Executive Summary — Core Web Vitals Performance Crisis
Impact on paid channel performance across AU · US · UK · CA · NZ  —  April to June 2026
Action required
What are Core Web Vitals? Core Web Vitals (CWV) are Google's standardised measurements of how fast and stable a webpage feels to a real user on their device. They measure three things: how quickly the main content loads (LCP), how much the page jumps around while loading (CLS), and how long before a user can actually interact with the page (TBT/INP). Google collects this data from real Chrome users over a rolling 28-day window — a dataset called CrUX — and uses it directly to score landing page quality in its ad auctions. A poor CWV score is not just a technical problem: it is a direct signal to Google and Meta that your landing pages deliver a bad user experience, and both platforms respond by making your ads more expensive to run.

What we measure and why it matters

LCP — Largest Contentful Paint
How fast does the main content load?
Measures how long it takes for the biggest visible element (hero image, headline) to appear. If this is slow, users see a blank or partial page and bounce before engaging.
Good <2.5sNeeds work 2.5–4sPoor >4s
TBT — Total Blocking Time
Can users interact with the page?
Measures how long the browser is "frozen" running JavaScript before a user can tap, scroll, or click. High TBT makes a page feel broken — especially on mobile where CPU is limited.
Good <200msNeeds work 200–600msPoor >600ms
CLS — Cumulative Layout Shift
Does the page jump around?
Measures unexpected visual shifts — buttons moving, text jumping — while the page loads. A high CLS causes users to mis-tap or lose their place, destroying trust instantly on mobile.
Good <0.1Needs work 0.1–0.25Poor >0.25

How poor CWV directly penalises paid performance

Step 1
Poor mobile CWV
LCP, TBT & CLS all critical on mobile landing pages. 926 URLs rated "poor" — 61% of all tracked mobile URLs.
Step 2
Quality Score drops
Google reads CrUX data and lowers landing page Quality Score. Lower QS = higher minimum CPC to win the same auction position.
Step 3
CPC & CPM inflate
Google charges more per click. Meta's relevance score also drops, inflating CPM. Both platforms penalise simultaneously — confirmed in our data.
Step 4
CPA blowout
Paying more per click for the same (or worse) conversion rate = CPA spikes. Smart bidding then overspends chasing unreachable targets, compounding the damage.

Why mobile specifically — and why the 28-day lag?

Mobile accounts for the majority of paid traffic and is disproportionately affected by slow page loads — mobile devices have less CPU and memory than desktop, so JavaScript-heavy pages that perform acceptably on desktop can be critically broken on mobile. Desktop recovered on April 23 when an engineering fix was deployed — but that fix was never applied to mobile. The result: desktop Quality Scores recovered while mobile continued to degrade, unseen.

The 28-day lag between the May 1 regression and the June 1 CPA collapse is explained by how Google collects CWV data. CrUX is a rolling 28-day average of real Chrome user sessions — it does not update instantly. For the first four weeks after the regression, the CrUX average was still partially weighted by clean pre-regression data, masking the true severity. By June 1, the window contained 28 full days of degraded mobile data, and Quality Scores collapsed across all markets simultaneously.

