NATIVE APP · BEHAVIOURAL TELEMETRY

Native App — Deep Analysis

For CEO · CMO · Growth Fleet · Data pulled live 2026-07-20 · GA4 property 177471782, streams platform ∈ (Android, iOS)

The one-line story. The native app works at the top and leaks at the bottom — 82% onboarding, 46% reach a first workout, but only ~6% are still active at D28. And the Jul-2026 MAU "recovery" masking that leak is bought Android volume in markets that don't pay. This is behavioural evidence for the fleet's existing verdict: churn is the binding constraint.

⚠ GA4 measures behaviour, not money. Every revenue figure below is CROSS-SOURCE from RevenueCat/Stripe. Nothing was written to any live system; the cockpit was not edited.

Verification legend — every number is tagged, nothing is estimated unless labelled

✅ VERIFIED — live GA4 this session 🔗 CROSS-SOURCE — cockpit / RevenueCat 🧮 DERIVED — maths on verified inputs 🟡 DISTORTED — real but bug-inflated, don't quote ⚠ ANOMALY — data-quality bug found ❌ NOT AVAILABLE — needs RevenueCat

Native active users / 90d

10,600

✅ iOS 3,367 · Android 7,195

Onboarding completion

82%

🧮 2,243 / 2,730

D28 retention (mature)

~6%

✅ engagement, not subs

App subscribers

214

🔗 RevenueCat

App MRR

$3,089

🔗 28% of $10,978 total

Installs / 90d

8,350

✅ vs 214 paying

1Executive summary — the six findings

"The app is dark" is false. ~10,600 native users/90d have been sitting inside a property we already query — nobody filtered to the app streams. Access already worked. ✅

Activation is genuinely strong. 82% complete onboarding, 46% reach a first workout, 68% finish a started workout. The top of the product works. ✅

Retention is the leak — and it's the "unmeasurable" deal-blocking number. ~13% D7, ~6% D28 (mature). iOS retains ~2× Android. Now measured, with history. ✅

The MAU "recovery" is an illusion. The Jul rebound is entirely low-retention, low-ARPU Android installs; iOS fell ~59% straight through. Leaky bucket, not health. ✅

Money truth is in RevenueCat, correctly. 214 subs / $3,089 MRR; RC lifecycle events live. GA4 owns only pre-paywall behaviour. 🔗

Three instrumentation bugs must be fixed before board use. Double/triple-tagging, an iOS screen_view firehose, and purchase-on-initiation. ⚠

2Scale & trend — the MAU illusion

Native monthly active users, stacked iOS + Android. ✅ VERIFIED

Jan 2024Jan 2025Jan 2026Jul 2026

Read it decomposed. iOS 3,109→1,270 (−59%) and still falling. The entire Jul rebound to 5,362 is Android (1,047→4,092, +291%) — the worst-retaining, lowest-ARPU segment. The MAU line must not be celebrated without this split.

3Activation funnel ✅ VERIFIED · unique users · 90d

Onboarding completion

82%

2,243 / 2,730 🧮

Onboard → first workout

46%

the biggest drop 🧮

Workout completion

68%

696 / 1,022 🧮

⚠ Naming note: the onboarding survey is tagged take_personalized_assessment (2,730). A separate assessment_started (278) is the in-app re-assessment — a different action. Kept distinct.

4Retention — the deal-blocking number, now measurable

Weekly cohorts, native, cohortActiveUsers = engagement return. ✅ VERIFIED ❌ subscription retention needs RevenueCat

Cohort (wk of)	Size	wk1 ~D7	wk2	wk3	wk4 ~D28
Jun 08	173	20%	12%	9%	6%
Jun 15	184	17%	9%	9%	7%
Jun 22	1,390	14%	6%	3%	1%*
Jun 29 SURGE	3,393	11%	4%	1%	0%*
Jul 06	889	11%	3%	—*	—*
Jul 13	424	8%	—*	—*	—*

* right-censored — cohort too recent to have reached that week; ignore.

iOS ~D7

18%

iOS ~D28

Android ~D7

10%

Android ~D28

The tell: the 3,393-user Jun-29 surge cohort retains worst (11% D7). Bigger cohort, worse retention — the signature of low-intent acquisition.

