Eight seconds off the page a paid ad lands on
God Loves Nihon · godlovesnihon.jp
An evangelistic site whose visitors almost all arrive from paid ads, on a phone. Nothing was redesigned: the same pages now appear in 2.6 seconds instead of over ten, and report where readers stop.
The problem
This site is the landing page for advertising campaigns: its visitors are overwhelmingly people who tapped an ad on a Japanese mobile connection, and the ministry has already paid for that tap by the time the page starts loading. On Wix, a throttled phone waited more than ten seconds for the largest element to appear and pulled down about three megabytes to get there — long enough to lose a visitor whose click was already bought.
What we did
Nothing was redesigned. We rebuilt the same pages as static files the ministry owns outright, cutting what a visitor has to download before the page appears, and moved hosting to a CDN with strong coverage in Japan and no charge for bandwidth, so a campaign that works costs nothing more to serve. We also rebuilt the measurement: these pages are long scrolling stories with almost no clicks, so a pageview alone could not tell a reader who reached the prayer from someone who left at the first screen. The pages now report how far people read, which sections they reach and which calls to action they follow, into a single fresh property rather than the two overlapping ones the old site ran. Legacy URLs redirect to their new homes, so nothing already advertised, linked or indexed breaks.
Result
On a throttled phone the largest element now appears in 2.6 seconds instead of 10.6, and the page weighs about 400 KB instead of about 3 MB. What that is worth in campaign terms belongs to the ministry’s own ad reporting; what we can measure is the wait, and the wait is eight seconds shorter. The rebuild is finished and measured but not yet live — godlovesnihon.jp still serves the Wix version until the ministry decides to switch, which is their call and reversible. The new tracking is deliberately silent until then: it only reports from the live domain, so preview traffic cannot pollute the figures the campaigns will be judged on.
What changed for the people using it
| Aspect | Before | After | Change |
|---|---|---|---|
| Who can move the site | Only through the Wix account | The ministry owns the code and can host it anywhere | Ministry controls it |
| Control of the markup | Generated by Wix, not editable | Hand-authored and readable | Fully editable |
| Old page addresses | Would break on any move | Redirected to their new homes | Preserved |
| Ongoing cost | Wix subscription | Static hosting on a free tier | No longer billed |
| What a campaign can be judged on | A pageview, then silence | How far people read, the sections they reach, the buttons they follow | Drop-off is visible |
| Analytics setup | Two overlapping GA4 properties plus a tag manager | One fresh property | Reports agree |
What changed, measured
| Measure | Before | After | Change |
|---|---|---|---|
| Performance score | 73 | 97 | +24 pts |
| Accessibility score | 92 | 95 | +3 pts |
| First Contentful Paint | 2.0 s | 0.9 s | −55% |
| Largest Contentful Paint | 10.6 s | 2.6 s | −75% |
| Total Blocking Time | 117 ms | 0 ms | −100% |
| Page weight | 3,097 KiB | 416 KiB | −87% |
- Date measured
- 5 August 2026
- Methodology
- Mobile Lighthouse with simulated throttling, home page. Performance figures are the median of 3 runs; accessibility and SEO are single runs, because those audits are deterministic.
- Caveats
- “Before” is the live Wix site; “after” is the finished rebuild on its preview host. No SEO score is shown: the rebuild is deliberately hidden from search engines until launch so it cannot compete with the live site, which Lighthouse scores as a failure.