Break-even CPI is the cost per install at which the net revenue an average install earns you, by a date you choose, exactly repays what you paid for it. To calculate it for a subscription app, multiply your install-to-trial rate by your trial-to-paid rate by the net value of one subscriber up to that date, after store commission, VAT and refunds. The date is not a detail. The same install can be worth nothing at day 3, 24p at day 14 and 86p a year later, so a subscription app does not have one break-even CPI. It has a curve, and the most expensive mistake in paid acquisition is reading the wrong point on it.
Below is the formula, the funnel method for building it, a worked example that runs an annual and a monthly plan through the same funnel, the gross-versus-net trap hidden in most revenue dashboards, and how to turn the result into the cost per trial that Meta actually reports. Every number in the worked example is hypothetical and labelled as such. The published benchmarks come from RevenueCat's State of Subscription Apps 2026, with definitions checked against the report text.
The break-even CPI formula
Start with the definition, because each word in it carries a decision.
Break-even CPI (at horizon H) = net revenue per install earned by H
Funnel method = install-to-trial × trial-to-paid × net value per subscriber by H, plus net revenue from users who pay without a trial
Target CPI = break-even CPI × acquisition efficiency
- Net revenue means what reaches your bank: after the store's commission, after sales tax where the price includes it, and after refunds. Using gross revenue overstates break-even CPI by anything from about a fifth to about 70%, depending on market and store.
- Per install means every install in the cohort, including the ones that never open the app twice. This is why download-to-paid conversion sits at the heart of the calculation.
- Horizon H is the date by which you need the money back. It is a finance decision, set by runway and appetite for growth, not by the ad platform's reporting window.
- Acquisition efficiency is the share of break-even you are willing to pay. An app that wants every pound of media to return £1.25 of net revenue within the horizon uses 0.8. An app deliberately buying growth might use 1.0 at a longer horizon.
The funnel version is the one worth building, because it shows you which term to work on. Revenue per install tells you where you are. The funnel tells you why. Our guide to subscription app CAC works the same arithmetic from the other direction, starting with spend and ending with a cost per subscriber.
Why a subscription app has a break-even curve, not a number
An e-commerce install that buys pays on the day. A subscription install pays in instalments, and the first instalment often arrives after a free trial ends. Adapty's 2026 report finds that around 90% of trial starts happen on the day of install, and RevenueCat finds that 50.6% of all paid conversions that eventually happen do so on install day, with 19.2% arriving in week six or later. Both of those are shares of conversions, not shares of installs.
The consequence for break-even CPI is simple. Any figure calculated before the cohort has finished converting and renewing understates what an install is worth, and any figure calculated from a long horizon overstates what you will have in the bank by the time the media bill is due. RevenueCat's own benchmark shows the effect on revenue per install: the median across all categories is $0.23 at day 14 and $0.34 at day 60, and the report calls a 1.5 times multiplier between the two typical. Health and Fitness moves from $0.48 to $0.66 over the same period, and North American users from $0.38 to $0.55.
So the useful question is never "what is my break-even CPI?". It is "what is my break-even CPI at the horizon I have decided to be judged on?", and the answer is usually lower than people expect.
A worked example: one funnel, two plans
The following app is hypothetical. It is a UK fitness app selling through the App Store with a 7 day free trial, and we run the same funnel through two price plans so the only thing that changes is how the money arrives. The inputs sit inside the ranges RevenueCat and Adapty publish, but they are not client data and are not offered as benchmarks.
| Input (hypothetical) | Value | Note |
|---|---|---|
| Install-to-trial | 12% | Share of installs starting a free trial |
| Trial-to-paid | 35% | Share of trial starts converting at the end of a 7 day trial |
| Paying share of installs | 4.2% | 12% × 35% |
| Annual price (UK, incl. 20% VAT) | £39.99 | £33.33 before VAT |
| Monthly price (UK, incl. 20% VAT) | £9.99 | £8.33 before VAT |
| App Store share, first year | 70% | 85% after a year of paid service |
| Refunds on first payment | 3% | Applied to the first billing period only |
| Annual first renewal | 30% | Share of annual subscribers renewing once |
| Monthly renewals | 55%, 70%, 78%, then 85% | 1st, 2nd, 3rd renewal, then each month after |
For context on those inputs: Adapty reports a 35.0% average trial-to-paid rate for Health and Fitness, and RevenueCat a 37.7% median for the category. RevenueCat puts median first renewal rates between 23% and 40% for annual plans and 53% to 61% for monthly plans, depending on category, and most categories' refund rates between 3% and 4%. The rates you need are the ones for install-to-trial and trial-to-paid from your own paid cohorts, not these.
