
• 43.44% of WordPress mobile origins pass (June 2025; ~45–46% by late 2025)
• Most contractor sites are WordPress + heavy page builders + shared hosting — they cluster on the fail side.
• 53% of mobile visits are abandoned if the site takes more than 3 seconds (Google/DoubleClick 2016, canonical).

Why this matters: a buyer who finds a contractor on a directory such as HomeStars, then searches the contractor's name on mobile, then clicks through to the site, abandons before LCP renders if the site is slow. The directory lead is then wasted at the contractor's own front door — a structural failure mode common to verticals that pay for directory traffic but underinvest in site performance.

The diagnostic frame is straightforward: PageSpeed Insights on the homepage and at least one representative deep page (a service page or case study) on mobile reveals whether LCP, INP, or CLS is the limiting metric. The typical causes on contractor sites are heavy page-builder JavaScript (Elementor, Divi, WPBakery), unoptimised hero images, shared hosting with poor TTFB, or render-blocking third-party scripts. WordPress's primary failure mode is LCP and TTFB, not INP (Lead-generation directories for trades and home services).

A field-sample audit of Ontario general-contractor websites has not yet been conducted. The placeholder estimate of "60–70% likely to fail CWV on mobile" is therefore [Directional], derived from the HTTP Archive 2025 WordPress pass rates adjusted upward for the builder-heavy construction vertical.

LCP killers (catalogue)

Ranked by observed frequency in the field (HTTP Archive 2025 + practitioner consensus):

1. Hero image not preloaded and not given fetchpriority="high". Single largest cause. Google's own test on Google Flights: a single fetchpriority="high" moved LCP from 2.6s to 1.9s.
2. Lazy-loading the LCP image (loading="lazy" on the above-fold hero). Common WordPress default since 5.5; the CMS now excludes the first 3 images, but custom themes routinely undo this.
3. Slow TTFB — shared hosting with 800ms–1.5s server response; uncached dynamic queries; missing CDN.
4. Render-blocking CSS — Elementor stylesheets, theme.css from heavy themes.
5. LCP image hidden from the preload scanner (via data-src, JS lazy-load libraries, CSS background-image). Web Almanac 2025: 7% of sites still do this.
6. Unoptimised image format — JPEG/PNG instead of WebP/AVIF. WebP is 25–34% smaller than JPEG at equivalent SSIM, with Google's own FAQ summarising the average as ~30% (Google WebP Compression Study, updated August 7, 2025 — https://developers.google.com/speed/webp/docs/webp_study). Confidence: Verified (Google primary).
7. Wrong-resolution images — no srcset/sizes, serving 4000px desktop heroes to mobile.
8. Third-party scripts blocking the main document while the LCP image waits in queue.

Remediation pattern: inline critical CSS → preload + fetchpriority="high" the hero → modern image formats → srcset/sizes → move TTFB-heavy work to the edge, static generation, or cache. The LCP element should additionally carry explicit width/height attributes (preventing CLS) and never be hidden from the preload scanner via data-src, a CSS background-image, or a JS-injected URL.

INP killers (catalogue)

Ranked by observed frequency in practitioner reports and HTTP Archive analysis:

1. Long-running JavaScript on the main thread — particularly framework hydration (React, Vue, Angular) where Evaluate Script tasks exceed 50ms.
2. Tag managers with too many tags — a single Google Tag Manager container with ~18 tags can multiply Total Blocking Time roughly 20× relative to a clean baseline (Chrome Aurora team measurement). This is the largest single category of avoidable INP cost on most marketing sites.
3. Chat widgets (Intercom, Drift, Tawk, HubSpot) that render UI and attach event handlers before the user interacts.
4. A/B testing scripts (Optimizely, VWO) that fire synchronously during page load and on interactions; anti-flicker holds cost both LCP and INP.
5. Heatmap and session-recording tools (Hotjar, FullStory, Microsoft Clarity).
6. Heavy framework hydration with no code-splitting — bundles >300 KB compressed.
7. Untransferred work that could run inside requestIdleCallback or off the main thread.
8. Synchronous input handlers performing layout-thrashing DOM reads and writes inside the event callback.

Source: Chrome Aurora team and Chrome DevRel write-ups on INP optimisation patterns; HTTP Archive 2025 Web Almanac (Third Parties + Performance chapters). Confidence: Industry-consensus.

