5.3 MB
average product photo
14,800 photos of 2,500 lamps were served as camera JPEGs of 4–8 MB, with no responsive sizes. The main photo on a product page weighed 6.8 MB on its own.
Finding 1 of 4
A fictional business. The figures are a modelled scenario, not a client result.
Fenerche Lamps · Online lighting shop, 2,500 lamps · Plovdiv
Mobile LCP from 7.0 s to 1.8 s, then +41% organic sessions
7 months · Bulgaria · Signal plan · Image compression · Lazy loading · Script deferral · Caching & CDN
Mobile LCP · p75
7.0 s
Before
1.8 s
Month 7
01The challenge
Camera-sized product photos of 4–8 MB, render-blocking scripts and no caching left mobile LCP at 7 s, and rankings and sales slipped after a core update.
Fenerche photographs every lamp in its Plovdiv showroom, lit and unlit, from six angles, and the photos went online straight from the camera. On a phone over 4G a product page took seven seconds to show its main photo, and many shoppers left before the lamp ever appeared. When a broad core update reshuffled lighting searches in Bulgaria, the slow pages slid down and conversions slid with them.
One product page, before
18.4 MB on the first load
Eight weeks after the core update
02Diagnosis
5.3 MB
average product photo
14,800 photos of 2,500 lamps were served as camera JPEGs of 4–8 MB, with no responsive sizes. The main photo on a product page weighed 6.8 MB on its own.
Finding 1 of 4
11
render-blocking scripts in the head
Of 23 third-party scripts (chat, reviews, three analytics tags, a heat-map tool, two retargeting pixels), 11 loaded synchronously before anything was drawn: 1,850 ms of total blocking time.
Finding 2 of 4
1.6 s
connection and server response on mobile
Every page was rebuilt on every request and static files were sent with Cache-Control: no-store, so a returning shopper downloaded all 18.4 MB again.
Finding 3 of 4
-36%
mobile conversion after the update
In the eight weeks after the update, 312 tracked keywords fell from an average position of 8.9 to 15.3, and mobile conversion dropped from 1.4% to 0.9%.
Finding 4 of 4
03The strategy
Built on our Signal plan: technical SEO measured in field data and shipped one change at a time, so every notch can be credited to its fix.
Notch 1
All 14,800 photos re-encoded to AVIF with WebP and JPEG fallbacks, at six responsive widths, with width and height set so nothing jumps. The average photo went from 5.3 MB to 140 KB.
Notch 2
Photos below the fold, the 40-tile category grids and the room-setting galleries now load only as the shopper scrolls near them. The main photo is preloaded with fetchpriority="high", so it no longer waits in line.
Notch 3
Seven scripts removed, twelve deferred until after the first paint or the first tap, critical CSS inlined. Chat and reviews load when a shopper reaches for them.
Notch 4
Pages cached at the edge with stale-while-revalidate, static files fingerprinted and cached for a year, HTTP/3 switched on. Connection and server response on mobile fell from 1.6 s to 0.9 s.
04Turn the dimmer
Turn the knob to add the fixes in the order we shipped them, and watch the product page paint sooner on a mid-range phone.
Drag the knob, use the arrow keys on it, or pick a notch.
Notch 0 of 4
Camera-sized photos, eleven blocking scripts, no caching. The filament barely glows.
Needs work: above 2.5 s
What the phone waits for
The product page, second by second
1 s
2 s
3 s
4 s
5 s
7 s
As found: LCP 7.0 s, first-load weight 18.4 MB.
05The engagement
Month 1Field LCP 7.0 s
CrUX and real-user monitoring set up, an image inventory and a script census, and the drop mapped keyword by keyword.
Month 2Field LCP 6.1 s
All 14,800 photos re-encoded; new uploads are compressed automatically in the shop's admin.
Month 3Field LCP 4.5 s
Category grids, galleries and related products load on approach; the main photo is preloaded.
Month 4Field LCP 3.6 s
Third-party scripts cut from 23 to 16, twelve of them deferred; critical CSS inlined.
Month 5Field LCP 2.6 s
Edge cache, long-lived static files and HTTP/3; a repeat visit now downloads about 0.1 MB.
Month 6Field LCP 1.9 s
The 28-day CrUX window passes all three Core Web Vitals on mobile; a performance budget now guards every release.
Month 7Field LCP 1.8 s
Category searches back on page one and mobile conversion above its pre-update level.
06Results
Mobile LCP (p75, field data)
7.0 s → 1.8 s
Organic sessions (month 7 vs. month 1)
+41%
Mobile conversion rate
0.9% → 1.6%
Product page weight (first load)
18.4 MB → 1.2 MB
Modelled scenario for a fictional business — not a client result. Rankings and traffic are never guaranteed; see our Terms.
| Query (as typed) | Before update | After update | Month 7 |
|---|---|---|---|
| магазин за осветление пловдив | #1 | #2 | #1 |
| пендел лампа | #3 | #9 | #2 |
| лампи за таван | #4 | #11 | #3 |
| нощна лампа | #5 | #12 | #4 |
| аплици за стена | #6 | #14 | #5 |
| модерен полилей | #7 | #15 | #5 |
| led осветление за кухня | #8 | #17 | #6 |
| градинско осветление | #9 | #19 | #8 |
Of 312 tracked keywords, 58 were in the top three before the update, 21 just after it and 71 by month 7.
We find where the seconds go and win them back one notch at a time, on our Signal plan.