Customer acquisition cost for a subscription app is the acquisition spend required to produce one paying subscriber. The formula is spend divided by subscribers, and almost nobody gets it wrong. What people get wrong is which spend, which subscribers, and which window. Four different numbers are all called CAC in normal conversation, they routinely disagree by more than twenty per cent, and the most common way of calculating it produces a figure that gets worse every time you grow. This is how to work out each version, when to use it, and how to stop the calendar from lying to you.
Most CAC explainers are written for software companies with a sales team and a monthly plan. A subscription app has neither. It has an install, a paywall, a free trial that resolves a week or a month later, a platform taking a cut before the money reaches you, and an attribution system that reports some of this in modelled aggregate. Every one of those changes the calculation. The generic formula survives. The generic answer does not.
The four numbers people call CAC
Before calculating anything, decide which of these you actually need. The arguments teams have about whether CAC is £40 or £70 are almost always two people quoting different rows of this table at each other.
| Metric | Numerator | Denominator | What it is for |
|---|---|---|---|
| CPI | Paid media spend | Installs | Media buying efficiency and creative comparison |
| Cost per trial | Paid media spend | Trial starts | The in-account proxy you can actually optimise against day to day |
| Paid CAC, or subscriber acquisition cost | Paid media spend | New paying subscribers from that spend | Channel and campaign decisions, and the number to compare against net revenue |
| Fully loaded CAC | Media, creative production, tooling, attribution, agency fees or salaries | New paying subscribers | Whether the business makes money, and what you show a board |
| Blended CAC | All acquisition cost | All new paying subscribers, paid and organic | Company-level efficiency, useful to investors, dangerous for channel decisions |
The two that matter most are paid CAC and fully loaded CAC. Paid CAC tells you whether a channel is working, because media spend is the cost that moves when you move budget. Fully loaded CAC tells you whether the company is working. Quoting one while deciding with the other is the single most expensive habit in this category.
Why CPI is not CAC, and how far apart they sit
Cost per install and customer acquisition cost are separated by every conversion rate between the install and the payment. For a trial-based subscription app that is two rates:
CAC = CPI ÷ (install-to-trial × trial-to-paid)
Take an app converting 12% of installs into trials and 30% of trials into paying subscribers. The multiplier is 1 ÷ (0.12 × 0.30), which is 27.8. A £2.00 install becomes a £55.56 subscriber. That number is worth sitting with, because it reframes what a change in CPI is actually worth. A fifty pence rise in CPI is not a fifty pence problem. It is a £13.89 problem. The same leverage runs the other way, which is why a cheaper install that damages trial conversion can make you less money at a better CPI.
This is also why comparing your CPI against a published benchmark answers less than it appears to. Our mobile app CPI benchmarks for 2026 are useful for sanity checking a media buy, but two apps at an identical £2.00 CPI can sit £30 apart on CAC purely on funnel differences, and the one with the worse CPI is often the better business. If your instinct is to fix the install cost first, our guide to lowering CPI on mobile apps covers how to do that without quietly breaking the rates underneath it.
The cohort matching error, which costs the most
Here is the mistake nearly every app team makes, including teams that understand everything above. They take last month's Meta spend, divide it by last month's new paying subscribers, and call the result CAC. It is the obvious thing to do. It is also wrong, and it is wrong in a direction that punishes you precisely when you are doing well.
The reason is timing. Adapty's State of In-App Subscriptions 2026, drawn from around 16,000 apps and $3bn in subscription revenue, reports that roughly 90% of trial starts happen on the day of install. So the delay between spending money and learning whether it worked is not spread across the funnel. It sits almost entirely in one place: the gap between the trial starting and the trial converting. In practice, your CAC reporting lag is your trial length. A seven day trial means this week's subscribers were bought last week. A thirty day trial means this month's subscribers were bought last month.
Now watch what that does to a calendar month calculation. Take an app whose spend goes from £20,000 in month one to £40,000 in month two, with the funnel rates above, so month one's cohort eventually produces 360 subscribers and month two's eventually produces 720. If conversions land roughly half a month after the spend, the subscribers recorded during month two are a blend of the two cohorts, around 540 rather than 720.
| Hypothetical month | Spend | Subscribers recorded in month | Calendar CAC | True cohort CAC |
|---|---|---|---|---|
| Month 2, scaling up | £40,000 | 540 | £74.07 | £55.56 |
| Month 3, scaling back down | £20,000 | 540 | £37.04 | £55.56 |
The distortion is 33% in both directions, and it is symmetrical. Calendar month CAC overstates your cost by a third in the month you scale, then understates it by a third in the month you pull back. All three cohorts here have identical economics. Nothing about the business changed. Only the budget line moved.