WordPress's INP pass rate is 85.9% — INP is not the typical WordPress failure mode; LCP and TTFB are.

CLS killers (catalogue)

Ranked by observed frequency:

1. Images and embeds without explicit width/height or aspect-ratio — the browser allocates wrong space, then reflows.
2. Web fonts with font-display: swap when the swap font has different metrics from the fallback (FOUT-induced shift). Web Almanac 2025: the median site uses 4 fonts; only 0.5% use the CLS-safe font-display: optional.
3. Late-loading ads, embeds, cookie banners that inject above-the-fold content.
4. App-injected banners on Shopify and similar (promo bars, free-shipping bars).
5. Dynamically-loaded above-the-fold content from JavaScript (consent managers, geo-personalisation).

Remediation pattern: reserve space with aspect-ratio or min-height → use font-display: optional with preload, or swap plus size-adjust metric overrides → place cookie banners outside the document flow (bottom-anchored fixed) → audit every plugin/app that can inject above-the-fold DOM.

Revenue and bounce evidence (the case-study cluster)

The published evidence linking page-performance changes to revenue, conversion, and bounce-rate movements spans more than a decade. The numbers below are quoted verbatim from the original sources; the confidence label on each indicates how cleanly performance is isolated from confounds.

Walmart (2012, foundational)

Devon Auerswald, Walmart Web Performance presentation, 2012:

"100 ms improvement = up to 1% incremental revenue; 1 s improvement = up to 2% increase in conversion."

Source: Walmart Web Performance presentation, 2012 (foundational; cited continuously across the performance community since). Confidence: Verified.

The first widely-disclosed e-commerce performance-revenue measurement at scale. The magnitudes (1% revenue per 100ms) became the default mental model in the industry, and have held up directionally across the Vodafone and Deloitte/Google replications.

BBC (web.dev, 2024)

"BBC lost an additional 10% of users for every additional second the site took to load."

Source: https://web.dev/learn/performance/why-speed-matters Confidence: Verified.

The cleanest "speed loss curve" stat in the public literature. The 10%-per-second loss is roughly linear over the relevant range (1–5 seconds), which makes it a useful rough calculator: if LCP is 4 seconds and the median is 2.5, roughly 15% of users are lost to that 1.5-second penalty.

Pinterest PWA rebuild (2017, foundational)

Pinterest's mobile-web rebuild as a Progressive Web App delivered 60% faster loads, +40% conversions, +40% time spent, and +44% ad revenue.

Source: Addy Osmani, "A Pinterest PWA Performance Case Study" (2017, foundational; results replicated in 2018–2025 case studies of similar mobile-web rebuilds). Confidence: Verified.

The strongest single-site evidence that mobile performance investment compounds into revenue at scale. Pinterest at the time had the engineering resources to instrument it cleanly. SMB scale will not see 44% ad-revenue lifts (no ad revenue), but the conversion-and-engagement direction generalizes.

Vodafone (2021)

See Measurement frameworks for small-business websites. Vodafone's web team ran a true A/B test: a 31% LCP improvement produced an 8% lift in sales — one of the cleanest published performance-only attribution studies.

Renault (2021)

Renault analyzed 4 months of data across multi-million sessions. A 1-second LCP improvement was associated with -14 percentage points of bounce rate and +13% conversions.

Source: web.dev/case-studies/renault (2021). Methodology: linear regression on multi-month real-user data. Not a true A/B test; pre/post or regression-based. Confidence: Medium — performance correlation is well-isolated but not in an A/B sense.

Good for the general "faster pages, higher conversion" argument; not as airtight as Vodafone or Rakuten 24.

Rakuten 24 (2022)

Rakuten 24 (Japan) ran an A/B test in 2022 optimizing Core Web Vitals: CLS improved 92.7%, FID 7.95%, FCP 8.45%, TTFB 18.03%, and an ~0.4s LCP gain.

Outcome: +33.13% conversion rate, +53.37% revenue per visitor, -35.12% exit rate.

Source: web.dev/case-studies/rakuten (2022). Methodology: true A/B test — the only change between variants was performance. Confidence: High (performance is cleanly isolated).

