App growth metrics that matter: CPI, CAC, trial-to-paid, LTV Retention first, then everything else.

the app growth metrics that matter are the ones that tell you whether a new user stays and pays back what you spent to get them: CPI, CAC (blended and by channel), cost per retained user, day 1, 7 and 30 retention, trial-to-paid, LTV and payback. fix retention and activation before you touch acquisition.

most dashboards lead with the cheapest looking number. this page defines each metric the way an operator reads it, shows which one to fix first, and explains why consumer AI apps play by tighter rules than most categories.

CPI: cost per install

CPI is ad spend divided by installs, as reported by the ad platform or your MMP. it is the easiest number to get and the easiest to be fooled by. a low CPI tells you the ad is cheap to act on, not that the people who installed will ever open the app a second time.

use CPI to compare creatives and campaigns inside one platform. do not use it to compare channels, and never use it alone to decide whether growth is healthy. an install that never returns is a cost, not a user.

CAC: blended vs channel

CAC is what it costs to acquire a user. channel CAC is what each ad platform reports for itself. blended CAC is total acquisition spend across every channel, including creative and tools, divided by total new users from your own analytics. make budget decisions on blended, because platform numbers overlap.

add up what Meta, TikTok and Apple Search Ads each claim and you will often count more users than you actually got, because each platform takes credit for the same person. channel CAC is still useful for ranking campaigns within one platform. it just should not set the total budget.

Cost per retained user

cost per retained user is spend divided by the users who actually came back, usually on day 1 or day 7. it is the most honest version of CAC, because it only counts people who got value from the first session. i treat it as the real CAC, and the dashboard number as a hint.

the rule i use: nothing counts as a user unless they came back. this number also shows why onboarding fixes lower CAC without touching bids. if more installs finish onboarding and return, the same spend produces more retained users.

Day 1, day 7 and day 30 retention

retention is the share of a cohort that opens the app again on a given day after install. day 1 tells you whether the first session landed. day 7 tells you whether a habit is forming. day 30 tells you whether you have a product people keep. read them by weekly cohort, never as one average.

cohorts matter because a single average hides change. if last week’s users retain worse than the week before, something broke: a new ad pulling the wrong audience, an onboarding change, a paywall moved earlier. compare each cohort to the ones before it, and to the creative or channel that brought it in.

Trial-to-paid conversion

trial-to-paid is the share of users who start a free trial and then pay when it ends. for subscription apps it is where retention turns into revenue. a weak rate usually means the trial did not show enough value, the paywall came too early, or the ad promised something the product does not do.

track it by acquisition source. users from one channel or creative can start trials eagerly and convert poorly, which makes that channel look good on CAC and bad on revenue. subscription tools like RevenueCat can break this out by cohort.

LTV: lifetime value

LTV is the total revenue a user brings in over their time with the app. early on you cannot know it, so you estimate it from retention curves and revenue per retained user in your first cohorts. treat early LTV as a forecast with wide error bars, and update it every month as cohorts age.

the common mistake is using an optimistic LTV to justify a high CAC. if your LTV assumes users stay a year and your oldest cohort is eight weeks old, the number is a guess. build LTV from what cohorts have actually paid, then add a cautious projection on top.

Payback period

payback is the number of days until the revenue from a new user covers what you spent to acquire them. it turns CAC and LTV into one practical question: how long is your cash tied up in each new user? shorter payback means you can reinvest sooner and scale without running out of money.

a subscription app with strong retention can afford a longer payback. a free app with thin monetisation needs a very short one, or growth burns cash. track payback by cohort and by channel, not as a single company number.

Which number to fix first

fix activation and retention before acquisition. if users do not come back, a lower CPI or CAC just buys a faster, cheaper version of the same drop off. get day 1 and day 7 retention stable across a few weekly cohorts, then trial-to-paid, and only then push acquisition cost down.

the order i use:

1. activation: do new users reach the first meaningful action in their first session?
2. retention: do they come back on day 1 and day 7?
3. monetisation: does trial-to-paid hold across cohorts?
4. acquisition: now lower CAC and scale spend.

acquisition sits last not because it matters less, but because every fix above it makes each ad dollar worth more.

Healthy ranges for consumer AI apps

consumer AI apps have a hard constraint: early users are often worth cents, and every active user costs money in inference. install costs of $3 to $15 will kill a consumer AI app at that stage. in my experience, these apps struggle to survive above roughly $0.50 blended CAC.

that ceiling comes from the economics, not from a benchmark report. when a user is worth cents in the first months, you cannot buy them for dollars and hope LTV catches up, because inference costs keep running while you wait. the way down is creative volume and organic channels feeding paid, not bid tweaks. ZuAI grew from 10K to 2M users at $0.02 CAC in eleven months on that approach, which proves the method works, not that it is a target for your app. the method is in how to lower CAC for a consumer AI app.

for retention, trial-to-paid and LTV, i will not quote ranges here. compare against your own earlier cohorts first, then against your category’s published benchmarks, such as RevenueCat’s annual subscription app report or your MMP’s benchmark data. a category average is a sanity check, not a goal.

How the numbers fit together: a worked example

here is how the metrics connect for one month of a small subscription app. these are illustrative numbers, not from a real account. the point is how each definition changes the story, not the values themselves, so do not read them as benchmarks for your category or your app.

• acquisition spend (media, creative and tools): $5,000
• installs: 12,500
• users back on day 7: 2,500
• trials started: 500
• trials that convert to paid: 150

metric	calculation	result
CPI	5,000 / 12,500	$0.40
cost per day 7 retained user	5,000 / 2,500	$2.00
cost per paying user	5,000 / 150	$33.33

same budget, three very different stories. a $0.40 CPI looks cheap. the $33.33 cost per paying user is the number to hold against LTV, and payback asks how many months of subscription it takes to earn that back.

for how these metrics fit into the whole role, read what an app growth marketer does, and if you are budgeting for help, see what an app growth marketer costs or browse the learn hub.