Step 1: net value of one payment
The annual plan costs £39.99 including VAT, so £33.33 before VAT. Apple pays developers 70% of the price in a subscriber's first year, minus applicable taxes, which leaves £23.33. Knock off 3% for refunds on the first payment and the net first-year value of an annual subscriber is £22.63, out of a price tag of £39.99. The monthly plan at £9.99 nets £5.83 per payment, or £5.65 on the first payment after refunds.
Step 2: paying share of installs
12% of installs start a trial and 35% of those convert, so 4.2% of installs become paying subscribers. We assume, to keep the example clean, that everyone who converts does so when the trial ends on day 7 and that nobody pays without a trial.
Step 3: break-even CPI by horizon
| Horizon | Break-even CPI, annual plan | Break-even CPI, monthly plan |
|---|---|---|
| Day 3 (trial still running) | £0.00 | £0.00 |
| Day 14 | £0.95 | £0.24 |
| Day 60 | £0.95 | £0.37 |
| Day 90 | £0.95 | £0.47 |
| Day 180 | £0.95 | £0.66 |
| Day 365 | £0.95 | £0.84 |
| Day 400 (after the first annual renewal) | £1.31 | £0.86 |
The annual plan's curve is a staircase. Nothing until the trial ends, then 4.2% × £22.63 = £0.95 on day 7, then flat for a year until the 30% who renew pay again at the higher 85% rate, which adds 36p and takes the two-year figure to £1.31. The monthly plan's curve is a slope. It starts at 24p and climbs as each surviving subscriber pays again, reaching 84p by the end of the first year, when about 7% of the original monthly subscribers are still paying. RevenueCat reports 12-month retention of 8% to 9% for monthly plans, depending on paywall model, so the renewal assumptions here are, if anything, slightly cautious.
Three conclusions follow, and they apply well beyond this example.
- At the date most campaigns are judged, break-even CPI is close to zero. Three days after install, with the trial still running, neither plan has earned anything. An account manager reading ROAS at day 3 is reading the part of the curve that tells you nothing.
- The same funnel supports four times the CPI at day 14 on an annual plan as on a monthly one. That is cash timing, not customer quality. We held the paying share constant across both plans, which real apps never manage, because the plan on offer changes who converts. It shows how much of break-even CPI is set by the paywall before a single ad runs.
- On this example, a £1.00 CPI loses money at every horizon inside a year. That is true on both plans, even though a £1.00 install sounds cheap. A CPI figure on its own says nothing about whether you can afford it.
The dashboard trap: gross revenue is not break-even
The fastest way to calculate break-even CPI is to take revenue per install from your subscription dashboard and call it a day. Most teams who do this set their ceiling far too high, because dashboard revenue is usually gross.
RevenueCat's documentation is explicit about its own product. Its revenue chart is net of refunds, and it offers separate views for revenue net of taxes and for proceeds, which removes taxes and store commission as well. Its realised LTV is calculated before store commission, taxes and fees. The State of Subscription Apps report defines revenue per install simply as total revenue divided by total installs and does not say whether commission has been removed, so treat its RPI benchmarks as gross too unless you know otherwise.
Converting gross to what the developer receives depends on market and store:
| Situation | Developer receives | Calculation |
|---|---|---|
| UK price including 20% VAT, App Store, first year | About 58% | 70% of the price before VAT (0.70 ÷ 1.20) |
| Same, App Store, after a year of paid service | About 71% | 85% ÷ 1.20 |
| UK price including 20% VAT, Google Play subscription | About 71% | 85% (after a 10% service fee and a 5% billing fee) ÷ 1.20 |
| Price with no tax included, App Store, first year | 70% | Commission only |
The store rates come from Apple's subscription documentation and Google Play's fee schedule for auto-renewing subscriptions, which in the UK, EEA and US is now a 10% service fee plus a 5% billing fee. Quoting the 10% alone, as many headlines did, makes break-even CPI about 6% too generous. Apple's Small Business Program pays 85% from the first payment, which changes the first row considerably if you qualify. A UK App Store cohort showing £1.00 of dashboard revenue per install in its first year supports a break-even CPI of about 58p, not £1.00.
