At a glance
Lab Performance Score over time, daily time-series of the Lighthouse Performance Score (0-100) on mobile profile. The deterministic engineering trend that pairs with field-data trends (crux_pass_rate_trend) for the full picture. Lab data updates in real-time after each audit run, while field data has a 28-day rolling lag. Use lab trend for: build-pipeline gating, post-deploy verification, engineering iteration. Use field trends for: ranking-impact decisions, monthly executive reporting.
| What it counts | Time-series of the per-URL Lighthouse Performance Score across daily audit runs. Each data point is the score from that day’s audit (with 7-day rolling smoothing applied to reduce run-to-run variance). |
| Sample type | Lab data from scheduled Lighthouse audits. Field equivalent (closest analog): crux_pass_rate_trend. |
| Why lab trend updates faster than field trend | Lab measures one synthetic run; field needs 28 days of accumulated user data. A deploy fixing performance shows in lab trend within 24-48 hours; field trend takes 28 days for full reflection. |
| Run-to-run variance smoothing | Single-day lab scores fluctuate ±5-10 points due to Lighthouse run-to-run variance. The trend uses 7-day rolling smoothing to filter noise; sharp deploys still appear as inflection points within 3-7 days of the actual deploy. |
| Common drift patterns | Same as LCP trend (since LCP dominates the score). Step regressions on deploys; gradual drift on accumulated changes. The score is more sensitive than individual sub-metrics because all five contribute simultaneously; a small regression in TBT plus a small regression in LCP can drop the score visibly even when each individual sub-metric stays in band. |
| CI/CD usage | Lighthouse CI uses this score as a build budget gate. Common pattern: fail builds when score regresses >5 points or drops below 60 absolute. Catches performance regressions before they reach production. |
| Currency | n/a, 0-100 score time-series. |
| Time window | T/7D/30D rolling |
| Alert trigger | 7-day rolling score drops > 10 points vs 30-day baseline OR current score < 50 (red band). |
| Sentiment key | psi_perf_score |
| Roles | owner, marketing, 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, mobile lab score trend over 6 months ending Wednesday 15 May 26.| Month | Score (7D rolling) | Δ vs prior | Field CWV pass rate (compare) | Notes |
|---|---|---|---|---|
| Nov 25 | 72 | (baseline) | 84% | Healthy lab + field |
| Dec 25 | 70 | -2 | 82% | Stable |
| Jan 26 | 64 | -6 | 78% | Lab moved first, field followed at 28-day lag |
| Feb 26 | 58 | -6 | 72% | Hero carousel deploy hit lab immediately, field caught up |
| Mar 26 | 50 | -8 | 68% | Marketing-stack expansion + cache config change |
| Apr 26 | 44 | -6 | 65% | BFCM imagery + cache stayed bad |
| May 26 | 42 | -2 | 64.8% | Current state, lab and field aligned |
| Cumulative drift | -30 points | -19 percentage points | 6-month silent score erosion |
- Lab trend leads field trend by 4-6 weeks. Each deploy / change shows up in lab within days; field absorbs over the 28-day rolling window. Reading lab and field trends together provides a leading-indicator + truth-source view.
-
The Jan 26 6-point drop in lab was the early warning. Lab caught the regression ~28 days before field-CWV-pass-rate visibly dropped. A
+5 point lab score regressionalert would have fired then; investigating immediately would have prevented the cascade. - The Mar 26 8-point drop is the largest single-month regression. Two contributing changes (marketing stack + cache config) deployed in the same sprint window. CI/CD gating at “score must not drop > 5 points vs baseline” would have blocked at least one of these deploys, preventing the compound regression.
- Current 42 score is in red band (below 50). The optimisation cycle to recover to 70+ requires the work documented across the per-CWV cards: image optimisation, third-party deferral, render-blocking fix, cache restoration. Lab score is the primary build-gate metric; CI configuration update recommended to prevent recurrence.
- Lab vs field divergence is informative. When lab and field move together (this site’s pattern), real-user conditions roughly match emulation. When lab improves but field doesn’t, real users have slower devices/networks than emulation assumes; investigate audience profile.
- Compare lab trend with field trends. Lab leads; field confirms.
- Identify deploys / changes correlated with lab inflection points. Lab captures changes within 24-48 hours; cross-reference deploy log.
- Apply CI/CD gating to prevent regressions reaching production.
| Time horizon | Action |
|---|---|
| First 1 hour | Identify deploys correlated with lab score drops. |
| First 24 hours | Roll back or fix-forward problematic deploys. |
| Day 7 | Lab trend confirms recovery. |
| Day 28 | Field metrics catch up. |
Sibling cards merchants should reference together
| Card | Why merchants reach for it |
|---|---|
psi_perf_score_summary | Static lab score snapshot. |
psi_performance_score | Per-URL lab score gauge. |
crux_pass_rate_trend | Field CWV pass rate trend; the field counterpart. |
crux_lcp_trend | LCP-specific field trend. |
crux_inp_trend | INP-specific field trend. |
crux_cls_trend | CLS-specific field trend. |
psi_biggest_regression | Per-URL regression detection. |
crux_regression_timeline | Composite regression detection. |
Reconciling against the vendor’s own dashboard
Where to look:- Lighthouse CI dashboard, historical lab scores from CI builds.
- PageSpeed Insights, current snapshot.
- Internal CI/CD logs, deploy-correlated score history.
| Reason | Direction | What to do |
|---|---|---|
| Audit cadence. Vortex IQ runs scheduled audits (daily); CI runs per-commit. | Different granularity | Use CI for per-deploy attribution; Vortex IQ for trend monitoring. |
| URL set. Vortex IQ audits configured URL list; CI typically audits build artifacts. | Different scope | Confirm same URLs for direct comparison. |
| Smoothing. Vortex IQ uses 7-day rolling. | Smoothed | Single-day deploys show within 3-7 days of actual deploy. |