App marketing KPIs in 2026: the 7 core metrics you need to know

November 2, 2021

Tin Le

Updated on October 17, 2026: Adtek rewrote this entire article with the latest data, platform features and regulations from 2025 to 2026.

App marketing KPIs are the set of metrics that answer three questions about an app: how much it costs to acquire a new user, whether that user stays, and how long it takes them to pay back the ad money you spent. In 2026, the minimum set is CPI (or cost per first action), day 1, day 7 and day 30 retention, ARPU, LTV, cohort ROAS and churn. Everything else, such as CTR, store conversion rate or session count, is there for diagnosis when one of those seven metrics starts to drop.

The original version of this article was written in 2021 and listed nearly 30 metrics. The 2026 edition cuts that down to seven core metrics, updates the market data and explains how to read the numbers now that Apple has moved to AdAttributionKit.

The seven core app marketing KPIs for 2026

Metric	How to calculate it	The decision it supports
CPI (Cost per install)	Ad spend / installs from ads	Which channel acquires users more cheaply, which creative needs replacing
CPA for the first action	Spend / number of people who complete registration, a first top-up or a first purchase	Use it instead of CPI when the app has a clear activation step (e-wallets, marketplaces, fintech)
Retention D1, D7, D30	Users who reopen the app on day N / users who installed in the cohort	Whether onboarding works, whether the product keeps users
ARPU	Revenue in the period / active users in the period	What each user is worth, including users who never pay
LTV	Revenue (or gross profit) a cohort generates in N days / users in the cohort	The highest CPI you can pay and still make money
Cohort ROAS (D7, D30)	Cohort revenue up to day N / cost of acquiring that cohort	Whether to raise or cut budget for a channel or campaign
Churn	Users lost in the period / users at the start of the period	Whether the app is losing people faster or slower than it acquires new ones

Three points are often misunderstood. ARPU divides by all active users, while ARPPU divides only by paying users, so a freemium app with a high ARPPU can still have a very low ARPU. LTV should be calculated on gross profit if the app sells physical goods. ROAS only means something when it is tied to a day mark: a D7 ROAS of 40% and a D30 ROAS of 40% describe two very different situations.

Metrics from the 2021 version that moved to the diagnostic tier

CTR, click-to-install rate (CTI), store conversion rate, organic share, sessions, DAU/MAU, uninstall rate, eCPM and install fraud rate still need tracking, but one level down. When CPI rises 30% in a week, open CTR and store conversion rate to see whether the problem sits in the creative or on the store page. ROX (return on experience) from the old version has been dropped because there is no consistent way to calculate it.

The 2025 to 2026 app market shifted its focus from downloads to revenue

According to Sensor Tower’s State of Mobile 2026 report (January 2026), downloads on iOS and Google Play grew just 0.8% in 2025, to nearly 150 billion. In-app purchase revenue grew 10.6% to USD 167 billion, and for the first time non-game apps earned more than games. For games, Sensor Tower advises UA teams to set targets around payback time rather than install volume.

Adjust’s Mobile App Trends 2026 report (February 2026), on the other hand, recorded a 10% rise in installs in 2025 and a 7% rise in sessions. The two figures do not contradict each other: Sensor Tower estimates the whole market, while Adjust measures the apps that use its tools. Whenever you put a benchmark into a report, always cite the source and the dataset.

Vietnam: installs have plateaued and iOS is gaining share

AppsFlyer’s State of App Marketing in Vietnam 2025 report (October 2025), based on 13.6 billion installs from 2017 to 2025, shows that app installs in Vietnam reached 2.17 billion in 2024 and then fell 2% in 2025. The iOS share of installs rose from 22% in 2017 to 33% in 2025. For finance apps, iOS was forecast to account for 54% of installs in 2025, according to a Vietnamese-language news story on the report (October 2025). That means most of a finance app’s valuable users sit on the platform that is hardest to measure.

Retention D1, D7, D30: the app marketing KPI most often misread

Day N retention rate is the share of users in a cohort who come back and open the app on the Nth day after installing. D1 reflects onboarding, D7 tells you whether users found value in the first week, and D30 tells you whether the app has become a habit.

These three marks are now moving apart. AppsFlyer’s State of Marketing in APAC 2026 report (August 2026), based on 30 billion installs from nearly 12,000 apps, found that D7 retention for finance apps on Android in Southeast Asia rose 45% year over year, while D30 fell 32%. D30 retention declined across all four verticals (finance, shopping, entertainment, gaming), all four regions and both platforms, with no exceptions.

The same report shows Android UA spend in Vietnam rising 194% for entertainment apps, 153% for finance apps and 106% for games, according to Thị trường Tài chính Tiền tệ, a Vietnamese financial news outlet (September 2026). When budgets climb sharply while D30 goes down, it is time to check your numbers again: part of a good D7 can come from new-user offers, 7-day reward missions or first-time cashback, and users leave once the offer runs out.

Where to get retention benchmarks

A single average across all industries is close to useless, because retention for news apps, casual games and e-wallets differs several times over. AppsFlyer’s free benchmark tool gives figures by vertical and country. According to its methodology page, the data is updated every quarter, retention is calculated on two-month cohorts, and the top and bottom 10% of apps are excluded; when Vietnam does not meet the data threshold, the tool shows subregion figures instead. Even so, comparing this month’s cohort with the same month last year for your own app is more reliable than any industry average.