Key findings from the data

5 markets, 2 platforms, 1 cause
AU, US, UK, CA and NZ all experienced CPA spikes simultaneously on Jun 1. Google CPC and Meta CPM both inflated in the same window. Cross-platform, cross-market synchronisation rules out creative fatigue, audience saturation, seasonal factors, or bidding strategy as root causes.
Campaigns launched into a broken environment
Mother's Day switch-off (May 10), Flash Sale (May 11–14), Free PT Melbourne (May 16), NCA (May 18) and Father's Day (May 25) all ran during the silent degradation window. The Flash Sale artificially suppressed AU CPA — masking the underlying penalty. Father's Day launched 4 days before the full collapse and recorded zero GA4 conversions.
The fix exists — it just wasn't applied to mobile
Desktop CWV recovered on Apr 23, confirming the engineering team identified and resolved the root cause. The mobile branch was not updated. This is an incomplete deployment — not a new investigation — making the fix the highest-priority unblocking action.
Every day of delay extends the recovery window
Once the mobile fix is deployed, CrUX needs 28 days to re-average with clean data, followed by 2–4 weeks of smart bidding re-learning. Full CPA normalisation is estimated 6–8 weeks from fix date. Each day the fix is delayed pushes the entire recovery timeline out by one day.
Impact summary — blended CPA change since regression (May 1 → Jun 8)
US CPA
+51%
$53.71 → $81.23
CA CPA
+39%
$63.61 → $88.58
UK CPA
+29%
$32.95 → $42.61
AU CPA
+14%
$58.35 → $66.65
NZ CPA
+11%
$87.53 → $97.33
Mobile "poor" URLs
926
↑ from ~700 in March — 61% of total
Desktop "poor" URLs
174
Recovered Apr 23 — fix exists
US Meta CPM (Jun 1)
$20.20
vs $15.20 baseline (+33%)
CA Meta CPM (Jun 1)
$14.41
vs $9.07 baseline (+59%)
Blended CPA by market — CWV events & campaign overlay
Apr 23 — desktop fixed
May 1 — regression deployed
May 29 — CrUX 28d window closes
Jun 1 — CPA collapse
Campaign events
Key correlation: All 5 markets spike simultaneously on Jun 1 — exactly 28 days after the May 1 regression, matching Google's CrUX rolling window precisely. The AU Flash Sale (May 11–14) artificially suppressed CPA during the silent degradation window, masking the underlying penalty. Father's Day launched May 25 — 4 days before full collapse — and recorded zero GA4 conversions despite active spend.
Google CPC — auction penalty signal
Quality Score penalty
CPC climbs from mid-May as CrUX re-averages daily. AU hits $2.32 on May 15 — Quality Score is degrading before CPA visibly collapses. The penalty starts immediately at May 1; CrUX just smooths it over 28 days.
Meta CPM — delivery cost signal
Relevance score penalty
CPM surges from late May across all markets. US +33% and CA +59% on Jun 1. Google and Meta penalising simultaneously on the same date is the clearest possible evidence that the root cause is the landing page, not channel strategy or campaign decisions.
Event timeline — CWV, campaign launches & CPA correlation
DateEventMarketsData signalHypothesis
Apr 23Desktop CWV fixedAllCPA stable. Desktop Quality Score begins recovering.Fix applied to desktop branch only — mobile branch missed entirely.
May 1Code regression deployedAllCPA appears calm. Mobile LCP/TBT/CLS all degrade silently.CrUX has 28 days of clean data masking the damage. No visible signal yet in ads.
May 10Mother's Day switch-offAU, US, CA, NZCPA drops — normal post-peak seasonal normalisation.Budget reduction masks early CWV penalty. Appears campaign-driven, not technical.
May 11–14Flash sale (AU)AUAU CPA dips to ~$45–47 — flash sale conversion boost.Promotional lift masks CWV penalty in AU specifically. Pages are broken while CPA looks healthy.
May 16Free PT Meta Melbourne liveAUAU Meta CPM begins climbing. CPC also rising.New spend on already-degraded mobile pages. CPM inflation partly driven by poor Meta relevance score.
May 18NCA campaign liveAUAU CPA $49–53. Appears reasonable but masking CWV damage.Launching into degraded mobile UX. Conversion data from this period is unreliable for future tCPA targets.
May 25Father's Day goes liveUS, CA, UKUS/CA CPA climbing. CPM spiking. Zero GA4 conversions on FD page.FD launches 4 days before full CrUX collapse. Landing page broken on mobile from launch day.
May 29CrUX 28d window closesAllCPC/CPM both elevated. CPA about to collapse.Google's rolling average now fully reflects degraded mobile CWV. Quality Scores drop across all landing pages.
Jun 1CPA collapse — all marketsAllUS +51%, CA +39%, UK +29%, AU +14%, NZ +11% simultaneously.Simultaneous multi-market, multi-platform spike. Definitively rules out creative, seasonal and bidding causes.
Jun 3Free PT Melbourne offAUAU CPA remains elevated at $62–69.Campaign switch-off does not improve CPA — confirms driver is technical (CWV), not campaign spend.
Action plan — prioritised by impact Recovery timelines included
P1
Engineering: deploy mobile CWV fix — complete the Apr 23 desktop fix on mobile
The desktop recovery on Apr 23 proves the fix exists. This is an incomplete deployment, not a new investigation. Target LCP <2.5s, TBT <200ms, CLS <0.1 on all mobile paid landing pages. Every additional day adds more degraded CrUX data that must be overwritten post-fix.
Engineering28-day CrUX recovery clock starts on deploy dateBlocking all other recovery
P1
Paid: reallocate budgets to best-performing channel per market during recovery
US and CA showing severe Google CPA blowout (+51%, +39%). Shift budget toward Meta where CPA is lower. NZ historically favours Meta. Do not set aggressive tCPA targets while smart bidding is re-learning against degraded data.
Paid MediaImmediate — reduces daily overspend
P1
Paid: switch Google tCPA to Maximise Conversions during smart bidding re-learning window
tCPA targets set pre-regression are now unreachable. Smart bidding is overspending trying to hit impossible targets. Re-introduce tCPA targets 2 weeks post mobile fix, starting 20–30% above current observed CPA.
Paid MediaImmediate — stabilises auction behaviour
P1
Paid: pause or replace Father's Day landing page — zero conversions recorded
FD page launched May 25 directly into the CrUX degradation window and has recorded zero GA4 conversions across US, CA, UK despite active spend. Swap to a proven high-performing page until mobile CWV is fixed.
Paid MediaImmediate — stops confirmed wasted spend
P2
Engineering: add Lighthouse CI to deployment pipeline to prevent recurrence
No automated performance gate existed at deploy time. Lighthouse CI in CI/CD would have caught this regression before production — and before 28 days of CrUX data were contaminated. The mobile/desktop branch split in the pipeline also needs reviewing.
EngineeringPrevents recurrence — 1–2 sprint effort
P2
Analytics: invalidate NCA and FD conversion data collected during degraded period
Conversion data from May 1–present on degraded mobile pages is unreliable for tCPA target-setting. NCA (AU, live May 18) and Father's Day (US/CA/UK, live May 25) both launched into the degraded window. Reset tCPA baselines from clean post-fix data only.
AnalyticsStart immediately post-fix
P3
Engineering: isolated paid landing page architecture decoupled from CMS deployments
A lightweight static landing page layer — independently deployed and Lighthouse-tested — permanently decouples paid performance from product release cycles and prevents this class of incident recurring.
Engineering + Paid MediaLong-term — 4–6 week initiative
Estimated full recovery timeline: Mobile CWV fix deployed → 28 days for CrUX to re-average → 2–4 weeks smart bidding re-learning → full CPA normalisation in 6–8 weeks from fix deploy date. Every day of delay pushes this window out by one day.