What is CPA app marketing?
CPA stands for cost per acquisition or cost per action. In app marketing it means you set a price for one event that happens inside the app, and you pay that price each time the event is counted. For a subscription app the useful events are usually a free trial start and a paid subscription, because those are the steps closest to revenue.
The appeal is simple: an install that never opens the paywall costs you nothing. The catch is also simple: someone has to count the event and tie it to the creator, ad or link that caused it. The rest of this guide covers both halves, the price and the counting.
How do CPI, CPA, CPE and revenue share differ?
They differ in which event you pay for, and so in who carries the risk. The further the event sits from the first view, the closer it is to revenue, and the harder it is to count and credit.
| Model | You pay for | Good for | Watch out for |
|---|---|---|---|
| CPM or pay per view | Each 1,000 views | Reach you can count right away | Views are not installs |
| CPI (cost per install) | Each install | Free apps where installs are the goal | Installs that never open the app |
| CPE (cost per engagement) | An in-app step, such as finishing onboarding or a level | Apps where one early step predicts value | The step must be defined and counted the same way by both sides |
| CPA (cost per action) | A signup, trial start, paid subscription or purchase | Subscription apps with a trial or paywall | Delays, refunds, and actions that can't be credited |
| Revenue share | A cut of what the users you bring pay you | Subscription apps with steady renewals | Creators wait months to be paid, and tracking must be exact |
Revenue share has vocal supporters. WinWinKit, which runs affiliate programs for apps, argued in July 2026 that paying per view pays for attention rather than results. Mixed deals are common too: a rate per view for reach, plus a bonus per action. See pay-per-view influencer marketing and pay-per-install influencer marketing for the other two halves, and performance marketing for apps for the overview.
What conversion rates should a subscription app expect in 2026?
Expect a small share of installs to start a trial and a third or so of trials to pay, with wide gaps by category, country and trial length. RevenueCat's 2026 report, built from over 115,000 apps, puts the median D30 download-to-trial rate at 7.1% in North America and 5.0% in Western Europe, and the global median download-to-paid rate at 2.0% by day 35.
| Step | Reported figure | Segment | Source |
|---|---|---|---|
| Install to trial (D30) | 7.1% median | North America | RevenueCat 2026 |
| Install to trial (D30) | 5.0% median | Western Europe | RevenueCat 2026 |
| Install to trial (D30) | 9.1% Business, 6.9% Health & Fitness | By category, median | RevenueCat 2026 |
| Install to trial | 10.9% average | Global, trial funnels | Adapty 2026 |
| Trial to paid | 25.5% / 37.4% / 42.5% median | Trials of 4 days or less / 5-9 days / 17-32 days | RevenueCat 2026 |
| Trial to paid | 25.6% average, 35.0% Health & Fitness | Global / by category | Adapty 2026 |
| Install to paid (D35) | 2.0% median | Global | RevenueCat 2026 |
| Refunds in first billing period | 3-4% for most categories | By category, median | RevenueCat 2026 |
| Revenue per paying user, year 1 | $23 median | Global | RevenueCat 2026 |
Medians and averages are not the same thing, which is part of why the two reports differ. Use these rows to sanity-check your own funnel, not as your price. Your real numbers are in App Store Connect, Play Console and your subscription tool, and they are the only ones that should set a bid.
What do trials and paying users cost on paid ads?
Published cost-per-trial numbers are rare. The clearest 2026 set we found is Adapty's Meta Ads benchmarks, built from 5,000 subscription apps in 45 countries, July to September 2026. We found no published benchmark for what creators charge per trial or per paid subscription.
| What you buy | Reported cost | Source |
|---|---|---|
| A trial, when Meta optimizes for Start Trial | $20.28 | Adapty |
| A paying user, optimizing for Subscribe | $42.77 | Adapty |
| A paying user, optimizing for Start Trial | $57.32 | Adapty |
| A paying user, optimizing for Purchase | $82.23 | Adapty |
| A paying user, all campaigns | $64.93 | Adapty |
Adapty's point is that the cheapest trial is not the cheapest customer: trials bought by optimizing for Subscribe cost more but turned into payers for less. The same applies to creators. A bonus per paid subscription rewards the videos that bring buyers, not just curious tappers. For the wider budget picture, see how much app marketing costs.
How do you set a fair price per trial or paid subscription?
Work backwards from what a paying subscriber is worth to you. Multiply by your trial-to-paid rate to get the value of a trial, then decide what share of that value you are willing to spend to get it. Every number below is an example; replace each with your own.
- 01
Value of a paid subscriber
Use what you actually keep: proceeds after the store's cut and refunds, over a period you can wait for, such as the first year. Example: $40. For reference, the global median a paying user brought in during year one was $23 in RevenueCat's 2026 report.
