← Back to Academia

15.K3 - Кейс: Mobile App - retention-driven growth

Retention = growth для мобильных приложений. Если D30 retention < 15% - не масштабируйте UA. Fix retention first.

#advanced #case #block-15 #mobile


Навигация

15.K2 - Кейс: E-commerce - от 0 до масштаба


Retention-Driven Growth Model

Improve Retention
D1: 45% → 55%
D7: 25% → 35%
D30: 12% → 20%

Higher LTV
$5 → $8.50

Can afford higher CPI
$1.50 → $2.50

Scale UA profitably
More installs at same ROI

More users = more data
= better product decisions

Key insight: 5% improvement in D30 retention → 25-40% increase in LTV → can outbid competitors on UA.


Кейс: Финтех-приложение (mobile banking KZ)

Baseline Metrics

MetricBeforeIndustry Benchmark
D1 Retention42%40-55%
D7 Retention22%25-35%
D30 Retention11%15-25%
Activation Rate35% (registration → first transaction)40-60%
ARPU (monthly)₸2,500-
CPI (blended)₸1,200₸1,000-3,000

Diagnosis: D1 OK, but D1→D7 drop = 48%. Problem is between first open and habit formation.

Root Cause Analysis

Drop-off pointHypothesisData Source
D1→D3Onboarding too long (8 screens)Amplitude funnel
D3→D7No trigger to return after first usePush analytics
D7→D30Core value not discoveredFeature adoption metrics

Experiments Roadmap

#ExperimentMetricICEResult
1Shorten onboarding: 8 → 4 screensActivation rate8.3+12% activation ✅
2Push notification D1: «Complete setup → get ₸500 bonus»D3 retention7.7+8% D3 retention ✅
3Smart push D5: transaction reminderD7 retention7.0+5% D7 retention ✅
4Weekly financial summary (push + in-app)D30 retention6.7+4% D30 retention ✅
5Referral: invite friend → both get ₸1,000K-factor6.3K-factor 0.15 → 0.28 ✅
6Gamification: savings challengeD30 retention5.7+2% (modest) ⚠️

After Optimization

MetricBeforeAfterΔ
D1 Retention42%48%+6pp
D7 Retention22%30%+8pp
D30 Retention11%18%+7pp
Activation35%47%+12pp
ARPU₸2,500₸3,200+28%
LTV (6mo)₸6,750₸12,096+79%
Max CPI (profitable)₸1,687₸3,024+79%

Result: Can now spend 79% more per install and remain profitable → scale UA aggressively.


Mobile App Retention Toolkit

Push Notifications Strategy

TimingMessageGoal
D0 (install)Welcome, complete setupActivation
D1«Your account is ready - make first transfer»First action
D3«₸500 bonus waiting - complete 1 transaction»Habit start
D7«Weekly summary: you saved ₸12,500 this week»Engagement
D14«New feature: bill payments in 1 tap»Feature discovery
D30«Your monthly financial report is ready»Long-term habit

Rules:

  • Frequency cap: max 1 push/day
  • Opt-in optimization: ask AFTER first value (not on install)
  • Personalize: name, specific amounts, relevant features
  • Timing: based on user's typical active hours (Amplitude cohorts)

In-App Messages

TriggerMessageFormat
First openOnboarding flow (4 screens max)Full-screen
After first transaction«Great! Here's what else you can do»Bottom sheet
3 days inactive«Complete a transfer → earn ₸500»Banner
Feature unused after D7«Did you know you can pay bills here?»Tooltip

Feature Flags & Remote Config

ToolWhat it does
Firebase Remote ConfigChange app behavior without release (feature toggles, copy, thresholds)
Firebase A/B TestingServer-side A/B tests on config values
StatsigFeature flags + experiments + analytics
LaunchDarklyEnterprise feature management

Attribution for Mobile

MMP Setup (AppsFlyer)

In-App Events tracked: - af_complete_registration (activation) - first_transaction (aha-moment) - af_purchase (loan disbursed) - monthly_active (custom, fires on D30) OneLink: deferred deep linking for all campaigns - Web ads → App Store → open specific screen - Referral links → install → credited referrer SKAN Conversion Value Schema: Bit 0-1: Revenue bucket (0, ₸1-5K, ₸5-20K, ₸20K+) Bit 2-3: Engagement (registered, first tx, 3+ tx, weekly active) Bit 4-5: Time to convert (D0, D1, D2-3, D4-7)

🔧 Практика

  1. Для мобильного приложения (выберите тип: финтех / e-commerce / utility):
    • Определите retention benchmarks (D1/D7/D30)
    • Проанализируйте текущий D1→D7 и D7→D30 drop-off
    • Предложите 5 экспериментов (push, onboarding, feature, referral)
    • Рассчитайте: как +5pp D30 retention влияет на LTV и max CPI
  2. Создайте push notification calendar (D0-D30)
  3. Спроектируйте SKAN conversion value schema

Связанные заметки