Services

A 6.8-Second LCP and Half Your Ad Clicks Never See the Product: How Website Speed Optimization Works, from a Core Web Vitals Audit to Images, Caching and CDN

2026.09.27 · 5 views
A 6.8-Second LCP and Half Your Ad Clicks Never See the Product: How Website Speed Optimization Works, from a Core Web Vitals Audit to Images, Caching and CDN
“

Four options compared, a 4-week optimization process, Cloudflare and labor cost breakdown, six traps and fixes, and a 90-day roadmap verified with real-user data. ”

Share:

A handmade furniture brand spends NT$120,000 a month on Meta ads, yet GA4 shows a 71% bounce rate on its mobile landing page. Measurement revealed a 4.2 MB uncompressed hero image, a 6.8-second LCP on mobile 4G, nine tracking tags and a chat widget. Many people clicked the ad and left before seeing a single product. Speed optimization is not about a pretty score; it is about making sure paid traffic actually sees your content.

When It Fits × When It Doesn't

Good fit for speed optimizationHold off
You run ads and mobile bounce rate exceeds 60%Under 1,000 monthly visits: fix content and visibility first
Search Console marks most URLs "Needs improvement" or "Poor"A redesign is due within six months: build it in there
Home or product page LCP exceeds 4 secondsYou are on a closed platform like Wix or Shopify with little room to tune
Lots of plugins or tags, and buttons feel laggyThe server keeps crashing: fix stability first

Alternatives Matrix

OptionCostProsTrade-offs
Install a caching plugin yourselfNT$0–3,000/yrDone in a dayOften conflicts with other plugins; treats symptoms
Add a CDN onlyCloudflare Free and upFaster static filesA 4 MB image is still 4 MB through a CDN
Upgrade hostingNT$500–3,000 more per monthBetter TTFBWasted if the host is not the bottleneck
Outsourced audit + optimizationNT$30,000–120,000Data-driven fixes with acceptance criteriaTakes 3–4 weeks and internal cooperation

Rule of thumb: look at the four LCP subparts (TTFB, resource load delay, load duration, render delay) and fix whichever one is slow, instead of paying for a new host first.

4-Week Optimization Process

PhaseTimingDeliverablesTools
BaselineWeek 1Baseline report for the top 20 pages, ranked bottlenecksPageSpeed Insights, Search Console, WebPageTest
Quick fixesWeek 2WebP/AVIF images, fetchpriority on the LCP image, font subsettingSquoosh, Chrome DevTools
Caching and CDNWeek 3Caching strategy document, cache rulesCloudflare, Redis
Third parties and interactivityWeek 4Script budget, long-task splitting, acceptance reportGoogle Tag Manager, Lighthouse

Acceptance follows the Core Web Vitals thresholds on web.dev: LCP ≤ 2.5 s, INP ≤ 200 ms, CLS ≤ 0.1, measured at the 75th percentile of real users.

Real Cost Breakdown

  • Audit report: NT$8,000–15,000, about 8–12 hours of measurement and analysis
  • Optimization work: brochure or WordPress sites NT$30,000–60,000; e-commerce or custom systems NT$60,000–120,000
  • CDN: per Cloudflare's official plans, Free is usable, Pro is US$20/month and Business US$200/month (billed annually)
  • Hidden costs: 4–10 hours to re-export and re-upload images, time coordinating with marketing to remove tags, 2–4 hours of regression testing each quarter after theme or plugin updates
  • Maintenance: monthly performance monitoring and reporting NT$2,000–5,000

Implementation Reality vs Client Expectations

Clients assumeWhat actually happens
A score of 100 means ranking firstGoogle's documentation treats it as one part of page experience; content still dominates
Results show up tomorrowField data is a 28-day rolling average, so it takes about four weeks to fully reflect
The tech team can fix it aloneThe heaviest scripts are often marketing's tags, which need a cross-team decision
Do it once and it lasts foreverOne new plugin or oversized image and you are back where you started within three months

Six Traps and How to Avoid Them

  • Only watching lab scores: accept against real-user data in the Search Console Core Web Vitals report
  • Lazy-loading the hero: never lazy-load the above-the-fold LCP image; add fetchpriority="high" instead
  • Images without dimensions: reserve width and height for every image and ad slot to prevent CLS spikes
  • Caching logged-in or cart pages: exclude personalized pages with cookies from cache rules and test checkout
  • Tags only ever get added: set a script budget; every new tag needs an owner and a removal date
  • Testing on desktop only: always measure on a mid-range Android phone over 4G

Success Metrics and 90-Day Roadmap

  • 30 days: lab LCP under 2.5 s, homepage transfer size cut by 50%+, zero checkout errors
  • 60 days: more than half of URLs rated "Good" in Search Console, mobile landing page bounce rate down 10+ percentage points
  • 90 days: all three metrics pass at the 75th percentile; compare ad conversion rate and cost per conversion; set up monthly performance regression alerts

Decision Checklist

  • ☐ Do you run paid ads every month?
  • ☐ Is mobile bounce rate above 60%?
  • ☐ Is homepage LCP above 4 seconds?
  • ☐ Does Search Console show "Poor" URLs?
  • ☐ Does the site have more than 20 plugins?
  • ☐ Are there more than 5 tracking tags?
  • ☐ Is any homepage image larger than 500 KB?
  • ☐ Do taps on mobile feel delayed?
  • ☐ Is marketing willing to review its tags with you?
  • ☐ Is there a budget of NT$30,000 or more?

Seven or more checks and it is worth doing; fewer than four, start with free image compression and plugin cleanup.

FAQ

Do we need a PageSpeed score above 90?

No. Google evaluates whether real users' 75th-percentile LCP, INP and CLS meet the thresholds; the lab score is only a diagnostic tool.

How long until we see results?

Lab data changes the same day; PageSpeed Insights field data is a 28-day rolling average, so it takes about four weeks to fully reflect.

Will switching hosts fix it?

Only when TTFB is the problem. For most SMB sites the bottleneck is images, third-party scripts and fonts, so a new host often costs more without helping.

Can this be done on WordPress?

Yes. The usual approach is removing unnecessary plugins, converting images to WebP/AVIF, and adding page caching and a CDN, usually without a rebuild.

Next Step

ScriptWalker's Website Performance Audit & Optimization service starts at NT$30,000 (standalone audit report from NT$8,000), covering baseline measurement, image and caching work, Cloudflare setup, a script budget and 90 days of post-launch data tracking. Send us your URL and a free 30-minute consultation will pinpoint the three things slowing you down most.

Share: