Skip to main content
Card class: Card

At a glance

Composite regression timeline across all CWV trends, surfaces the dates of significant CWV regressions across the LCP / INP / CLS / TTFB / FCP trends in a single chronological view. The “what changed when” diagnostic surface: instead of reading 5 separate trend cards, this card shows every regression event ranked by severity and decorated with the affected metric. Critical for incident-response retrospective: when leadership asks “why did our pass rate drop?”, this card produces the answer in one screen.
What it countsChronological timeline of regression events detected across all CWV trends. Each event includes: date, affected metric (LCP / INP / CLS / TTFB / FCP / score / pass rate), magnitude (delta), severity classification, and inferred cause when correlatable with deploy / content events.
Sample typeField data for CrUX-derived metrics; lab data for score-derived events. The composite uses both data sources to surface a unified timeline.
Detection criteriaA regression event is recorded when: (a) field metric moves > +2x typical run-to-run variance over a 7-day rolling window; (b) lab metric moves > 5 points / 200ms / 0.05 within a single audit; (c) threshold crossing: the metric crosses a Google-defined band boundary (e.g. CWV pass rate from above to below 75 percent).
Severity classificationCritical: threshold crossing (passing → failing band). High: regression > 30 percent of band width without crossing threshold. Medium: regression > 15 percent without crossing. Low: persistent drift beyond noise but small in magnitude.
Inference of causeWhen deploy log or CMS audit log is available, the timeline cross-references regression dates with shipped changes. High-confidence inferences (deploy on the same day, single-metric pattern matching the change-type): annotated as “likely cause: X”. Lower-confidence: surfaces candidate events without claiming attribution.
Why this card mattersThe trio of TTFB + FCP + LCP trends already reveals regression class (server vs render-blocking vs LCP-element). This card aggregates the trio + INP + CLS trends + score trend + pass rate trend into a unified timeline. Saves 10-15 minutes of cross-card cross-referencing per incident-response investigation.
The “leadership read” framingWhen leadership asks “why is our CWV failing?”, the card produces a 2-3 line answer: “Three regressions in the last 6 months: Feb 26 hero carousel deploy (+460ms LCP), Mar 26 marketing-stack expansion (+660ms LCP, +60ms INP), Apr 26 BFCM imagery (+580ms LCP, +0.02 CLS). Recovery requires reversing each contribution; estimated 6-week cycle.”
Currencyn/a, timeline view.
Time windowT-180D rolling timeline (default 6-month look-back).
Alert triggerNew critical-severity event detected (threshold crossing).
Sentiment keynull (composite view).
Rolesowner, operations

Calculation

Calculated automatically from your Website Performance (PageSpeed + CrUX) data. See the At a glance summary above for what the metric tracks and the worked example below for a typical reading.

Worked example

A UK-based BigCommerce fashion store, regression timeline over 6 months ending Wednesday 15 May 26.
DateSeverityAffected metricsMagnitudeInferred cause
14 Jan 26HighLCP, FCPLCP +260ms, FCP +280msKlaviyo popup CSS landed as render-blocking external stylesheet
03 Feb 26CriticalPass rate (84% → 78%)-6pp pass rate; LCP +460msHero carousel deploy (heavier images, 4 PNGs at 2.8MB)
17 Feb 26CriticalPass rate (78% → 72%)Threshold crossing below 75%Compounded with Klaviyo popup, pushed below threshold
12 Mar 26CriticalTTFB, scoreTTFB +260ms, score -8BFCM-prep dynamic-pricing widget disabled CDN caching
18 Mar 26HighINPINP +60msKlaviyo SDK update added behavior tracking
25 Mar 26HighINPINP +80msFilter widget refactor introduced long tasks
09 Apr 26HighLCPLCP +580msBFCM hero campaign images uploaded unoptimised
22 Apr 26MediumINPINP +40msHotjar session recording added
Most recentFailingAll composite64.8% pass rateSustained sub-75% for 3 months
What the timeline is telling us:
  1. Eight distinct regression events identified over the 6-month window, with two critical-severity threshold crossings (Feb 03 and Feb 17). The cumulative effect is the current 64.8% pass rate.
  2. Two critical events were missable in real-time because each individual event seemed minor. Feb 03 was a single hero deploy; Feb 17 was a “stabilisation” of the previous deploy. The threshold crossing happened because cumulative effect crossed 75%, not because any single deploy was catastrophic.
  3. The Mar 12 critical TTFB regression had a different shape. Single dominant cause (cache config change), single dominant metric (TTFB), single dominant severity (critical). Easier to attribute and easier to fix than the cumulative-drift pattern.
  4. Clear leadership story: “We had 8 regression events over 6 months; 3 critical, 4 high, 1 medium. Two threshold crossings put us in CWV failing band. The dominant causes were image-related (hero carousel, BFCM imagery) and cache-related (BFCM dynamic-pricing widget). Recovery requires reversing the image and cache changes, estimated 6-week cycle, expected to cross back above 75% threshold by week 4-5.”
  5. Recovery prioritisation by impact × ease-of-fix:
    • Mar 12 cache config: highest leverage, easy fix (config rollback). Tackle first.
    • Apr 09 BFCM imagery: high leverage, mechanical fix (image format conversion). Tackle second.
    • Feb 03 hero carousel: high leverage, requires re-export workflow. Tackle third.
    • Jan 14 Klaviyo CSS: medium leverage, easy fix (defer). Tackle in parallel with above.
    • Mar 18, Mar 25, Apr 22 INP events: medium leverage, more invasive (refactor). Tackle in week 3-4.
  6. Defence going forward: enable critical-severity alerts (threshold crossings) + high-severity alerts (>30% band-width regression). The timeline becomes a continuous monitoring surface rather than a retrospective tool.
The diagnostic flow:
  1. Read top-3 by severity. Critical events first, then high, then medium.
  2. Cross-reference deploy / content / vendor logs for each event date.
  3. Plan recovery sequence by impact × ease-of-fix.
  4. Re-enable alerts for ongoing monitoring.
Rapid-response playbook:
Time horizonAction
First 1 hourIdentify critical-severity events; understand each cause.
First weekBegin recovery on highest-leverage easiest-fix events.
First monthField metrics begin reflecting cumulative recovery.
Quarter onwardContinuous monitoring via alerts.

