13 May – 13 Jul 2026 · Firebase events · 467 notifications, of which 97% are the automated daily
book-recommendation broadcast to all users (1–3×/day). The only distinct campaigns: the Jarir announcement and the onboarding welcome push.
Delivered
11.85M
Android-measurable only — iOS delivery isn't logged
Opened
55,452
≈ 0.47% open rate · 89 dismissals for every open
Led to listening
9,882
Listening within 1h of a tap ≈ 160 sessions/day
Subscriptions
95
Within 24h of a tap, over 2 months — push is not a sales channel
The funnel: the leak is at the open
Once someone taps, push works — most open the app and 1 in 6 starts listening within the hour.
Almost nobody taps. Each bar shows the conversion from the step above it.
Delivered (Android)
11,847,798
Opened
55,452
0.47% — the leak
Opened the app
34,787
62.7% of opens
Started listening
9,882
17.8% of opens
Subscribed (24h)
95
0.17% of opens
Weekly deliveries kept growing…
Millions of notifications delivered per week (Android).
…while the open rate drifted down
Weekly open rate (%). The mid-May step-up is an iOS measurement change, not real growth.
What people actually open: targeted beats broadcast, and the book matters
Open rate by message (sends with ≥ 1,000 deliveries). The triggered welcome push beats every broadcast;
among broadcasts, fiction & thrillers run 3–6× ahead of heavy non-fiction on the same audience.
Send time: evening wins
Open rate by send slot (Cairo time). The 9am slot carries the most volume at the worst rate.
Read the numbers with these in mind
!
iOS delivery is invisible. All 11.85M “delivered” are Android. iOS users get and open pushes (31% of opens) but iOS reach can’t be sized — open rates are directional, Android-anchored.
!
iOS opens jumped ~7× the week of 25 May — almost certainly an app-release or instrumentation change. Worth one question to engineering.
?
“Conversion” is same-device timing, no control group. Some of that activity would have happened anyway — treat it as an upper bound.
What I'd do with this
Shift from broadcast to targeted. The welcome push (1.24%) proves relevance ≈ 2.6× the broadcast average. Candidates: continue-your-book nudges, new-episode alerts for series a user follows, genre-based daily picks.
Test frequency down, not up. 89 dismissals per open, and the open rate fell as weekly volume rose — classic fatigue. A/B 1×/day vs the current 1–3×/day; watch opt-outs and listening, not just opens.
Move weight from 9am to the evening. 11pm converts ~2× the 9am slot, yet 9am is the single biggest slot (2.04M deliveries).
Pick titles like a merchandiser. A simple playbook for whoever queues the daily book: fiction/thriller sustains 0.7%+; heavy classics sit near 0.2%.
Fix measurement before the next review: iOS delivery logging (or adopt FCM-console numbers as the iOS denominator) and confirm the 25-May iOS change.
Judge push on listening re-engagement (~160 listening sessions/day today), not subscriptions.