- 02
Your trial-to-paid rate
From your own data. Example: 40% of trials convert.
- 03
Value of a trial
$40 x 40% = $16. One trial start is worth $16 to you, on average.
- 04
Your spending limit
Choose the share of value you will spend to acquire it. Spending up to half: max $8 per trial, or max $20 per paid subscription. Spend more and you need a longer payback.
- 05
If you pay for both
Cost per paid subscriber = trial bonus / trial-to-paid rate + subscription bonus. A $2 trial bonus and a $15 subscription bonus: $2 / 40% + $15 = $20. That sits right at the limit.
- 06
Add everything else
Any view rate, fees and free tools count too. Judge the result as total spend divided by paid subscribers, not bonus by bonus.
| Trial-to-paid rate | Value of one trial | Max per trial | Max per paid subscription |
|---|---|---|---|
| 25% | $10 | $5 | $20 |
| 35% | $14 | $7 | $20 |
| 45% | $18 | $9 | $20 |
Two rules of thumb follow. Pay more per paid subscription than per trial, because the paid event already includes the trial risk. And if your subscriber is worth less than what paid ads charge per payer (the $42.77-$82.23 range above), ads at the median price would not pay back inside the period you chose.
How are trials and subscriptions counted per creator or campaign?
On iPhone, Apple counts them for you when the user arrives through an App Store Connect campaign link. On Android, Google Play counts installs per link, but trials and payments per link need a few lines of code in the app. In both cases, only people who tap the link are counted.
| Method | What it counts | Limits |
|---|---|---|
| App Store Connect campaign link (iOS) | Each link gets a campaign token; Analytics shows impressions, page views, downloads, usage, sales and subscriptions per token (Apple). The subscription event report has a Campaign column with free trial starts, trials converted to full price and refunds (Apple). No code in the app. | Download within 24 hours of the tap; the most recent link gets credit; a metric shows only once it reaches 5 |
| Google Play link with utm_campaign (Android) | Store listing visitors and acquisitions per campaign (Play Console Help) | No purchases or subscriptions per campaign; small campaigns are grouped under Other |
| Play Install Referrer + subscription tool (Android) | Code reads the referrer once on first launch; it is kept 90 days and doesn't change unless the app is reinstalled (Android Developers). The app writes the code to a custom attribute in a tool like RevenueCat, whose webhooks then carry it with each trial and payment (RevenueCat). | Needs an app update; counts only installs after that version |
| MMP (AppsFlyer, Adjust, Branch) | Adjust lists the Play referrer as deterministic matching on Android (Adjust); in-app events like trials can be sent with it | An extra SDK and contract; on iOS, web-to-app matching without tracking consent is modeled, not exact |
| AdAttributionKit / SKAdNetwork (iOS) | Installs and conversion values from ads signed by registered ad networks (Apple) | Not built for a link in a creator's profile; postbacks arrive at least 24 to 48 hours after the event |
Apple's per-campaign subscription data got much better this year. On March 25, 2026, Apple announced more than 100 new metrics for in-app purchases and offers, cohorts by download source, and two new subscription reports. RevenueCat also says plainly that it is not an attribution network: it only knows where a user came from if the app tells it.
One shortcut is off the table. Matching a link tap to an app user by device signals is fingerprinting, and Apple's privacy rules say apps that do it, or use SDKs that do, may be rejected from the App Store.
What are the catches of paying per trial or subscription?
Four things make CPA harder than it looks: the event happens late, some of it is undone, much of it can't be credited, and creators carry more risk. Plan for each before you set a price.
- Delays. A trial converts only when it ends, days or weeks after the video. Store reports also arrive later than view counts, so action bonuses are paid later than views.
- Cancellations and refunds. In RevenueCat's data, 55.4% of cancellations of 3-day trials happen on day 0, and most categories refund 3-4% of first paid periods (RevenueCat 2026). Decide up front whether a refunded subscription still earns its bonus.
- Searchers aren't credited. Many viewers watch a video, then search the store by name. Neither Apple nor Google ties that install to the creator's link, so counted actions are a floor, not the total.
- Small numbers hide. Apple shows a campaign metric only once it reaches 5, so a small creator's 3 trials may not appear at all.
- Creators may say no. Pure CPA puts all the risk on the creator, including actions they caused but can't be credited for. Many will ask for a base rate, which is why mixed deals are common.
- Last tap wins. If a user taps two creators' links, Apple credits the most recent one (Apple).
How does Markapping price trial and paid subscription bonuses?
