Chad Case File 003c · Arros QD · London · Jul – Sep 2026 vs 2025

Three months on, traffic is 2.7× last year.

The third report on Arros QD, after the 48-hour rebuild and the first month. Each of the first three full months on the Chad-built site, compared with the same month a year earlier on the old one. July–September sessions rose from 10,422 to 27,911, up 167.8%.

+167.8%
Sessions, Jul–Sep vs 2025
10,422 → 27,911
+154.2%
New users, Sep vs Sep 2025
2,343 → 5,956
+804.7%
Organic Search, Sep vs Sep 2025
321 → 2,904
+128.5%
Sessions excl. paid, Aug vs Aug 2025
2,838 → 6,485
What we did

Compared each of the first three full months on the rebuilt site with the same month of 2025, in GA4, filtered to arrosqd.com.

What changed

Sessions rose 205.6%, 154.1% and 143.3% in July, August and September. With paid channels removed, August and September still rose 128.5% and 103.5%.

Limits

GA4 under-recorded sessions from 24 August to 13 September, so August and September are understated. Booking tracking started on 19 August and has no 2025 comparison.

00 — The launch

Traffic tripled in the nine days after launch, before any paid media.

The rebuilt site went live on 26 June 2026. Matched nine-day windows: 17–25 June before launch, 27 June–5 July after it. Paid Search started on 6 July and Paid Social on 15 July, so the second window is pre-paid.

MetricPre-launchPost-launch, pre-paidChange
Total sessions9052,751+204.0%
New users5481,852+238.0%
Organic Search sessions881,324+1,404.5%
▸Launch day: GA4 flagged a new-user anomaly on 26 June itself: 54 expected, 186 recorded. First full month: July sessions rose from 4,352 in June to 10,771 (+147.5%).

The launch evidence in full is in the first-month report. The sections below test whether it held.

01 — Year over year

Every month at least 2.4× the year before.

Comparing a month with the same month a year earlier holds the season constant. July–September 2025 ran on the previous site; July–September 2026 on the rebuild, launched 26 June 2026.

Month20252026Change
July3,52410,771+205.6%
August3,3068,399+154.1%
September3,5928,741+143.3%
July–September10,42227,911+167.8%
▸New users grew faster than sessions: 2,362 → 5,895 in August (+149.6%) and 2,343 → 5,956 in September (+154.2%). The growth is new visitors, not repeat visits.
▸Current run-rate: from 14 to 30 September the site averaged 473 sessions a day, against 271 a day across August (+74.8%).
02 — Excluding paid media

Without paid channels, sessions still more than doubled.

Paid Search started on 6 July 2026 and Paid Social on 15 July. The rows below remove Paid Search, Paid Social, Display and Cross-network, calculated from each report's channel table.

Month2025, excl. paid2026, excl. paidChange
August2,8386,485+128.5%
September3,1476,403+103.5%
!Channel tables do not sum exactly to the reports' headline totals, so these rows are calculated from channel totals alone. The July report has no channel comparison with July 2025.
03 — Channel mix

Organic Search went from 9% of traffic to the largest channel.

Sessions by channel, September 2025 vs September 2026. In September 2025, 78% of sessions were Direct: people who already knew the restaurant.

ChannelSep 2025Sep 2026Change
Organic Search3212,904+804.7%
Direct2,7412,664−2.8%
Paid Search3561,322+271.3%
Paid Social0877new
Referral17658+3,770.6%
Organic Social087new
AI Assistant050new
▸August shows the same shift: Organic Search 302 → 3,147 (+942.1%), now the largest single channel. Paid Social, Organic Social, AI Assistant and Display all generated zero sessions in August 2025.
04 — Commercial relevance

90.1% of September sessions started on the homepage, menus or lunch page.

Top landing pages, 1–30 September 2026. August's share on the homepage and menus was 83.7%; July's was 79.3%.

Landing pageSessionsEngagementAvg. time
/ (homepage)5,58563.8%158.4s
/menus1,22146.6%172.2s
/menus/main-menu53885.5%117.5s
/lunch52970.7%177.8s
/spaces6564.6%299.3s
/experiences6170.5%127.4s
▸Menu downloads: 633 in August → 1,410 in September (+122.7%). /lunch: 28 sessions in August → 529 in September, 82% of them from Paid Search.
05 — Measurement

Three things the raw GA4 numbers get wrong.

Found by Chad while producing the monthly reports. Each one inflates or hides a number that would otherwise be reported at face value.

booking_complete counts bookings that never touched the websiteSevenRooms
3,315 of 4,192 Paid Social sessions in September went straight to SevenRoomsPaid Social
The property also records sevenrooms.com: 4,854 sessions in SeptemberRaw totals
A booking count is not a website conversion count. SevenRooms' own booking page carries Arros QD's analytics tags, so a booking made from Google Maps' Reserve button or a bookmark fires the same event as one made through the site. Every session figure on this page is filtered to arrosqd.com. Bookings are not reported as website results until they are reconciled with SevenRooms' own reservation report.
06 — Conclusion

The rebuild changed how the restaurant is found.

"In the three months after Chad's rebuild went live, arrosqd.com recorded 27,911 sessions against 10,422 in July–September 2025, a 167.8% increase." Organic Search grew from 321 to 2,904 sessions a month to become the largest channel. With every paid channel removed, August and September sessions still rose 128.5% and 103.5% year over year. Traffic concentrated on the homepage, menus and lunch page, and new users grew faster than sessions.

Next phase

!Data notes: GA4 recorded unusually few sessions on every channel from 24 August to 13 September; those days are included as recorded. July includes 480 sessions from one suspicious referral on 24 July; excluding them, July is +192.0% on July 2025. The September report gives September 2025 as 3,592 sessions in its overview and 3,509 in its channel table; 3,592 is used here.

See what a rebuild does to your own traffic.