Subscription Metrics for Mobile Apps: A Measurement Guide
Subscription revenue is not a single number. It is the output of a funnel: installs, trial starts, trial conversions, renewals, cancellations and refunds. Each stage has its own metric, and each metric has a different place where it must be measured. This guide maps the funnel, explains what each metric means, and shows where the measurement has to happen so the numbers you act on are actually accurate.
How subscription revenue actually accumulates
A subscription app earns through two mechanisms: selling new subscriptions and keeping existing subscribers. Improving either one raises revenue; improving both compounds growth. The practical implication is that every point between install and cancellation is a place where revenue can be gained or lost.
Recurring revenue is more predictable than one-off purchases, but it comes with an expectation: users keep paying only while the app keeps delivering new content, updates or recurring value. An app that stops changing tends to see subscribers leave. That is why the metrics below are not just accounting figures — they are signals about whether the product is still worth paying for.
A useful way to picture the business is a funnel from install to cancellation. The stages are: install, virality, paid acquisition, trial started, trial converted, renewal, cancellation and refund. Each stage has a metric, and each metric should be tracked from the point where it is actually observable.
Health check: revenue and MRR
Two base metrics tell you whether anything you change later actually improved the business.
Revenue is the money generated. With mobile subscriptions it is harder to track than it looks. If a trial is offered, no revenue appears until the trial ends, and most revenue arrives through renewals. Renewals happen whether or not the user opens the app, so revenue events must be tracked on the server side, independently of in-app usage.
App store fees also matter. Fees are commonly 30% or 15%, and the 30% rate can drop to 15% for some users once they have stayed subscribed beyond a year. That means proceeds (take-home) should be separated from gross revenue. A blended take-home rate between 70% and 85% is realistic, and the difference has a real effect on cashflow forecasting.
Monthly recurring revenue (MRR) is the metric subscription businesses cite most often, and it is usually more informative than raw revenue. It smooths the picture in two ways. First, it acts as a de-noising filter, which helps when an app sells several durations such as monthly, annual and lifetime. Second, Apple and Google pay out the full amount of a longer subscription at purchase, which creates revenue spikes that can hide trends from shorter, cheaper plans. MRR clarifies those trends.
MRR is based on currently active subscriptions and their durations, which change daily, so it can be expressed as a daily metric. There is no single official formula; what matters is that you pick one method and use it consistently as a health indicator.
Top of funnel: installs, virality and paid acquisition
Installs are the entry point. More installs generally mean more subscriptions and more MRR, but before spending to drive volume it is worth finding leaks — the obvious places where users drop off. The same funnel and metrics apply whether you have 10 installs a day or 10,000.
Virality is when one user brings another, who brings another. Growth can be exponential because users do the distribution work. Paid acquisition is the most common install strategy for mobile businesses. The basic test is unit economics: if acquiring a user costs less than the revenue that user eventually produces, and the pattern can be repeated at scale, the acquisition channel is viable.
Trial started
Free trials are a common and effective way to sell subscriptions, though they do not suit every app. The metric at this stage is how many users begin a free trial.
Measurement is usually straightforward because trial starts can be captured from in-app events. One caution: some mobile analytics SDKs that claim to track in-app purchases automatically do not count free trials correctly, so verify the trial-start number against a second source before trusting it.
There is a lot of room to experiment here. Onboarding is the first sales pitch: it should explain the value proposition clearly and quickly. A useful research exercise is to download top-grossing apps in your category and note how long their onboarding takes, how much text it uses, and which engagement tactics they use before presenting a plan.
After onboarding, presenting a subscription option usually makes sense even if the app is usable without one. The upsell screen is another experiment surface. Offering more than one plan — different durations and prices — can change willingness to pay. Price, trial length, subscription length, and even the number and order of options can all affect trial start rate. A higher-priced annual plan, for example, can make a monthly plan look like better value.
The upsell should not be the only place a subscription is offered. Burying an upgrade button in settings tends to hurt trial starts. A pattern that works is to show the upsell whenever a user tries to reach a premium feature, while leaving premium entry points visible to free users so they keep encountering the reminder. More advanced tests might vary price based on how many times a user has already seen the upsell.
Trial converted
Up to this point the funnel has produced no revenue, and acquisition may already have cost money. Trial conversion measures how many trial starts become paid users.
This stage, and everything after it, can only be measured accurately on the server side. Trials convert through the app stores automatically, independent of any in-app event a device could report.
Conversion varies with trial length and with how much users depend on premium features. A 33% trial conversion rate is a reasonable baseline to start from; targeted audiences can convert much higher. There is usually less room to experiment with conversion than with trial starts, but both should be measured — a high trial start rate means little if users do not convert.
Timing of opt-outs is informative. Did users drop out immediately after first seeing premium content, or later, just before the trial converted? That points to whether to invest in earlier value or later value.
Also watch subscriptions that begin without a free trial. App stores allow a free trial to be redeemed only once, so a user who cancels a trial and subscribes again goes straight to paid. A high share of no-trial subscriptions suggests users cancel and come back to pay; testing a longer trial period is one way to see whether more of them convert.
Renewal and cancellation
Renewals are where subscription revenue is made. Keeping subscribers renewing for longer directly raises MRR, and renewal revenue arrives even with zero new installs. It is often more useful to think in terms of reducing cancellations than increasing renewals.
Cancellation, or churn, is the rate at which users unsubscribe. It can happen during a trial or after any number of renewals. Mobile app churn is usually substantially higher than in B2B SaaS: on the low end 5% to 10%, on the high end more than 30% per month.
