Report Date: 2026-04-27 | Data: through 2026-04-26 | For: Analysis Meeting
3 consecutive zero-booking days (Apr 24-26)
Weekly bookings dropped -77.4% (W1: 3.1/day to W2: 0.7/day). CPO surged to ¥196,274 (+368%). ROAS collapsed to 8.8%.
Two key Meta Ads changes:
explore.tomogo-travel.com/book-your-tour. Full-switch test to evaluate LP conversion performance.Rationale: Based on last-click attribution, concentrate reach on the segment with proven conversion. LP switch is an intentional test — the LP's original CVR was not high, so we're running a clean A/B by switching all traffic at once and monitoring the trend.
| Gender | W1 Installs | W2 Installs | Change | W1 Purchases | W2 Purchases |
|---|---|---|---|---|---|
| Female | 13 | 2 | -85% | 1 | 0 |
| Male | 8 | 5 | -38% | 3 | 1 |
Key insight: Female impressions actually increased (+12.6%) while installs crashed -85%. This points to creative fatigue or the destination change specifically disrupting women's path to install.
This aligns with the "Women Discover → Men Book" hypothesis — when women stop installing (top-of-funnel), men's downstream purchases also decline.
| Age | W1 Installs | W2 Installs | W1 Purchases | W2 Purchases | Cost Change |
|---|---|---|---|---|---|
| 35-44 | 9 | 3 | 3 | 0 | ±0% (¥185K both weeks) |
| 25-34 | 7 | 3 | 0 | 1 | +16% |
| 55-64 | 4 | 0 | 0 | 0 | +7% |
Male/35-44 accounted for 3 out of 4 total purchases in W1. This entire segment went silent in W2 despite identical spend.
Apr 25 Meta cost was -42% vs Apr 20. The decline started Apr 23 and bottomed on Apr 25. Stories placement took the biggest hit (-54%). Likely caused by budget pacing adjustments after the ad set consolidation reduced auction diversity.
| Country | W1 Purchases | W2 Purchases | W1 Cost | W2 Cost |
|---|---|---|---|---|
| US | 1 | 1 | ¥272,738 | ¥273,491 |
| JP | 2 | 0 | ¥17,765 | ¥21,397 |
| AU | 1 | 0 | ¥92,519 | ¥88,292 |
JP was the most cost-efficient segment in W1 (2 purchases on just ¥18K — tourists already in Japan with high intent). The LP destination change may have disrupted their short conversion path.
Combined effect of three factors:
These factors cascaded: fewer women discovering TOMOGO! → fewer men booking → zero-booking weekend.
To assess whether the female install collapse (Hypothesis A) impacts not just this week but future weeks, we analyzed the lag correlation between female installs and total bookings using historical data (Feb 1 - Apr 26, 2026; 85 days).
| Lag (days) | Correlation r | Strength | Interpretation |
|---|---|---|---|
| 0 (same day) | 0.09 | None | No same-day effect |
| 3 | 0.16 | Weak | |
| 5 | 0.30 | Moderate | Effect starts building |
| 7 | 0.47 | Moderate-Strong | Clear 1-week connection |
| 10 | 0.66 | Strong | |
| 12 | 0.70 | Strong (Peak) | Female Install peaks ~2 weeks before Booking |
| 14 | 0.69 | Strong |
Finding: Female installs correlate most strongly with total bookings 10-14 days later (peak r = 0.70 at lag +12 days). Same-day correlation is near zero (r = 0.09).
This quantitatively supports the "Women Discover → Offline Propagation → Men Book" two-layer hypothesis. The time gap represents the offline word-of-mouth cycle.
| Lag | Correlation r | Interpretation |
|---|---|---|
| Same week | 0.14 | None |
| +1 week | 0.40 | Moderate |
| +2 weeks | 0.64 | Strong |
| +3 weeks | 0.70 | Strong |
| Week | Female Installs | Bookings 2 Weeks Later | |
|---|---|---|---|
| Feb 2 | 345 | → | 23 (Feb 16) |
| Feb 9 | 242 | → | 27 (Feb 23) |
| Mar 2 | 32 | → | 23 (Mar 16) |
| Mar 16 | 12 | → | 11 (Mar 30) |
| Apr 6 | 18 | → | 5 (Apr 20) |
| Apr 13 | 13 | → | ? (Apr 27) Expected: low |
| Apr 20 | 2 | → | May 4: Risk of further decline |
| Lag | r | Interpretation |
|---|---|---|
| 0 days | -0.08 | The female-to-male connection is not captured within Meta's attribution window — indirect evidence that the path runs through offline channels (word-of-mouth → brand search → booking) |
| 7 days | -0.13 | |
| 14 days | -0.20 |
Caveats:
Does increasing female reach drive more bookings? We compared four female metrics (Reach, IMP, Click, Install) against total bookings at various lags to find out.
| Metric | Lag 0 | Lag +5d | Lag +7d | Lag +10d | Lag +12d | Lag +14d | Lag +21d | Interpretation |
|---|---|---|---|---|---|---|---|---|
| Female Reach | -0.13 | -0.13 | -0.02 | 0.07 | 0.04 | -0.03 | 0.25 | Near-zero correlation |
| Female IMP | -0.04 | -0.01 | 0.10 | 0.18 | 0.15 | 0.07 | 0.30 | Weak positive only |
| Female Click | -0.56 | -0.77 | -0.73 | -0.64 | -0.60 | -0.58 | -0.37 | Strong negative at all lags |
| Female Install | 0.09 | 0.30 | 0.47 | 0.66 | 0.70 | 0.69 | 0.80 | Strong positive (leading indicator) |
Three key findings:
| Week | F. Reach | F. Click | F. Install | Bookings | Note |
|---|---|---|---|---|---|
| Feb 2 | 201,898 | 2,818 | 345 | 16 | High installs → booking peak 2 weeks later |
| Mar 16 | 213,054 | 5,693 | 12 | 23 | Highest reach, yet installs collapsed |
| Mar 30 | 162,445 | 7,997 | 8 | 11 | Most clicks — bookings plummeted |
| Apr 20 | 145,646 | 8,684 | 2 | 5 | Reach up, clicks at all-time high, installs collapsed → zero bookings |
The Apr 20 week paradox: Female reach was +18% WoW, clicks hit a monthly high of 8,684. Yet bookings collapsed to 5 (with three consecutive zero days). The decisive factor: female installs dropped to just 2.
"More reach = more bookings" is false. Whether women install the app is the only leading indicator that matters.
Strategic implications:
| # | Topic | Key Question |
|---|---|---|
| 1 | LP test decision | Continue the LP test or revert? What's the threshold for reversal? |
| 2 | Seg B/C revival | If female install collapse is linked to ad set consolidation, should we partially restore Seg B/C? |
| 3 | Creative refresh | Impressions up but installs down = creative fatigue signal. When do we introduce new creatives? |
| 4 | JP segment strategy | Should we design dedicated targeting for tourists currently in Japan? (highest efficiency segment) |
Proposed immediate actions: