Harbor & Pine Performance Report

Recommendations

Prepared by your analyst

Maintenance — handled by your plan

Convert room-scene and product images to AVIF 1.4s estimated LCP savings across 5 templates — the single biggest lever Scheduled · next maintenance window
Defer gallery slider + non-critical checkout scripts Slider is dead weight on 2 templates; deferring analytics on checkout saved 80ms INP in testing Scheduled · next maintenance window
Preload the serif display font Late discovery adds ~0.3s to first render on every template Scheduled · next maintenance window

Opportunities — worth pursuing

Lazy-load below-fold room scenes on collection pages 14 full-size images load up front on /shop; 4 are in the first viewport ≈1 h · fits your included hours
Serve one optimized pair for the hover crossfade Both card images ship at full size — ~600KB saved per collection view ≈2 h · fits your included hours
Rework the checkout payment-script loading order Checkout INP is 285ms on Android — the worst number on the site, on the page that takes the money ≈3 h · fits your included hours

Analyst notes

Cross-referenced: RUM × CrUX × Lighthouse

The July gallery update regressed INP within hours — and RUM caught it the same day. CrUX wouldn’t have shown it for weeks (28-day rolling average). Rolled back July 18; the field trend confirms full recovery by early August.

Mobile LCP runs 0.9s behind desktop, and it’s image sizing, not the server. TTFB is nearly identical on both — the gap closes when collection cards stop shipping desktop-size images to phones (the ≈2 h crossfade fix).

Checkout is the slowest template for interaction, on the device most shoppers use. 285ms INP on Android traces to payment iframes plus analytics competing on the main thread. Deferring what isn’t needed before first tap is the highest-value fix on the list.

Core Web Vitals — real users

Source: CrUX (Chrome UX Report), mobile · trailing 28 days

What Google measures your shoppers experiencing, and judges rankings on.

LCP · loading

2.4s

✓ Good — under 2.5s

INP · interactivity

210ms

! Needs work — target ≤200ms

CLS · stability

0.04

✓ Good — under 0.1

Lab performance

88 / 100

✓ Good

Six-month trend

Source: CrUX History API · weekly, mobile

75th-percentile field values by week. Dashed line marks Google’s “good” threshold.

LCP · seconds

INP · milliseconds

The July INP spike was the gallery plugin update — caught and rolled back within the plan — see the regression log below.

Real-user monitoring by device

Source: TALLKAROL RUM (web-vitals → CRM) · 12,400 page views sampled

Your actual visitors, all browsers, live — not limited to Chrome’s 28-day sample.

Device	Share of views	LCP p75	INP p75	CLS p75	Status
Mobile	64%	2.8s	230ms	0.05	! LCP + INP over target
Desktop	31%	1.9s	110ms	0.02	✓ Good
Tablet	5%	2.3s	160ms	0.03	✓ Good

Worst segment: Chrome on Android · collection pages. Regressions in RUM alert the same day and open a ticket in the support portal automatically.

Vitals by template

Source: TALLKAROL RUM · p75, Aug 1–31

Template	LCP mobile	INP mobile	CLS mobile	LCP desktop	INP desktop	Status
/ (home)	2.3s	140ms	0.03	1.6s	70ms	✓ Good
/shop (collection)	2.9s	190ms	0.06	2.0s	95ms	! Mobile LCP
/product (template)	2.7s	220ms	0.04	1.9s	120ms	! Mobile
/cart	2.2s	180ms	0.02	1.5s	90ms	✓ Good
/checkout	2.4s	285ms	0.03	1.8s	150ms	✕ INP — fix queued
/journal (blog)	2.1s	120ms	0.03	1.4s	65ms	✓ Good

Lab scores by template

Source: PageSpeed Insights API (Lighthouse, mobile) · avg of 3 runs, Aug 29

Template	Performance	Accessibility	Best practices	SEO	Status
/ (home)	92	100	96	100	✓ Good
/shop (collection)	84	100	96	98	✓ Good
/product (template)	78	98	92	98	! Image weight
/cart	86	100	96	96	✓ Good
/checkout	74	98	88	96	! Payment scripts
/journal (blog)	90	100	96	100	✓ Good

Top fixes queued

Source: PageSpeed Insights API — Lighthouse opportunities

Fix	Est. savings	Templates affected	Status
Serve images in AVIF/WebP	1.4s	5	Scheduled
Reduce unused JavaScript (gallery slider)	0.6s	2	Scheduled
Preload key font	0.3s	6	Scheduled
Properly size images (collection cards)	0.9s	1	! Needs the ≈2 h opportunity

Update-window regression log

Source: PageSpeed Insights API · lab runs before/after each maintenance window

Every update window is bracketed by lab runs, so a regression is caught the day it ships — not in next month’s numbers.

Window	What shipped	Before	After	Outcome
Jun 18	Core + 6 plugin updates	84	86	✓ Clean
Jul 16	Gallery plugin 4.2 + 4 others	86	79	✕ Regression — gallery 4.2
Jul 18	Gallery rolled back to 4.1	79	85	✓ Recovered
Aug 20	Core + 5 plugin updates, image lazy-load fix	85	88	✓ Improved