Conversion check: Oct 6-12 vs Sep 29 - Oct 5
Source: GA4 property 400681367 (tickadoo.com) and Howard order API. Pulled 13 Oct 2026, ~12:40 CEST. Oct 13 is a partial day and is not in the comparison.
Most GA4 sessions are a bot
About 20k sessions a day, roughly 80% of everything GA4 reports, come from headless browsers at three impossible screen sizes (1280x1200, 1366x1366, 800x600). Across Sep 15 - Oct 13 they made zero purchases. On Oct 11 the same bot spiked to 187k sessions (US, Brazil, France, Mexico, UK). Conversion rates on this page use real sessions, with those excluded; raw GA4 CVR (about 0.2%) mostly measures the bot.
More orders, smaller baskets
Purchases are up 12% and Howard orders up 11%, but GA4 revenue is down 15% (95k vs 112k), so the average order is about 24% smaller (209 vs 275). Worth checking whether the mix moved toward cheaper experiences or fewer big seat map orders.
Purchases per day (GA4)
Grey = Sep 29 - Oct 5, blue = Oct 6-12, red = today (partial).
Conversion rate per day (purchases / real sessions)
Oct 11 looks high because real sessions dipped on a Sunday, not because of the bot spike, which is excluded.
Howard orders created per day
Successful POST /api/orders/bookings. Independent of GA4, so unaffected by bots or attribution.
Bot sessions per day
Sessions at 1280x1200, 1366x1366 or 800x600. Today Singapore (1366x1366, about 2k so far) is back, plus 3.3k from China at 1920x1080, not excluded here.
What changed recently
- Fixed (6 Oct): 1 adult is preselected again on Headout experiences whose supplier minimum is 0 (frontend#3168).
- Measurement (13 Oct): seat map funnel events and device class added (howard#7268, frontend#3222); mobile seat map Continue now tracked (frontend#3224). Analytics only, no checkout behaviour change.
- Measurement (5 Oct): frontend#2747 changed GA4 session attribution.