We have watched this arithmetic end perfectly healthy scaling attempts. A founder raises budgets, sees reported CAC jump by a third four weeks later, concludes the channel has saturated, and cuts back. The following month CAC looks excellent, which confirms the diagnosis. Both readings were artefacts of the denominator. The fix is not clever: define a cohort by install date, wait for the trials in it to resolve, then divide. It costs you a reporting delay equal to your trial length and it is the difference between measuring your business and measuring your budget schedule.
What belongs in the numerator
Paid CAC counts media spend. Fully loaded CAC counts everything it takes to produce a subscriber:
- Media spend across every paid channel.
- Creative production: editors, creators, licensing, AI tooling, usage rights on UGC.
- Subscription infrastructure and attribution: your revenue platform, your MMP, analytics.
- People: agency fees, or the loaded salary cost of in-house acquisition and creative staff.
One structural point worth knowing before you sign anything, and we have an obvious interest in it so treat it accordingly. How your agency charges changes the shape of your fully loaded CAC as you grow. A percentage of spend scales linearly with media, so it never dilutes: it is a permanent proportional tax on the numerator. A flat monthly fee is a fixed cost, so its contribution to CAC falls every time you scale. We publish flat monthly fees rather than a percentage for exactly this reason, though the honest caveat is that a flat fee is worse for you than a percentage at small spend, which is the mirror image of the same arithmetic. Our breakdown of what Meta ads agencies actually charge covers the models in full.
The other half of the equation: net revenue, not gross
CAC only means something next to what a subscriber returns, and the figure it has to be compared against is net of platform commission. A lot of models still assume 30%. For subscriptions that is usually wrong, and wrong by enough to change decisions.
Apple's published terms pay out 70% of the subscription price during a subscriber's first year, rising to 85% once a subscriber accumulates one year of paid service, with 85% applying immediately to App Store Small Business Programme members. One detail inside that is easy to miss and matters if you run long trials: Apple excludes free trials from days of paid service, so a thirty day trial does not bring a subscriber a month closer to the higher rate.
Google Play charges 15% on auto-renewing subscriptions. Since the rollout completed across the EEA, UK and US at the end of June 2026, that 15% is expressed as a 10% service fee plus a 5% billing fee where Google Play Billing is used, with the flat 15% still applying in markets the new structure has not yet reached. The headline number fell. The amount most developers pay on a subscription did not. If you build a model on the 10% without the billing fee, your allowable CAC comes out about 6% too generous.
Platform mix therefore changes your allowable CAC directly, on top of the install cost differences covered in iOS against Android CPI.
A worked example, end to end
All figures below are hypothetical and chosen to be plausible rather than drawn from any client account. A subscription app running one month of Meta spend:
| Step | Working | Result |
|---|---|---|
| Media spend | Monthly Meta budget | £40,000 |
| Installs | £40,000 ÷ £2.00 CPI | 20,000 |
| Trials | 20,000 × 12% | 2,400 |
| Cost per trial | £40,000 ÷ 2,400 | £16.67 |
| Paying subscribers | 2,400 × 30% | 720 |
| Paid CAC | £40,000 ÷ 720 | £55.56 |
| Creative production | Monthly | £6,000 |
| Agency fee | Flat monthly | £3,250 |
| Fully loaded CAC | £49,250 ÷ 720 | £68.40 |
Now put revenue beside it. Say the app sells a £59.99 annual plan on iOS. Apple's first year rate returns £41.99 net. If 45% of those subscribers renew once, the renewal lands at the 85% tier, which is £50.99 net, worth £22.95 per acquired subscriber on average. Two year net revenue per subscriber is therefore £64.94.
Against paid CAC of £55.56, that reads like a working business with a sensible margin. Against fully loaded CAC of £68.40, the same app loses £3.46 per subscriber over two years. Both numbers are correct. They support opposite decisions. This is not a contrived gap either: creative and agency costs are 19% of total acquisition cost here, which is unremarkable for an app at this spend level.
Note also what is doing the damage. Paid CAC is comfortable. The business is underwater because of costs that sit outside the ad account, and no amount of media optimisation addresses that. The lever with real headroom is the 30% trial-to-paid rate: lifting it to 36% produces 864 subscribers from the same spend and drops fully loaded CAC to £57.00, which turns the two year position from a £3.46 loss into a £7.94 profit per subscriber without touching CPI at all.
What CAC should change inside a Meta account
This is where most CAC articles stop, having produced a number with nowhere to go. Three practical consequences.
You cannot bid to a number you learn a month late
If your trial is thirty days, true CAC for today's spend is not knowable until October. That is far too slow to run an account against. The working answer is to operate day to day on cost per trial, which is available within hours, and true up against cohort CAC at whatever interval your trial length allows. Cost per trial is only a valid proxy while trial-to-paid is stable, so the trial-to-paid rate by cohort becomes a number you watch specifically to check whether your proxy still holds.
