Breaking down what “bookings are dropping” actually means — by booking vs zero-booking days, booking-count, and APP/WEB.
Period: 2026-08-06 to 09-20 / Source: [TOMOGO!] Analysis Report(51) _daily_data + Amplitude MCP / Bookings = Active Bookings (excl. cancellations, by booking date)
Conclusion: bookings themselves are not falling. The last 7 days = 8, within the Jul–Aug weekly baseline (7–12); statistically only −1.3σ, not an anomaly. What shrank is the booking “top of funnel” (user-acquisition volume), driven mainly by the deliberate ANW shutdown + ad-spend reduction. Booking quality/depth has actually improved — notably the 8/19 full frontend redesign lifted WEB CVR 4.5× (see 5 below).
1. Bookings are not falling
Bookings are flat; what collapsed is ad clicks
Weekly bookings hold in a 7–12 range (the 17 in the week of 8/31 is a spike). Ad clicks, meanwhile, collapsed from 24k to 4k. The felt “drop” is really the shrinking of clicks (= inflow).
Note: the click collapse aligns with the 9/7 ANW shutdown + ad-spend cut — removal of low-quality inventory, i.e. by design.
2. What dropped vs what's normal
Only the entrance (clicks) fell. Tour views and CVR are normal-to-improved
Indexed to “average of days a booking occurred = 100%”. Only clicks (33%) stand out as low. Tour detail views 99% = normal; CVR 115% = improved. → The initial hypothesis “tours aren’t being viewed” is rejected.
3. Quality by APP/WEB & platform
Tour views per user are actually deepening on iOS and Web
Tour-detail views per user: iOS +31%, Web +23% above the full-period average. Each person browses more — the issue is “headcount”, not “how they browse”. Only Android declined (89%), the sole regression.
Source: Amplitude MCP (platform = user scope, View Tour Detail Page).
4. Conversion rate (CVR)
APP CVR at 0.83% — an 8-week high. WEB switched on: 0 → 0.3
Even with sessions cut, APP conversion is at an 8-week high (0.83%). WEB also lifted from 0 to 0.3 since 8/31 (likely the web booking flow taking hold — hypothesis). = the cut removed low-quality sessions; remaining users are higher quality.
The 8/19 full redesign is working (= the main improvement)
Around the 8/19 frontend overhaul (Web/Guest App/Admin), WEB shows CVR ×4.5 and +34% views per user. The view lift exceeds iOS (+20%) — it lands most directly on Web. WEB inflow fell that same day (ad cut) yet conversion rose = quality, not volume. Bookings turned on ~9/1, ~2 weeks after the redesign = matching the WEB lead time.
WEB session CVR
0.05%→0.23%
4.5x
Tour views per user
2.15→2.90
+34%
WEB bookings (monthly)
8.3→12.7
+53%
Caveat: WEB bookings are only 14 in absolute terms — a “strong promising signal” stage. 8/19 also updated Guest App/Admin, so APP quality gains (APP CVR high 0.83%) may also stem from it. Source: Excel CVR WEB% + Amplitude (Web VTD/user).
Meta ads: campaigns generating bookings (facts)
Source: DATABEAT ⑰ by campaign / month-to-date (2026-09-01–09-23, through prior day). Meta records bookings as the AppPurchase event. Entry = WEB landing-page views ([WEB&APP] campaigns, Meta dashboard).
Meta campaign
CV (bookings)
App / Web
CPO
[WEB&APP] RT_Purchase (retargeting)
6
4 / 2
¥38,969
[WEB&APP]CV_Create Account
5
4 / 1
¥45,251
[APP_Booking]IOS_Advantage+_CP
3
2 / 1
¥231,622
Meta total
14
10 / 4
—
Fact: top booking contributor is retargeting (RT_Purchase) at the lowest CPO ¥38,969. The acquisition campaign CV_Create Account also drove 5 bookings. Booking events are App-heavy (App 10 / Web 4).
Watch-items & next moves
1
Only Android’s view depth declined (89%) — opposite of iOS/Web. Check for an Android-specific UX/delivery issue.
2
Sign-ups down 65% vs July — with lead time, this risks thinning future booking volume.
→
Top test theme: strengthen WEB inflow (not APP). Now that the redesign made WEB a “converting box”, it’s worth revisiting the Web→APP focus and routing inflow back to WEB ads/paths. Test small first, and check whether WEB’s CVR × acquisition cost beats APP.
→
Other topics: (1) cause of the Android depth decline (89%); (2) recovery of sign-up volume (−65% vs July).