AmazingCo performance crisis dashboard: mobile speed issue causing CPA spike across all markets

Active incident — mobile performance crisis causing CPA collapse across all 5 markets
Root cause confirmed: May 1st code deployment degraded mobile LCP, TBT, and CLS globally. This is a technical infrastructure issue, not an ads or audience problem.
CPA impact — May 31 → June 1
Australia
+57%
$54 → $85
United States
+16%
$69 → $80
Canada
+22%
$63 → $77
United Kingdom
+26%
$40 → $50
New Zealand
↑ spike
$50 → $75 (7-day avg)
CPA target
$55
All markets breached
Worst drop-off rate
33%
new-customer page
Max TBT observed
14s
vs 200ms target
✓ Hypothesis validated — May 1st is the origin
Your hypothesis is correct. The Lighthouse data and Google Search Console INP spike both point to May 1st as the trigger date. The 3-week lag to the June 1st CPA collapse is explained by Google's 28-day rolling CrUX average and Meta/Google ad algorithm learning windows — not a delayed problem start.
Incident timeline
May 1, 2026
Code deployment breaks mobile performance Root cause
A global script, GTM tag, or theme change is pushed. Mobile LCP and INP degrade across all pages simultaneously. No immediate CPA signal.
May 1–20
Silent conversion drain
Mobile users click ads, hit slow pages, abandon. Ad algorithms begin bidding more aggressively to compensate for lower conversion rates, slowly inflating spend.
May 21
First visible CPA spike Warning
AU: $65 → $77 overnight. US: $65 → $70. UK and CA also tick upward. CrUX 28-day average now reflects the damage from May 1st.
May 25–27
Secondary degradation event Critical
Search Console graphs show a second upward step in LCP and INP. Likely a second deployment or config change. CPAs pushed to new highs.
June 1, 2026
Full account collapse Active
28-day data cycle fully propagated. AU +57%, US +16%, CA +22%, UK +26%. Ad algorithms have re-priced based on the new, broken conversion rates.
Page speed — Lighthouse mobile
LCP
12–25s
Target: <2.5s
Across all 3 pages
TBT
7–14s
Target: <200ms
Browser frozen
CLS
0.891
Target: <0.1
Page jumping wildly
The TBT of 7–14 seconds means the mobile browser is completely frozen when a user taps "Book Now". The button appears broken. Users leave.
Ghost traffic — clicks vs arrivals
PageChannelLostStatus
mystery-picnicMeta15.7%Healthy
new-customerMeta33.3%Critical
us-fathersdayMeta11.2%Healthy
mystery-picnicGoogle28.7%Warning
new-customerGoogle33.8%Critical
us-fathersdayGoogle27.1%Warning
On /new-customer, 1 in 3 paid clicks never reaches the page. That budget is being vaporised before any conversion is possible.
Competing theories — all ruled out
TheoryRuled out because
Audience saturationSimultaneous cross-market spike — saturation is gradual and market-specific
Creative fatigueCTR remains healthy — people are clicking. Failure is post-click
Seasonal shiftSeasons don't cause 33% ghost traffic before anyone sees the page
Wrong audienceOrganic bounce rates are also high — the page is the problem, not the ad audience
Platform settingsBidding is a symptom, not a cause. Optimising bids on a broken site doesn't fix the site
Action plan
Short term — this week
1. Audit deployment logs: April 30 – May 1, May 25–27
2. Defer/remove heavy JS & third-party scripts (Freshchat, GTM)
3. Convert hero images to WebP, add fetchpriority="high"
4. Hardcode image/video dimensions to fix CLS 0.891
5. Re-run Lighthouse after each fix — target LCP <4s
Long term — 30–90 days
1. Performance budgets in CI/CD pipeline
2. Real-time Core Web Vitals monitoring (SpeedCurve/Calibre)
3. Full third-party script audit — defer non-essential
4. Mobile-first paid landing pages, minimal dependencies
5. Quarterly CPA vs CWV correlation review
What will not fix this
Increasing ad budget
Refreshing creative or switching audiences
Adjusting bidding strategies