Analyze churn by lifecycle position. Is it after the first renewal or the tenth? A cohorts table helps spot trends and shows whether churn is rising or falling over time. Churn is typically higher at the first renewal and lower with each consecutive renewal, because longer-subscribed users are less likely to leave. Watch for periods where churn jumps well above the prior period and investigate what changed.
Voluntary vs. involuntary churn
Most churn is voluntary: users choose to cancel. Involuntary churn happens when a subscription is cancelled without the user deciding to leave. It may account for roughly 20% of all cancellations.
The most common involuntary cause in mobile is billing failure. If the app stores cannot charge a user, the subscription is cancelled. This is a recoverable revenue opportunity: notify users by email, push and in-app messaging when a card problem is detected, and help them fix it before cancellation. Recovering half of those users could reduce overall churn by about 10%.
For voluntary churn, engagement tactics help: feedback surveys to learn why users cancel, product changes aimed at the stated reasons, and promo codes to encourage resubscription.
Refunds
A refund is the costliest form of cancellation: the subscriber is lost and the revenue is returned. At minimum, track refunds so a meaningful share of cashflow is not flowing back out. Ideally, the revenue and MRR figures you report already net out refunds.
Stage-by-stage measurement map
Use this map to decide where each metric is captured and what decision it supports. The 'where measured' column reflects the constraint that store-driven events cannot be reliably observed from the device alone.
Worked example: one change, one metric
Suppose an app has 1,000 trial starts in a month and converts at the 33% baseline, producing 330 paid subscribers. If a longer trial lifts conversion to 40%, the same 1,000 starts produce 400 paid subscribers — 70 more subscribers without any extra acquisition spend. Whether that is good depends on the next stage: if the longer trial also raises early churn, the MRR gain shrinks. This is why the funnel should be read as a chain, not as isolated numbers.
The same logic applies to involuntary churn. If roughly 20% of cancellations are involuntary and half of those can be recovered with billing-issue messaging, overall churn falls by about 10% — a change that flows straight into MRR.
Where to go from here
Start collecting these metrics now if you are not already. Retroactive analytics get harder as installs and subscriptions accumulate, so the earlier the instrumentation exists, the more history you will have to compare against.
A funnel gives you a structure for understanding how the business behaves and where improvement is possible. A few stages will usually stand out as the obvious place to start. Experiment in small increments; in a subscription model, small gains compound over time.
When you test a change to one metric, step back and check its effect on the whole business. Trial starts, conversion, renewals and churn all feed the same MRR number, and a gain in one stage can be offset by a loss in another.
| Stage | Metric | Where measured | Decision it supports |
|---|---|---|---|
| Revenue | Gross revenue and proceeds | Server side | Cashflow forecasting after app store fees |
| MRR | Monthly recurring revenue | Server side, based on active subscriptions | Overall business health and trend |
| Install | Installs | Acquisition and attribution sources | Whether top-of-funnel volume is the constraint |
| Virality | Referral-driven installs | Attribution sources | Whether organic sharing is contributing growth |
| Paid acquisition | Acquisition cost vs. user value | Attribution and revenue data | Whether a paid channel is viable at scale |
| Trial started | Trial starts | In-app events (verify SDK accuracy) | Onboarding and upsell effectiveness |
| Trial converted | Trial conversion rate | Server side only | Trial length and premium value fit |
| Renewal | Renewals | Server side | Retention strength over time |
| Cancellation | Churn rate and cohort timing | Server side, cohort analysis | Where in the lifecycle users leave |
| Voluntary vs. involuntary churn | Churn split by cause | Server side plus billing-issue signals | Whether recovery messaging is worth building |
| Refunds | Refund volume | Server side | Whether cashflow is leaking back to users |
The 'where measured' guidance reflects the constraint that store-driven events (renewals, trial conversions, billing failures) occur independently of the device. Figures referenced elsewhere in this guide are dated captured evidence from the original publication and should be re-checked against current store terms and your own data.
Frequently asked questions
Why track MRR instead of just revenue?
Revenue spikes when a longer subscription is paid out in full at purchase, which can hide trends from shorter, cheaper plans. MRR smooths those spikes and gives a steadier read on the size and direction of the subscriber base. It is also useful when an app sells several durations at once.
Which subscription metrics can only be measured on the server?
Trial conversions and everything after them — renewals, cancellations and refunds — need server-side measurement. Trials convert through the app stores automatically, independent of any in-app event a device could report. Revenue events also occur whether or not the user opens the app, so they must be tracked independently of in-app usage.
What is a reasonable trial conversion baseline?
A 33% trial conversion rate is a reasonable starting baseline. Conversion varies with trial length and with how much users depend on premium features, and targeted audiences can convert considerably higher.
What churn range should a mobile subscription app expect?
Mobile app churn is usually substantially higher than in B2B SaaS. On the low end it runs 5% to 10%; on the high end it can exceed 30% per month. Churn is typically highest at the first renewal and falls with each consecutive renewal.
How can involuntary churn be reduced?
Involuntary churn is most often caused by billing failure. Notify users by email, push and in-app messaging when a card problem is detected, and help them resolve it before the subscription is cancelled. Recovering half of affected users could cut overall churn by roughly 10%.
Why do some subscriptions start without a free trial?
App stores allow a free trial to be redeemed only once. A user who cancels a trial and subscribes again goes straight to a paid subscription. A high share of no-trial subscriptions suggests users cancel and return to pay, which is a signal to test longer trial periods.