One of the strongest "performance changes only" lift numbers in published literature. Pairs with Vodafone and redBus for a three-case A/B-test cluster.

redBus (2023)

redBus (India) optimized the date-change interaction on the search page, improving INP by ~72%. Outcome: +7% sales, with 6–7.2% CR uplift across operating countries.

Source: web.dev/case-studies/redbus-inp (2023). Methodology: RUM before/after on the key flow; not a true A/B, but performance was the isolated change. Confidence: Mostly yes — high attribution because the affected flow and code change are documented.

The cleanest published INP-specific lift number. Use when arguing for INP investment specifically, not generic "speed."

Deloitte / Google — "Milliseconds Make Millions" (2020)

Deloitte Ireland and Google measured 4 industry-wide site-speed metrics and found that a 0.1-second improvement correlated with:

• +8.4% retail conversions
• +9.2% average order value (retail)
• +10.1% travel conversions

Source: "Milliseconds Make Millions" — Deloitte Ireland / Think with Google PDF, 2020 (foundational, still widely cited). Confidence: Verified (Deloitte primary).

The most-cited industry-wide study of site speed → revenue. Magnitudes are conservative (compared to individual-site studies like Vodafone, which show larger lifts), which makes it more credible as a baseline estimate for SMB contexts.

Caveat for SMB application: none of the foundational performance-revenue studies (Vodafone, Pinterest, Walmart, BBC, Deloitte) measured small-business marketing sites. Extrapolating to a regional SMB such as a dental practice is plausible but unproven. Treat the magnitudes as directional indicators that the relationship exists, not precise predictions for any specific site.

The Economic Times (2021)

The Economic Times (India) brought LCP from 4.5s to 2.5s and CLS from 0.25 to 0.09 (passing all CWV thresholds). -43% bounce rate site-wide.

Source: web.dev/case-studies/economic-times-cwv (2021). Methodology: pre/post CrUX field data, AMP + non-AMP. Partial — not a true A/B. Confidence: Medium.

Companion case — Economic Times INP (2023): INP reduced ~4× via 30× TBT reduction + Next.js migration → -50% bounce, +43% pageviews on topic pages. Useful for the "INP investment pays off" argument specifically.

Akamai / SOASTA (2017, ~10B visits)

Akamai's 2017 State of Online Retail Performance analyzed approximately 10 billion user visits across top online retailers.

• 100ms delay → -2.4% desktop, -7.1% smartphone, -3.8% tablet conversions.
• 2-second delay → -36.5% desktop, -26.2% mobile, -25.1% tablet conversions.

Counterintuitive finding: desktop users were more conversion-sensitive, but mobile users were more bounce-sensitive — mobile users leave faster; desktop completers abandon further down the funnel.

Source: Akamai State of Online Retail Performance (ir.akamai.com / akamai.com newsroom). Confidence: High for the measurement; dated for application — cite as a historical retail e-commerce benchmark from 2017, not a universal current law. The widely repeated "1-second delay = 7% conversion loss" myth derives from this study and is misapplied to non-retail contexts (see the unverifiable-stats register below).

Google neural-net (Think with Google, 2018)

Google's neural-net analysis of 11 million mobile landing pages (Think with Google, 2018, 213 countries, 90% prediction accuracy):

• Mobile load 1s → 3s: +32% bounce probability.
• Mobile load 1s → 5s: +90% bounce probability.
• Mobile load 1s → 10s: +123% bounce probability.

Confidence: High for the general curve, medium for exact numbers (baselines have shifted since 2018). Source: Think with Google, Mobile Page Speed Benchmarks, 2018.

The cleanest single non-vendor "bounce vs. load time" curve in the literature. Quote with the year so the audience knows it is directionally honest, not a 2026 number.

Mobile CWV pass rates 2025 and the 3-second abandonment threshold

Mobile CWV pass rates — the structural baseline

Per the HTTP Archive Web Almanac 2025 (using July 2025 CrUX):

• 48% of mobile sites pass all three Core Web Vitals (LCP, INP, CLS).
• 56% of desktop sites pass.
• Year-over-year mobile pass rate: 36% (2023) → 44% (2024) → 48% (2025).

Source: almanac.httparchive.org/en/2025/performance. Confidence: Verified.