As a sense check on scale: RevenueCat puts the median app's download-to-paid rate at 2.0% within 35 days and the median realised value per payer after one year at $23, measured by developer headquarters. Multiplying two medians does not give you the median app, so this is an order of magnitude and nothing more, but 2.0% × $23 is $0.46 per install before store fees. For a typical subscription app, year-one break-even CPI is measured in tens of cents, not pounds.
Blended data flatters break-even CPI
The second common error is building the funnel from all installs. Organic and search installs cost nothing and tend to convert better, because the user came looking. In its Apple Ads benchmark report, Adapty measures install-to-paid at 1.92% for Apple Ads traffic against 0.91% for other paid channels. That is a different panel from RevenueCat's and the two figures cannot be subtracted from each other, but the direction is consistent with what we see in accounts: paid social installs convert below blended averages.
Break-even CPI for Meta should therefore be built from Meta cohorts only. If you cannot separate them, because attribution on iOS is partial and delayed, say so in the model and apply a haircut to the conversion rates rather than pretending the blended figure applies. Our explainer on SKAdNetwork postback delays covers why iOS paid cohorts resolve slowly and incompletely.
Turning break-even CPI into the number Meta shows you
Most subscription apps running Meta at any scale optimise for a trial start or a purchase rather than an install, so the cost you manage day to day is rarely CPI. It is cost per trial or cost per subscriber. Break-even CPI converts into both with one division:
Break-even cost per trial = break-even CPI ÷ install-to-trial = trial-to-paid × net value per subscriber
Break-even cost per subscriber = break-even CPI ÷ install-to-paid = net value per subscriber
In the annual-plan example, a £0.95 break-even CPI at a 12% trial rate is a £7.92 break-even cost per trial on a first-year horizon, and £22.63 per subscriber. Apply the 0.8 efficiency factor and the working target is about £6.34 per trial. This is the number to hold the account to, because it is the number the account can see.
Two cautions. First, if you set Meta's cost per result goal, previously called cost cap, Meta treats it as an average target for the optimisation event rather than a ceiling on every result, per Meta's help centre. Setting it at break-even means you should expect some days above break-even. Second, cost per trial only holds as a proxy while trial-to-paid holds. If a new creative doubles trial volume by attracting people who start trials out of curiosity, cost per trial falls and the account gets worse. We cover which metric to act on, and when, in CPI vs CPA vs CAC for subscription apps.
What moves break-even CPI, and who owns each lever
Because the funnel method multiplies its terms, a 10% improvement in any conversion rate raises break-even CPI by 10%. What differs is who can move each term, and how much of the change survives contact with the rest of the funnel.
| Lever | Effect on break-even CPI | Who can move it |
|---|---|---|
| Install-to-trial | Proportional: 10% more trials per install lifts break-even CPI 10% | Creative promise, onboarding, first paywall |
| Trial-to-paid | Proportional | Trial experience, price, quality of the users the ads attract |
| Net price per payment | Proportional only if conversion holds, which it rarely does exactly | Pricing, plan mix, store and tax |
| Renewal rates | Zero inside the first billing period, large beyond it | Product, lifecycle, billing recovery |
| Horizon | Can multiply break-even CPI several times for monthly plans | Finance and runway, not the ad account |
The point most often missed is that creative sits on both sides of the comparison. It sets CPI, and it also sets install-to-trial and trial-to-paid, because an ad decides who installs and what they expect when they arrive. A concept that halves CPI by attracting people the paywall cannot convert has not improved the account. It has moved the break-even line down to meet it. This is why we judge creative on cost per trial and, once cohorts mature, on cost per subscriber, rather than on CPI.
Our Harmony Self Hypnosis case study is a public example of the two sides moving the right way together: new anxiety and stress angles halved CPI to under £2, and revenue rose over 50%. The case study does not publish the conversion rates or a break-even figure, and we will not invent them. What it shows is the only kind of CPI reduction worth celebrating, the kind that turns up in revenue.
When it is rational to pay more than break-even
There are honest reasons to buy installs above break-even for a period, and one dishonest one.
- Learning. New concepts, new markets and new audiences cost more before the account finds the people who convert. That spend is a research budget and should be sized and capped as one. Our guide to the Meta ads testing budget sets out how.
