App Metrics Explained: DAU, Retention, ARPU, LTV, CPI

Every app dashboard throws the same handful of acronyms at you, yet two teams can quote the same metric and mean different things. This guide defines the five numbers that matter most, shows how Apple and Google actually compute them, and flags the places where the definitions quietly diverge.

The five metrics at a glance

The formulas look simple. The difficulty is in the denominators, the time windows and the fact that the two stores count different things.

DAU and MAU

Daily active users is the count of distinct users who opened the app on a given day. Google Play Console defines DAU exactly that way, and defines MAU as users who opened the app within a rolling 28-day period, not a calendar month ³. Apple's App Store Connect reports "active devices" rather than users, and its usage metrics only include people who agreed to share diagnostics and usage data with developers, so the absolute figures are a sample, not a census ¹.

Two practical consequences follow. First, a Play Console MAU and an App Store Connect active-device count are not comparable, because one counts people across devices and the other counts opted-in devices. Second, the DAU/MAU ratio (often called stickiness) is only meaningful if both numbers come from the same system with the same window. GameAnalytics reported a median DAU/MAU of roughly 4–5% for PC games in 2025, which shows how low the "normal" range can be outside the big social apps ⁹.

Retention

Retention is the share of a cohort of new users who return on a later day. Apple's definition: the percentage of active devices that installed the app on a chosen day and opened it N days later. Devices that install but never open are excluded from both the numerator and the denominator, and a cohort's denominator can grow after the fact if someone installs on January 1 but first opens on January 10 ¹.

Google reports retention at days 2, 3, 7, 14 and 28 rather than days 1, 7 and 30, and it anchors each cohort to the day the user first opened the app, counting a return on any of their devices ³. Apple's peer-group benchmarks use day 1, 7 and 28 and compare you against the 25th, 50th and 75th percentile of a peer category, with differential-privacy noise added to every value ².

Classic versus rolling retention

"Classic" (or N-day) retention counts a user only if they are active on exactly day N. "Rolling" retention counts them if they return on day N or any later day, so it is always higher. GameAnalytics states its game benchmarks use classic retention ⁹; most third-party blog tables do not say which they use, which is one reason published numbers disagree so much.

ARPU and ARPDAU

Average revenue per user divides revenue by the number of users in the period. ARPDAU divides a single day's gross revenue by that day's DAU and is the standard for ad-supported games. Play Console reports ARPDAU in USD, using the PST8PDT time zone, so that peer comparisons line up ³. Note that Play revenue figures are estimated gross sales including tax, whereas Apple's benchmark metric is proceeds per paying user, meaning what the developer receives after Apple's commission, divided by paying users ^2,3. Comparing gross ARPU on one store with net ARPPU on the other will make Android look better than it is.

RevenueCat's 2026 report (covering 2025 data, 115,000+ apps) gives a useful reference point for subscription apps: median revenue per install at day 14 was $0.23 globally, $0.48 for health and fitness and $0.08 for games, while apps with a hard paywall reached $2.32 ⁷.

LTV

Lifetime value is the net revenue you expect from a user across their whole relationship with the app. The simplest working formula is:

LTV ≈ ARPU per period × expected lifetime in periods, or equivalently ARPU per period ÷ churn rate per period.

Because real lifetimes are unknown at the time you need the number, most teams use a realized LTV at a fixed horizon. RevenueCat reports year-one realized LTV per paying subscriber at a $23 global median, $44 or more for the top quartile, and a category spread from $35.64 (health and fitness) down to $11.22 (games) in 2025 ⁷. Apple's App Store Connect benchmarks offer a 35-day equivalent, "Day 35 proceeds per download" and "Day 35 download to paid conversion", which are the closest thing to a native LTV proxy the store provides ².

CPI

Cost per install is paid media spend divided by installs attributed to that spend. Business of Apps' February 2025 compilation put global iOS CPI at roughly $1.50–$3.50 and Android at $1.50–$4.00, with North America the most expensive region at $2.50–$5.00 and games costing more on iOS than on Android across every genre it listed ⁸. These are wide ranges from a secondary aggregator, so treat them as orientation rather than targets.

The number that actually matters is CPI against LTV. A $3 install is cheap for a health app with a $35 one-year LTV per payer and a 2.9% download-to-paid rate, and ruinous for a game at $11 and 1.0% ⁷.

Pitfalls that distort every one of these

Cohort versus calendar views

A "retention" chart plotted by calendar date mixes users who installed yesterday with users who installed a month ago. Only a cohort table, rows for install day and columns for days since install, lets you see whether a product change moved the curve. Apple's retention report is laid out exactly this way and adds an average row across visible cohorts ¹.

Attribution windows

An install only counts toward CPI if it happens within the attribution window of an ad engagement. Apple's SKAdNetwork allows 30 days from a StoreKit-rendered ad or attributable web ad click, but only 24 hours from a view-through impression, and the user then has 60 days to launch the app ⁴. Adjust's defaults are 7 days for a device-matched click and 24 hours for an impression, adjustable between 1 and 30 days for clicks ⁶. Shorten the window and CPI rises; lengthen it and you start claiming organic installs.

Delayed and coarse signals on iOS

SKAdNetwork 4 sends up to three postbacks for conversion windows of days 0–2, 3–7 and 8–35 after first launch, each delayed by a random 24–48 hours (first) or 24–144 hours (second and third). Lower-tier campaigns receive only a coarse conversion value or no second and third postback at all ⁵. iOS campaign LTV is therefore always a modeled figure, never a direct read.

Time zones and privacy thresholds

Play Console uses Pacific time for installs and UTC for financial data ³, while Apple blanks retention cells for low-volume cohorts and only shows benchmarks when a peer group has enough apps ^1,2. A "missing" day is often a threshold, not a bug.

How this shows up in the charts

Top-chart positions are driven primarily by install velocity, so CPI and attribution windows decide who can buy their way onto the Top Free list for a week. Retention and LTV decide who stays there. An app with a high D1 but a thin D30 tends to spike and fade; an app with a flat retention tail keeps generating organic installs that compound with paid ones. Neither Apple nor Google publishes its ranking formula, so the exact weight of engagement in chart rank is unverified. The pattern worth watching is which apps hold Top Grossing positions for months: those are far more likely to have download-to-paid and renewal numbers in RevenueCat's top quartile than to have the cheapest installs ⁷.