Privacy-first measurement changes how you read app marketing KPIs on iOS

Since Apple switched on App Tracking Transparency in 2021, most iOS data can no longer be read at the user level. Adjust’s Mobile App Trends 2026 report puts the average tracking opt-in rate in Q1 2026 at 38%, up from 35% in Q1 2025, counted among users who were shown the permission prompt. AppsFlyer published a rate of 50% in April 2025. The two numbers differ because the denominators differ, but the conclusion is the same: at least half of the iOS users who are asked can only be measured through Apple’s aggregated framework.

That framework used to be SKAdNetwork. With iOS 17.4 (March 2024), Apple introduced AdAttributionKit; Apple’s documentation says the system considers impressions from both frameworks when deciding which ad gets credit for an install. Three rules directly affect your KPIs:

There are three conversion windows: days 0 to 2, days 3 to 7 and days 8 to 35, counted from the first app open, with at most one postback per window.
The first postback arrives after its window closes, with a random delay of 24 to 48 hours; the second and third postbacks are delayed 24 to 144 hours. D7 ROAS from Apple data is therefore only available after 8 to 13 days.
Fine-grained conversion values only come with the first postback. Later postbacks, and campaigns with few installs, only get coarse values: low, medium, high.

The consequence: D30 retention and per-campaign LTV on iOS have to be estimated, not counted. The common approach is to use the users who opted in to tracking as a sample to infer the whole cohort, then cross-check against aggregated postbacks and incrementality tests. The conversion value mapping (for example, high means the user topped up within 48 hours) needs to be agreed between the product team and the ads team before the campaign runs.

Android has gone a different way. On October 17, 2025, Google announced it would retire the Attribution Reporting API along with several other Privacy Sandbox technologies on Chrome and Android, citing low adoption. As of this writing, Android has no measurement change comparable to ATT, so report Android and iOS figures separately and do not add them together.

App marketing KPIs by app stage

A newly launched app and an app with a million users should not be looking at the same set of metrics. This is the framework Adtek recommends:

Stage	Primary KPIs	Diagnostic KPIs	The question to answer
Launch (months 0 to 3)	D1 and D7 retention; onboarding completion rate	Store conversion rate, CTI, crash rate	Do users understand what the app does?
Channel discovery (months 3 to 9)	CPA for the first action, D30 retention by channel	CPI, organic share, install fraud rate	Which channel brings users who are still around after 30 days?
Budget scaling	D7 and D30 cohort ROAS, payback period	CPI by creative, ARPU, payer rate	If budget rises 50%, how much does D30 ROAS drop?
Maturity	LTV/CAC, monthly churn, DAU/MAU	Reactivation rate, ARPPU, purchase frequency	Is it cheaper to keep existing users or buy new ones?

Cohort ROAS and LTV: how to set your maximum CPI

Cohort ROAS takes the revenue from people who installed on a given day or week, accumulates it up to D7, D30 and D90, and divides it by the cost of acquiring that same group. The ROAS in your ads manager pairs today’s revenue with today’s spend, even though the two belong to different groups of people.

Here is an example with hypothetical numbers. An e-commerce app spends VND 100 million to buy 10,000 installs, a CPI of VND 10,000. This cohort generates VND 25 million in revenue by D7 and VND 60 million by D30, so D7 ROAS is 25% and D30 ROAS is 60%. With a 30% gross margin, gross profit by D30 equals only 18% of the cost. To know whether the cohort pays back within 6 months, you need the revenue curves of older cohorts beyond D30. Without that curve, every ROAS target is a guess.

Churn is the other side of retention. When monthly churn outpaces the rate at which you acquire new users, MAU falls even if the budget stays the same. Break churn down by cohort to see whether new users leave quickly (an onboarding problem or the wrong ad audience) or older users drift away over time (the product has run out of new value).

Adtek’s view: a checklist before you lock in your app KPIs

When two teams look at the same app and come up with two different ROAS figures, the cause is usually the definitions, not the ads. Before talking budget, answer these eight questions:

Is there a single MMP (AppsFlyer, Adjust or equivalent) acting as the shared source of numbers, or does each channel report its own?
Does “active user” mean opening the app, logging in or completing a specific action? This definition drives both retention and ARPU.
Has the first valuable action (registration, top-up, first order) been sent back to the MMP and the ad platforms?
Have the conversion values for SKAdNetwork and AdAttributionKit been mapped to business KPIs, or are they still on defaults?
Is LTV calculated on revenue or on gross profit, and does it net out refunds and cancelled orders?
Is ROAS reported by cohort or by transaction date?
Are iOS and Android figures kept separate, with the estimated portion clearly labeled?
Is there an install fraud threshold that automatically excludes bad sources?

Once those eight points are settled, a weekly dashboard usually needs only five to seven metrics. The next step is choosing channels for each stage, which is covered in our article on app marketing strategy.

If you need to review how your app is measured before raising next quarter’s budget, contact the Adtek team.

This article was rewritten and updated by Adtek in 2026, drawing on original content from AppsFlyer (2021).