WordPress specifically

• 43.44% of WordPress mobile origins pass CWV as of June 2025 per the CrUX Technology Report (summarized by Search Engine Journal); updated late-2025 data has it at ~45–46% (HostingStep; Ben Ryan).
• Failure mode is NOT INP (85.9% pass rate) but LCP and TTFB — only ~32% of WordPress sites have good TTFB, driven by shared hosting (CoreDash analysis; SEJ).

Confidence: Verified.

Page-experience weight in the December 2025 core update

Google's December 2025 core update increased the weight of page-experience signals beyond their previous tie-breaker role. Source: Ben Ryan / Search Engine Journal coverage. Confidence: Industry-consensus.

The 53% mobile-abandonment threshold — still canonical

The canonical source is Google/DoubleClick's "The Need for Mobile Speed" (September 2016, Think with Google):

• n=3,700 aggregated Google Analytics samples, March 2016.
• "53% of visits are abandoned if a mobile site takes longer than three seconds to load."

Sources: Think with Google; restated in Google AdSense Help. Confidence: Verified.

The 2016 study remains the single most-cited number on mobile abandonment because nothing comparably-sized has replaced it. Treat as Verified for the 3-second threshold; the underlying user behaviour has likely intensified, not weakened, in the intervening decade.

Zero-click reality

58% of Google searches result in zero clicks to external websites, per SparkToro research. WSI's 2025 AI-era visibility report restates:

"Zero-click searches now account for nearly 60% of Google activity, signalling a tectonic shift in how people engage with search results."

Sources: SparkToro; WSI 2025. Confidence: Verified.

Caveat — field-sample audit not yet conducted

A field sample of 25 Ontario GC websites is a research item that remains outstanding and should be run before publishing a consumer-facing article. Placeholder estimate based on platform mix + CrUX baseline: 60–70% likely to fail CWV on mobile [Directional — derived from HTTP Archive 2025 WordPress pass rates, adjusted upward for builder-heavy construction vertical].

Web Almanac 2025 — mobile page weight (2,559 KB)

Median mobile home-page weight reached 2,559 KB in July 2025; desktop 2,862 KB. That is +7.8% YoY. Average mobile Total Blocking Time: 1,916 ms — almost 10× the 200ms best-practice target.

Source: HTTP Archive Web Almanac 2025 (Page Weight + Performance chapters); also cited via corewebvitals.io. Confidence: Verified.

The compounding-tax framing: the average page grows heavier each year, not lighter. Plugin updates, embeds, and successive "modernisations" all add bytes. A site that defends a 1.0 MB budget remains materially differentiated from the 2.56 MB median, and the gap is widening rather than narrowing — Web Almanac data shows +7.8% median weight year-on-year through 2025.

Web Almanac 2025 CWV pass rates

The Web Almanac 2025 CWV pass-rate data is referenced in summary above (48% mobile, 56% desktop, 77% of pages achieving good INP in June 2025). Per-metric breakdowns on mobile from the HTTP Archive 2025 Web Almanac Performance chapter are 62% LCP good, 77% INP good, 81% CLS good — LCP is the gating metric for the median site.

Source: HTTP Archive Web Almanac 2025, Performance chapter — almanac.httparchive.org/en/2025/performance. Confidence: Verified.

Platform CWV pass rates (November 2025 CrUX Tech Report)

Mobile CWV pass rate, all-three-thresholds, per the HTTP Archive CrUX Technology Report, summarized by Search Engine Journal December 2025:

CMS	Nov 2025	Jun 2025	YoY direction
Duda	84.87%	83.63%	up
Wix	74.86%	70.76%	up
Shopify	not in Nov SEJ breakout	75.22%	(relatively flat)
Squarespace	70.39%	67.66%	up
Drupal	63.27%	59.07%	up
Joomla	56.92%	–	up
WordPress	46.28%	43.44%	up but slowest gainer
Global (all origins)	~48%	~49%	(flat)

Trend (June 2024 → July 2025 → November 2025):

• WordPress: 40% → 45% → 46.28%. Gain of ~+4 pp/year — half of Wix's +14 pp YoY and Duda's +11 pp YoY.
• 2025 Web Almanac quote (verbatim): "More extensible platforms improved less. WordPress and Drupal each gained around 4% year over year."