The optimisation event decides which costs Meta controls
Meta optimises for the event you give it. Optimise for installs and the system will find you cheap installs, leaving the whole 27.8x multiplier unmanaged and free to drift. Optimise deeper, for trials or purchases, and delivery starts accounting for some of that gap, at the cost of a thinner signal. That trade-off deserves its own treatment, and how Advantage+ app campaigns handle it covers the mechanics.
Meta's reported cost per purchase is not your CAC
The two will not reconcile, and chasing agreement wastes weeks. Attribution windows differ from your cohort definition, iOS conversions arrive through a delayed and partly modelled pipeline, and Meta reports gross purchase value while your CAC comparison needs revenue net of commission. Our piece on SKAdNetwork postback delays explains why the iOS timing gap exists. Treat the in-platform number as a directional signal for comparing ads against each other, and your own cohort data as the source of truth for what a subscriber cost.
When CAC is too high, in order of leverage
The worked example makes the priority order clear, and it is close to the reverse of where most teams start.
- Trial-to-paid conversion. The highest leverage number in the model, because it divides into everything. Paywall, trial length, onboarding and the match between what the ad promised and what the paywall sells.
- Install-to-trial conversion. Usually an onboarding and message match problem rather than a media one.
- Net revenue per subscriber. Pricing, plan mix, renewal rate and platform commission tier.
- Creative. Better concepts lower CPI and lift downstream conversion at the same time, which is the only lever that moves two terms of the equation at once.
- Media efficiency. Real, but the smallest of the five for most accounts, and the first place almost everyone looks.
That ordering is why we treat creative as an economics lever rather than a production line. When Steps & Beasts grew revenue 145% and active subscriptions 118%, the work spanned creative testing and onboarding conversion together, because those two sit on opposite sides of the CAC equation and moving only one of them is how apps end up with an excellent CPI and an unprofitable business.
Frequently asked questions
How do you calculate CAC for a subscription app?
Divide the acquisition cost of a cohort by the number of paying subscribers that cohort produced. The important word is cohort. Take the users who installed in a defined window, wait until their trials have resolved, count how many started paying, and divide the spend that bought those installs by that number. Dividing this calendar month's spend by this calendar month's new subscribers mixes users bought at different prices and gives a number that is wrong in a predictable direction.
What is the difference between CPI and CAC?
CPI is what you pay for an install. CAC is what you pay for a paying subscriber. They are separated by every conversion rate between the two, so CAC equals CPI divided by the product of install-to-trial and trial-to-paid. At a 12% install-to-trial rate and a 30% trial-to-paid rate that multiplier is 27.8, which means a 50p rise in CPI is a £13.89 rise in CAC. CPI is a media buying metric. CAC is a business metric. Neither substitutes for the other.
What is subscriber acquisition cost?
Subscriber acquisition cost is the paid media spend required to acquire one paying subscriber, as distinct from one install, one registration or one trial start. For a subscription app it is usually the most useful version of CAC at channel level, because it is the first point in the funnel where money actually arrives. It excludes creative production, tooling and agency or salary costs, which belong in fully loaded CAC.
Should CAC include agency fees and creative production costs?
It depends on the decision you are making. Paid CAC, which counts media spend only, is the right number for comparing channels and campaigns, because those are the costs that move when you move budget. Fully loaded CAC, which adds creative production, subscription tooling, attribution vendors and agency fees or in-house salaries, is the right number for working out whether the business makes money. Most teams need both and get into trouble by quoting one and deciding with the other.
What store commission should I use when working out CAC payback?
Not 30% in most cases. Apple pays out 70% of the subscription price during a subscriber's first year of paid service and 85% after a subscriber accumulates one year, with 85% applying immediately for App Store Small Business Programme members. Free trials are excluded from days of paid service, so a long trial does not move a subscriber toward the higher tier. Google Play charges 15% on auto-renewing subscriptions, and since the rollout completed in the EEA, UK and US in June 2026 that 15% is expressed as a 10% service fee plus a 5% billing fee when Google Play Billing is used.
Why does my CAC look worse in months when I increase spend?
Because calendar-month CAC penalises growth. Paid conversions arrive after the trial resolves, so in a month when you raise budgets the spend is large but a share of the subscribers recorded that month were bought by the previous smaller cohort. Dividing large spend by an understated subscriber count inflates CAC. The same arithmetic runs in reverse when you cut budgets, which makes retreating look efficient. Cohort matching removes the illusion in both directions.
Where this sits in the wider picture
CAC is one term in a larger system. Our pillar on subscription app marketing strategy covers the full path from install to profitable scale and how net revenue per install sets your ceiling, while Day 7 ROAS deals with the other half of the problem this article raises, which is what to do while you wait for a cohort to resolve.
If you have run your own numbers and the conclusion is that trial and paid conversion, not install cost, are what is holding the account back, that is the problem we work on. Our consumer apps page sets out the approach and the case studies show what it has produced. If your app is already spending and the arithmetic above says the ceiling is close, it is worth a conversation.