LTV — What It Is, Predicting User Lifetime Value

What Is LTV?

LTV (Lifetime Value) is the predicted net profit that one user will generate over the entire period of using the product. In mobile development, LTV is the cornerstone of app economics: this metric determines how much can be spent on acquiring users through advertising.

The metric combines three components: average revenue per period (ARPU), average user lifetime, and gross margin. Without LTV, it is impossible to justify a UA (User Acquisition) budget. Investors evaluate a startup by LTV multiplier: SaaS companies are valued at 5–10× LTV of annual subscription.

According to OpenView (2025), the median LTV in mobile apps is $14.50 for the Games category, $22 for Social, and $45 for Health & Fitness. The range is enormous — from $0.50 in Hypercasual to $500+ in Dating apps with long lifetime.

Why LTV Matters in Mobile Development

LTV solves three business problems. First — assessing advertising ROI. If LTV is less than CPI, the app is unprofitable. Second — feature prioritization: product changes that increase LTV are implemented. Third — revenue forecasting: knowing the LTV of new users allows predicting revenue 6–12 months ahead.

The second and third tasks require historical data. For a new app without history, category benchmarks are used. Sensor Tower and Data.ai provide LTV estimates by category and region. However, benchmarks are a reference, not the truth: a specific app’s LTV can differ 2–3 times from the market average.

LTV vs Lifetime Gross Profit

LTV can be calculated based on revenue (Revenue LTV) or profit (Profit LTV). Profit LTV subtracts cost of goods sold (COGS) from Revenue LTV: store commissions (15–30%), server costs, payment processing. Apple and Google take 15–30% of each purchase — this significantly reduces the real LTV. For correct comparison with CAC, use Profit LTV.

How to Calculate LTV

There are three methods for calculating LTV: historical, cohort, and predictive. Each method provides different accuracy at different horizons. Historical LTV is calculated for users who have already completed their lifecycle — it is accurate but inapplicable for active users.

The cohort method is the gold standard. Users are grouped by install date (Day 0 cohort), and a revenue curve is built for each cohort by day. Extrapolating this curve provides an LTV forecast. According to Amplitude (2025), cohort LTV for a 90-day window has a margin of error of only 5–8% when forecasting 12 months ahead.

data class CohortData(
    val installDate: LocalDate,
    val dayRevenue: List<Double>
)

fun calculateProjectedLTV(
    cohort: CohortData,
    dayN: Int,
    annualDiscountRate: Double = 0.10
): Double {
    val observedRevenue = cohort.dayRevenue
        .take(dayN)
        .sum()
    val projectedRevenue = extrapolateRevenue(
        cohort.dayRevenue, dayN, 365
    )
    return applyDiscountRate(
        observedRevenue + projectedRevenue,
        annualDiscountRate
    )
}

The extrapolateRevenue method uses a revenue decay curve. It is typically approximated by a power law: revenue on day N = revenue_day_1 × N^(-alpha). Alpha is calculated from historical data. ApplyDiscountRate brings future revenues to present value — standard financial practice for correct valuation.

Basic LTV Formula

The simplest LTV formula: ARPU × Average Lifetime. Average lifetime is Average Churn Rate^−1. If users churn at 10% per month, Average Lifetime = 1 / 0.10 = 10 months. With Monthly ARPU of $2, LTV = 2 × 10 = $20. This formula is useful for quick estimates but ignores ARPU changes over time.

LTV Prediction Models

LTV prediction is a complex task requiring advanced statistical methods. For accurate early-stage forecasts (Day 1 — Day 30), ML models are used that analyze behavioral signals: first purchases, session frequency, and engagement.

Probabilistic Models — BG/NBD (Beta-Geometric / Negative Binomial Distribution) — a classic approach for predicting repeat purchases. The model estimates the probability that a user will remain active and the expected number of transactions. The Fader & Hardie model (2013) remains the de facto standard for LTV without ML.

ML Models — Random Forest, XGBoost, and neural networks. They take into account dozens of features: traffic sources, device, country, first actions. Google Cloud recommendation.ai uses Deep Neural Networks for LTV prediction based on first 7 days of events. The accuracy of such models is 15–25% higher than probabilistic models (Google, 2025).

Hybrid approach combines a probabilistic model for baseline prediction and ML correction for groups with anomalous behavior. In practice, hybrid gives the best balance of accuracy and computational cost. Most platforms (Amplitude, Mixpanel) use hybrid models for automatic LTV.

Day-7 LTV Forecast

Day-7 LTV is the most popular early forecast. If a user has not made a purchase within 7 days, the probability of future purchase drops by 80%. Day-7 LTV is calculated by multiplying Daily ARPU over 7 days by the predicted Lifetime. The correlation of Day-7 LTV with 365-day LTV reaches 0.85 in the Games category.

LTV to CAC Ratio

LTV / CAC ratio is the primary unit economics metric for a mobile app. The rule for a healthy business: LTV/CAC >= 3. If the ratio is less than 3, the business does not scale — each acquired user generates insufficient revenue to cover marketing costs and operating expenses.

Venture investors require LTV/CAC > 3 for seed round and > 5 for Series A. According to OpenView (2025), the median LTV/CAC for mobile-first companies is 3.2. Top companies achieve 5–7. The payback period should not exceed 12 months for subscription models.

It is important to note that LTV/CAC is calculated based on predicted LTV, not historical. If the forecast is overstated — the ratio appears healthy while the actual business is unprofitable. A conservative forecast of LTV based on P50 (median) is more reliable than one based on the average.

LTV by App Category

LTV varies significantly by category. Understanding typical values helps set realistic goals and choose a monetization strategy. The data below is based on Data.ai and Sensor Tower reports for 2025.

Hypercasual Games have the lowest LTV — $0.20–$0.50. The model is ad-based: users play for 3–5 days, watch rewarded videos, then leave. With such LTV, CAC cannot exceed $0.10–$0.15, which is achievable only with mass traffic and optimized creatives.

Strategy Games have the highest LTV in gaming: $25–$50. Users stay for 6–18 months, actively purchasing boosters and subscriptions. High LTV allows spending up to $10–$15 per install on UA. Traffic quality becomes more important than volume.

Health & Fitness — LTV of $30–$60 thanks to subscriptions. The average user pays $8–$15 per month for 4–6 months. Subscriptions provide predictable LTV and high business valuation by investors.

Social Dating — LTV reaches $100–$500. Users pay for subscriptions and virtual gifts; lifetime can exceed 12 months. This is the most profitable mobile app category in terms of LTV, but CAC is also the highest — $20–$50.

How to Increase LTV

Increasing LTV is a strategic task solved through product improvement. All methods fall into three groups: improving retention, increasing ARPU, and reducing churn. Each group has a multiplicative effect on LTV.

First group — retention. Increasing Day-7 retention by 10% yields LTV growth of 15–20%. Methods: personalized onboarding, push notifications with relevant content, gamification (achievements, leaderboards, streaks). Duolingo is a classic example: streak mechanics keep users for years, ensuring LTV 3–4 times higher than the EdTech average.

Second group — ARPU. Methods are described in the ARPU article: subscriptions, ad placement optimization, offer personalization. A 10% increase in ARPU with the same retention gives a direct 10% LTV increase.

Third group — reducing churn. Win-back campaigns (reactivating churned users), improving customer support, analyzing churn points. Reducing monthly churn from 15% to 10% increases Average Lifetime from 6.7 to 10 months — LTV growth of 50% with unchanged ARPU.