Confidence: Verified (CrUX is the authoritative public dataset; SEJ's SQL methodology is disclosed).

Reading the numbers correctly: a WordPress site is roughly 1.8× less likely to pass CWV than a Duda/Wix/Squarespace/Shopify site (mobile, November 2025). The closed-CMS lead is widening, not narrowing. The earlier June 2025 numbers — Duda 83.63%, Shopify 75.22%, Wix 70.76%, Squarespace 67.66%, Drupal 59.07%, WordPress 43.44%, summarised by Roger Montti for Search Engine Journal in August 2025 ("2025 Core Web Vitals Challenge: WordPress Versus Everyone") — are preserved above for historical comparison.

Causal decomposition of the WordPress gap

The best decomposition the public data supports for the WordPress CWV gap (May 2026):

Factor	Share of gap	Evidence	Confidence
Hosting / TTFB	~40–50%	Only 32% of WordPress origins have good TTFB (CoreDash citing CrUX, June 2025); moving from shared to managed hosting routinely flips LCP from poor to good with no other change	High
Page builder bloat (Elementor, Divi, WPBakery)	~20–30%	2025 Web Almanac: builder optimisations "do not completely erase the performance gap." Elementor → Gutenberg migrations consistently show 1.5–2.5s LCP reductions	Directional
Plugin sprawl / third-party scripts	~15–20%	2024 Web Almanac Performance chapter identifies third-party scripts (behaviour tracking, consent providers, CDNs) as the primary cause of long Presentation Delay (poor INP)	Directional
WordPress core	~5–10%	Gutenberg-only sites on managed hosting can match modern-stack pass rates in the field	Directional
Theme	~5–10%	Heavy theme CSS/JS, render-blocking stylesheets	Directional

Source: synthesis across the 2024 + 2025 HTTP Archive Web Almanac Performance chapters, CoreDash CWV analysis of June 2025 CrUX data, and the November 2025 CrUX Technology Report. Confidence: Industry-consensus on direction; Directional on exact share split.

A well-optimised Gutenberg-only WordPress site on managed hosting can match an Astro-on-Cloudflare site in the field — the WordPress CWV deficit is overwhelmingly an artefact of the dominant deployment configuration (shared hosting + heavy builders + plugin sprawl), not of WordPress core itself.

Hosted-builder ceiling

Closed-CMS platforms (Duda, Wix, Squarespace, Shopify) pass CWV at higher rates in part because their asset pipelines, hosting, and bfcache support are controlled by the vendor. The trade-off is a ceiling: DIY hosted builders stop being sufficient when a business needs customisation, advanced SEO control, performance tuning beyond the vendor defaults, complex commerce or memberships, or data portability.

Source: practitioner consensus across DIY-platform migration write-ups (Duda → WordPress, Wix → Webflow, Squarespace → Shopify case histories). Confidence: Industry-consensus.

Businesses whose roadmap implies any of these needs incur a future migration cost — frequently to WordPress, Webflow, or Shopify — that is rarely priced into the original platform decision.

March 2026 LCP-tightening rumour (refuted)

Claim being refuted: Multiple SEO blogs (Digital Applied, MonsterMegs, IdeaFueled, ClickRank, Mewa Studio, w3era) claim Google tightened the LCP "Good" threshold from 2.5s to 2.0s with the March 2026 core update, citing a "Search Central blog post published March 18, 2026."

Refutation:

• The cited Search Central blog post does not exist. Google's official documentation, last updated December 10, 2025, still states LCP ≤ 2.5s.
• Search Engine Roundtable (Barry Schwartz, April 8, 2026): Google's only public communication about the March update was a Search Status Dashboard note. No threshold change announced.
• Whitehat SEO and BuildMVPFast both state explicitly that thresholds have not changed.
• The original Digital Applied post that started the rumour is the only source; everything else traces back to it.

Source: Google Search Central docs (current); Search Engine Roundtable April 2026; Whitehat SEO + BuildMVPFast confirmations. Confidence: Verified refutation.

Operational guidance: the 2.0s / INP-elevation claim should be treated as unconfirmed industry rumour until Google publishes otherwise; the 2026-confirmed thresholds table above remains the actual reference. The rumour is a textbook case for the citation discipline of requiring author + institution + date + URL on every quantitative claim — applying that discipline catches that the cited Search Central post does not exist.