5Monetisation — GA4 is broken here; RevenueCat is truth

GA4 behaviour (pre-paywall) ✅

Landed on pricing page	1,205
Premium offer viewed	1,273
Premium feature blocked	711
Trial started	102
Purchase	252 ⚠ broken

RevenueCat / cockpit (money truth) 🔗

App subscribers	214
App MRR	$3,089
Total MRR	$10,978
RC events live	trial/convert/renew/cancel

Why GA4 money is unusable: purchase == purchase_initiated (both 252), and purchase-users (252) exceed trial-users (102) — an impossible "247% trial→purchase". It fires on initiation. Use RevenueCat for every money number; use GA4 only for the pre-paywall behaviour RC can't see.

6Geography — why installs ≠ revenue

Native active users by country, 90d. ✅ VERIFIED

The largest native audiences (India, South Africa, Pakistan) are low-willingness-to-pay markets. This is the mechanism behind 8,350 installs → 214 paying subs. Native install volume is not a revenue lever unless geo-targeted. ✅ geo · MODELLED ARPU-by-country inference.

7Engagement quality

Engagement rate

77%

✅ plausible

Sessions / user

3.5

🧮 36,677 / 10,600

Avg session · views/session

51.7m · 28.6

🟡 DISTORTED — don't quote

iOS screen_view firehose: iOS logged 1,023,532 screen-views vs Android's 26,112 (~304/user). It's firing on a loop in the iOS build and inflates every time/view metric. Session-level engagement rate survives; duration and views/session do not.

8Data-quality register — fix before any board deck

#	Bug (all VERIFIED present)	Effect	Owner
1	Double/triple event-tagging (legacy + new names fire together)	Any single count ambiguous; handled here by alias map	Nic / product
2	iOS `screen_view` firehose (1.02M vs 26k)	Inflates session-duration & views/session	Nic / product
3	`purchase` fires on initiation (== `purchase_initiated`)	GA4 unusable for revenue	Nic / product
4	Native data from 2020-06-11 (pre-GA4-GA)	Verify property creation before claiming "6yr history" NEEDS-CONFIRM	Analytics

9What this means for the cockpit 🔗 read-only · cockpit NOT edited

Cockpit fact 🔗	What native GA4 adds ✅
App MRR $3,089 / 214 subs (28% of MRR)	The app is ¼ of revenue but the majority of daily product usage (~10,600/90d) — under-weighted in a web-MRR-led view.
Cancel queue 108 / $1,504 (paid churn)	Upstream free churn now visible: `app_remove` 2,599/90d vs only 354 `comeback_after_7_days`. The paid queue is the tip; D28≈6% is the iceberg.
Break-even $15K by 1 Nov	Activation isn't the blocker (82%/46%) — retention is. Money spent adding users leaks by D28. Confirms "churn is the constraint."
Meta MTD $2,268 · Google MTD $1,056	Leaky-bucket flag: the Jul Android surge (+291%) into low-ARPU geos at 10% D7. If paid UA funds it, spend buys non-retaining installs.
Exit / data-room (CXO)	We can now show D7/D28 cohort retention by platform with history — the deal-blocking gate. Low, but measured beats "we don't know."

Proposed (not built): a read-only "Native App" cockpit tab — after the §8 bugs are fixed so numbers are quotable. Awaiting go.

10Recommended actions by seat

Seat	Action	Why
Growth	Confirm if the Jul Android surge is paid UA; geo-target or pause.	10% D7 into low-ARPU geos = spend leaking.
CPO / product	Attack onboard→first-workout (46%) & the D7→D28 collapse.	Activation strong; retention is the constraint.
Analytics	Own the 3 instrumentation fixes; reconcile GA4 behaviour to RevenueCat.	Not board-quotable until fixed.
CXO / exit	Use D7/D28 native cohorts (with history) in the data room — honestly.	Turns "dark" into a measured number.
CEO	Read the MAU line decomposed (§2).	Prevents a false-positive strategic read.

Reproducibility. Tool: tools/app-analytics/native_app_analytics.py --json → docs/company/ANALYTICS/NATIVE-APP/native_app_report.md. Auth: robot google-analytics-claude-access@the-movement-athlete…, property 177471782, filter platform ∈ (Android, iOS). Cross-source: docs/company/ANALYTICS/GROWTH-SCOREBOARD/canonical_facts.json (Stripe/RevenueCat, 2026-07-20 11:12).

No number here is hallucinated. Each is tagged ✅ VERIFIED (live GA4 this session) · 🔗 CROSS-SOURCE (cockpit/RevenueCat) · 🧮 DERIVED (inputs shown) · 🟡 DISTORTED · ⚠ ANOMALY · ❌ NOT AVAILABLE. Companion markdown: NATIVE_APP_DEEP_ANALYSIS_2026-07-20.md.