Sibling cards merchants should reference together

CardWhy merchants reach for it
crux_lcp_trendLCP-specific trend; this card aggregates.
crux_inp_trendINP-specific trend.
crux_cls_trendCLS-specific trend.
crux_ttfb_trendTTFB-specific trend.
crux_fcp_trendFCP-specific trend.
crux_pass_rate_trendComposite pass rate trend.
psi_score_trendLab score trend.
psi_biggest_regressionPer-URL regression detection.

Reconciling against the vendor’s own dashboard

Where to look: This is a Vortex IQ-derived composite card; no direct external equivalent. The closest comparable view is reading GSC’s CWV trend chart while cross-referencing your deploy log manually, that’s exactly the work this card automates. Why the timeline may include events external tools wouldn’t see:
ReasonWhat it means
Sub-threshold but persistent regressions. Smaller deltas that don’t trigger a single-trend alert but accumulate.The composite catches drift; single-trend alerts catch step changes.
Cross-metric correlation. TTFB regression on date X plus FCP regression on date X = single root-cause event.Surfaces unified incidents that single-trend views fragment.
Cross-connector reconciliation: primarily internal (aggregates all sub-trend cards). Quick rule for support tickets: if a merchant says “Vortex IQ shows a critical event but GSC doesn’t flag it”, the most common cause is GSC’s threshold being more lenient than the composite’s regression-detection threshold. Vortex IQ surfaces drift earlier; GSC reflects ranking-impact events later.

Known limitations / merchant FAQs

Why does this card label some events as “high-confidence cause” and others as “candidate cause”? When deploy logs and CMS audit logs are integrated, the timeline matches regression dates with shipped changes. High-confidence: same-day deploy + matching change-type pattern (e.g. CSS deploy + FCP regression). Candidate: multiple shipped changes on the same day, can’t disambiguate. No inference: regression date doesn’t match any logged change. My timeline shows 8 events but I only remember 3 deploys. Why? Some regressions trace to non-deploy events: vendor-side updates (Klaviyo, Tidio pushed new versions without merchant action), content uploads (marketing replaced hero images via CMS), or infrastructure changes (CDN config drift). The timeline catches all regression events, including the ones the merchant didn’t directly cause. How far back does the timeline go? Default 6-month look-back. Extends to 12 months for deeper investigation if configured. Beyond 12 months, CrUX historical data quality degrades for some metrics and the timeline becomes less useful. Should I act on every event in the timeline? Prioritise by severity × ease-of-fix. Critical events (threshold crossings) demand action; medium and low events are informational. The card surfaces all events; the recovery plan addresses only the most leverage-positive subset. Can I export the timeline for retrospective post-mortems? The Vortex IQ dashboard supports CSV export of the timeline. The Vortex Mind Pre-Launch Readiness report includes the timeline in its outputs. Why is sustained drift sometimes flagged as critical? When sustained drift compounds to push a metric across a band threshold (e.g. pass rate from 76% to 73%), the threshold crossing is critical regardless of whether any single event was large. The card flags the crossing date even when no single event would qualify on its own.

Tracked live in Vortex IQ Nerve Centre

When Did Speed Get Worse? is one of hundreds of KPI pulses Vortex IQ tracks across Website Performance (PageSpeed + CrUX) and 70+ other ecommerce connectors. Nerve Centre runs the detection layer; Vortex Mind investigates the cause when something moves; Ask Viq lets you interrogate any number in plain English. Start for free or book a demo to see this metric running on your own data.