Unverifiable stats catalogue — speed/conversion claims Candid will not quote

The following statistics circulate widely but are not supported by primary evidence in 2026. A disciplined practitioner does not quote them in sales conversations or marketing copy:

1. "A 1-second delay reduces conversions by 7% across all sites." Traces to Akamai/SOASTA 2017 retail data. Defensible as a retail e-commerce benchmark from 2017, not a universal law. For a regional lawyer or HVAC contractor, this specific elasticity has not been demonstrated.

2. "Amazon loses 1% revenue per 100ms of latency." Reportedly Amazon internal research circa 2006 (Greg Linden). No public methodology, ~20 years old, Amazon-scale operations.

3. "Core Web Vitals doubled in ranking importance" / "Google made CWV twice as important." Cited by some marketing blogs but not in any Google primary statement. Google has consistently described CWV as one of many page-experience signals, tiebreaker class. See the CWV-tiebreaker section above.

4. "53% of mobile users abandon in 3 seconds" applied to 2026 baselines. The original study (Google/DoubleClick/SOASTA, 2016) is nearly a decade old. Networks, devices, and tolerance baselines have shifted. Quote it as historical context, not a current 2026 number.

5. "Faster site = better AI Overview citation rate." No defensible primary research supports this as of May 2026. The one direct analysis at scale — Dan Taylor's January 2026 Search Engine Land study — found the opposite of what an SEO marketer would assume, as catalogued in the CWV-as-AI-citation-gate section above.

6. Agency "case studies" claiming "redesign → +40% conversions" as performance wins. A redesign confounds content, layout, CTA copy, navigation, trust signals, and performance. Cannot attribute the lift to speed without an isolated A/B test. Reject these.

7. Conversion-rate elasticity for "professional services" specifically. No major isolated study (Deloitte, Akamai, web.dev cases, Portent, Contentsquare) breaks out elasticity for lawyers, accountants, dentists, or consultants in a defensible way. Extrapolate cautiously from Deloitte lead-gen and Portent B2B — but flag as "general B2B," not "professional services-specific."

Why this rule exists: sales conversations are won on credibility. A quoted statistic that does not survive scrutiny loses the room. See Agency methodology for small-business website projects for the broader sourcing discipline.

Sources and confidence

Confidence labels appear inline above. The aggregate source map for this page:

• Verified — primary Google documentation: developers.google.com/search/docs/appearance/core-web-vitals (2025-12-10); web.dev/articles/lcp; web.dev/articles/defining-core-web-vitals-thresholds; web.dev/learn/performance/why-speed-matters; web.dev/case-studies/* (Renault, Rakuten, redBus, Economic Times); Search Central documentation on page-experience signals.
• Verified — primary industry research: HTTP Archive Web Almanac 2025 (Performance + Page Weight chapters); CrUX Technology Report; Walmart 2012 (Auerswald); Pinterest PWA case study (Addy Osmani, 2017); Deloitte Ireland / Think with Google "Milliseconds Make Millions" 2020; Google/DoubleClick "The Need for Mobile Speed" 2016; Think with Google Mobile Page Speed Benchmarks 2018; Akamai State of Online Retail Performance 2017.
• Verified — independent ranking-factor studies: Perficient 17-week study (June–September 2021, n=1,188 keywords); Advanced Web Ranking March 2022 (n=3M pages); Backlinko 208,085-page study (2021).
• Industry-consensus: Search Engine Journal December 2025 coverage of CrUX Tech Report and December 2025 core update; HostingStep and Ben Ryan late-2025 WordPress CWV updates; CoreDash WordPress TTFB analysis.
• Verified refutation: Search Engine Roundtable April 8, 2026 (Barry Schwartz); Whitehat SEO and BuildMVPFast confirmations that the March 2026 LCP-tightening rumour is unsupported.
• Single-source / Directional: the 60–70% Ontario-GC CWV-fail estimate (pending 25-site field audit); SparkToro / WSI 2025 zero-click figure (58% of Google searches).
• Unverifiable / rejected: the seven claims in the unverifiable-stats catalogue above.

See Agency methodology for small-business website projects and Conversion rate — realistic expectations for small-business websites for the upstream research briefs that informed this synthesis.