Markapping is a performance-based creator marketing platform for mobile apps: app owners pay creators for results, not flat fees, at a rate per 1,000 views on TikTok and Instagram Reels plus optional bonuses per install, trial or paid subscription. It is not live yet. App owners and creators can join separate waitlists now, and it will run in the phone or computer browser with no app to install.
Every campaign pays a rate per 1,000 views, from $1.00, and the rate can only go up while the campaign is live. On top, you can add three optional bonuses, each priced on its own: per install, per trial started and per paid subscription. Views are always paid, so it is not a pay-per-subscription-only deal.
- 01
Each creator gets their own store link
Apple and Google Play count installs per link. On iPhone, Apple also counts trials and paid subscriptions per link through App Store Connect, with no code in your app.
- 02
Android needs a short code
About 20 lines in your app read Google Play's install referrer and pass the creator's code to your subscription tool, so trials and payments can be counted per link on Android too.
- 03
Fund the budget
From $500, paid up front, plus a 10% platform fee. Unspent budget comes back as a refund or credit. Creators keep 95% of what they earn.
- 04
Set the terms
View rate, any bonuses, a 7, 14 or 30-day counting window and caps. Every video gets a fake-view check, a person makes the final call on held videos, and the creator can appeal once.
- 05
Bonuses arrive later
View earnings follow the counting window. Install, trial and subscription bonuses are paid a few days later, when the stores' data arrives.
| Item | Price | Counted | Spent from budget |
|---|---|---|---|
| Views | $1.00 per 1,000 | 150,000 | $150 |
| Trials started (link-counted) | $2 each | 30 | $60 |
| Paid subscriptions (link-counted) | $12 each | 12 | $144 |
| Total | $354 of $500 |
In this example the app owner paid $550 up front ($500 plus the 10% fee), spent $354 of the budget, and gets $146 back as a refund or credit. That is $29.50 of budget per link-counted paid subscriber. Viewers who searched the store instead of tapping a link aren't credited, so the true cost per subscriber is likely lower. You can only reject a video for breaking a written campaign rule, and videos you don't review within 48 hours are approved.
For the full product, see how Markapping works for app owners and pricing. For creator deals in general, see influencer marketing for apps and performance-based UGC.
Questions
What is CPA in app marketing?
Cost per acquisition, or cost per action: you pay a set price each time a chosen in-app event happens, such as a signup, a free trial start, a paid subscription or a purchase. It differs from CPI, where you pay per install, and from pay per view, where you pay per 1,000 views.
How much should I pay per trial?
Take what a paying subscriber is worth to you, multiply by your trial-to-paid rate, then by the share you're willing to spend. Example: $40 x 40% x half = $8 per trial. For context, Meta campaigns optimized for trials bought them for a median $20.28 in Q3 2026 (Adapty).
What is a good trial-to-paid conversion rate in 2026?
RevenueCat's 2026 medians are 25.5% for trials of 4 days or less, 37.4% for 5-9 days and 42.5% for 17-32 days (RevenueCat). Adapty reports a 25.6% global average (Adapty). Compare against apps in your category and country.
Can I count trials from a creator's video on iPhone without an SDK?
Yes, for people who tap the creator's App Store Connect campaign link and download within 24 hours. Apple's subscription event report has a Campaign column with free trial starts, conversions to full price and refunds (Apple). Viewers who search the store instead aren't counted.
Can I count trials per link on Android?
Not from Play Console alone: it reports store visitors and installs per utm_campaign, not purchases (Play Console Help). Your app has to read the Play Install Referrer and pass the code to your subscription tool or an MMP.
Will creators accept pay per subscription only?
Some will, especially affiliates who already sell. Many won't, because they carry all the risk, wait weeks to be paid, and lose credit for viewers who search the store. A base rate per view plus a bonus per trial or subscription is an easier offer.
Is Markapping live?
Not yet. Markapping is opening soon, and app owners and creators can join separate waitlists now. It will run in the browser, for iOS and Android apps, on TikTok and Instagram Reels.
Sources
- RevenueCat: State of Subscription Apps 2026 (Mar 2026)
- Adapty: State of in-app subscriptions 2026
- Adapty: Meta Ads benchmarks for subscription apps (Q3 2026)
- WinWinKit: Future of WinWinKit (Jul 2026)
- Apple Developer: App Store Connect Analytics, campaign links
- Apple Developer: App Store Subscription Event report
- Apple Developer News: New In-App Purchase and subscription data now available in Analytics (Mar 25, 2026)
- Apple Developer: AdAttributionKit, receiving ad attributions and postbacks
- Apple Developer: User privacy and data use
- Play Console Help: Understand and grow your app's user base
- Android Developers: Play Install Referrer Library
- RevenueCat Docs: Customer attributes
- Adjust Help Center: Attribution methods