| Page type | Performance, before | Performance, after | Speed Index, before | Speed Index, after |
|---|---|---|---|---|
| Home | 30.0 | 52.0 | 17.2s | 9.5s |
| Product | 30.5 | 60.6 | 19.8s | 9.1s |
| Collection | 31.3 | 57.5 | 19.2s | 8.8s |
| Blog | 33.1 | 59.9 | 19.4s | 8.2s |
Audited June 12, 2026, Lighthouse 13.3.0, Moto G Power emulation with 4x CPU slowdown and slow 4G, matching Google's standard mobile testing conditions.
Excludes Prime Day (Jun 22 - Jul 5, 2026). Traffic declined by a similar amount in both years, yet conversion rate moved in opposite directions, ruling out a traffic-volume explanation.
Judges each channel against its own historical pattern, rather than requiring a flip from negative to positive to count. A channel that declined less this year than last year is a real, relative improvement.
| Channel | Rate change, 2026 | Rate change, 2025 | Verdict |
|---|---|---|---|
| Paid Social, mobile | +118.9% | -8.3% | Improved |
| Direct, desktop | +125.4% | -30.3% | Improved |
| Direct, mobile | +37.3% | -44.2% | Improved |
| Paid Search, desktop | +37.4% | -2.1% | Improved |
| Organic Search, mobile | -10.5% | +1.4% | Declined |
| Organic Search, desktop | -2.4% | -5.6% | Improved |
| Unassigned, mobile | +65.8% | -31.1% | Improved |
| Paid Search, mobile | +21.0% | +42.5% | Declined |
Organic Search, desktop still shows a decline in raw terms, but a smaller decline than last year, a real relative improvement, just not a full reversal. Organic Search, mobile and Paid Search, mobile are the two exceptions: both performed better last year than this year on these dates.
A Prime Day sale ran Jun 22 - Jul 5, 2026, inside this test's data. Even with those dates excluded, the graph below still shows a step up after the June 19 launch (+22%, p = 0.054).
Designed specifically to rule out "this was already trending in that direction," and to remove the sale's effect rather than let it inflate the launch's apparent size.
Combined across all 7 tested channels, weekly-averaged conversion rate, Jan - Aug 2026. The shaded band marks the Jun 22 - Jul 5 Prime Day sale, excluded from the trend and launch-effect calculation, though its real effect on the raw weekly data is still shown, not hidden.
Optimisation 01 delivered a measurable conversion improvement, once a Prime Day sale running inside the comparison window was correctly excluded: conversion rate up 19%, despite traffic declining by a similar amount to last year, ruling out a traffic-volume explanation.
It holds up under scrutiny: the same dates last year moved the opposite way, 6 of 8 channels performed better than they did last year, and the statistical test still shows a positive trend even with Prime Day removed entirely.