- A longer horizon you can fund. An app with runway and strong renewals can judge itself at month 12 instead of month 1. That is a legitimate choice, provided the renewal rates in the model come from your own cohorts rather than from hope.
- The dishonest one is paying above break-even because the dashboard is gross, the cohort is blended with organic, or the ROAS was read before the trials ended. That is not a strategy. It is a measurement error that has not been caught yet.
The five questions we ask before accepting a CPI target
When a subscription app gives us a target CPI, these are the questions we work through before a campaign goes live. If a number cannot survive them, it is not a target, it is a wish.
- At what horizon is this break-even, and does the business have the cash to wait that long?
- Is the revenue figure net of store commission, VAT and refunds, or is it straight from a dashboard?
- Were the conversion rates measured on paid social cohorts, or on all installs?
- What cost per trial or cost per subscriber does this imply, given that is what the account will optimise towards?
- Which part of the funnel is weakest against its category benchmark, and is anyone other than the media buyer working on it?
The last question matters because break-even CPI is set by the whole funnel. In our comparison of subscription app marketing agencies we split cohort revenue into installs, install-to-paid, revenue per payer and retention for exactly this reason: an acquisition partner can move the first term and influence the second, but break-even CPI rises fastest when someone is also working on the other two.
Frequently asked questions
What is break-even CPI?
Break-even CPI is the cost per install at which the net revenue an average install produces, by a date you choose, exactly repays what you paid for it. Pay less than that and the cohort is profitable by that date. Pay more and it is not. For a subscription app it has to be stated with a time horizon, because the same install is worth very different amounts at day 14, day 60 and day 365.
What is the formula for break-even CPI?
Break-even CPI equals net revenue per install by your chosen horizon. For a trial-based subscription app you can build it from the funnel: install-to-trial rate multiplied by trial-to-paid rate multiplied by the net value of one subscriber by that horizon, plus any revenue from users who pay without a trial. Net means after store commission, sales taxes such as VAT, and refunds.
What is the difference between break-even CPI and allowable CPI?
Break-even CPI is the point where an install exactly pays for itself. Allowable or target CPI is what you are actually prepared to pay, which is break-even CPI multiplied by an acquisition efficiency factor. An app that wants every pound of media to return 1.25 pounds of net revenue within the horizon uses a factor of 0.8, so a £1.00 break-even CPI becomes an £0.80 target.
Can I use revenue per install from RevenueCat as my break-even CPI?
Not directly. RevenueCat's documentation says its revenue and realised LTV figures are net of refunds but before store commission and taxes. For a UK App Store subscriber in their first year, where the price includes 20% VAT and Apple keeps 30%, the developer receives about 58% of the revenue figure. Treat dashboard revenue per install as a gross number and convert it to proceeds before using it as a break-even CPI.
How do I turn break-even CPI into a target for Meta ads?
Most subscription apps optimise Meta for a trial start or a purchase, not an install, so the cost Meta reports is per trial or per subscriber. Divide break-even CPI by your install-to-trial rate to get break-even cost per trial, or by your install-to-paid rate to get break-even cost per subscriber. Meta's cost per result goal, previously called cost cap, manages the average cost of results rather than capping every one.
What is a good CPI for a subscription app?
There is no good CPI independent of the funnel behind it. A £2.00 install is profitable for an app that turns a high share of installs into annual subscribers and ruinous for one that converts at the median. RevenueCat's 2026 data puts median revenue per install at $0.23 by day 14 and $0.34 by day 60, before fees, which is a useful reality check on how low break-even CPI is for a typical app.
The short version
Break-even CPI is net revenue per install by a chosen date, and for a subscription app it is a curve that starts at zero while the trial runs and climbs as subscribers pay and renew. Build it from the funnel so you know which term to fix, net it of commission, VAT and refunds before you trust it, measure it on paid cohorts rather than blended ones, and convert it into the cost per trial the ad account can see. A good CPI does not exist on its own. It exists only against the funnel behind it.
For where break-even CPI sits in the wider growth model, from install to renewal, read our subscription app marketing strategy guide, and for the cost side, the 2026 mobile app CPI benchmarks. If you want a team to build the model with your own cohort data and run Meta acquisition against it, our consumer apps page explains how we work with subscription apps, and if your app is already spending, our pricing is published.