Paid acquisition for subscription apps and digital products.
An ecommerce order is revenue on the day. A subscription is a promise that pays out over months. Most of what goes wrong in subscription ad accounts comes from forgetting that difference.
When did your last subscriber actually become profitable?
Most subscription teams cannot answer that from their ad reporting, and the reason is structural. The ad platforms were built around a purchase that happens once, close to the click. A subscription does the opposite. The first charge may arrive a week or a month after the click, the real value arrives over the months after that, and the platform stopped watching long before either.
That gap is where subscription budgets go wrong. Campaigns get judged on installs or trial starts because those are what the platforms can see. Budget flows toward whatever produces the cheapest early event. And months later the business discovers that the cheapest trials were also the ones least likely to pay.
This guide is for consumer and prosumer subscriptions: apps, digital products, memberships, and software a person buys on their own card. It covers the order in which the metrics matter, why trial design is a media decision, what to tell the platforms, how to read results by cohort, and how to know what you can afford to pay.
Why the ecommerce playbook does not fit
In ecommerce, the order is the revenue. A 3.0x ROAS on day one is a reasonable, if imperfect, statement about what the spend earned. In a subscription business, the day-one event is a leading indicator of a future that has not happened yet. A trial start is worth something only in proportion to the chance it becomes a paying subscriber and the length of time that subscriber stays.
Two things follow. First, the platforms cannot see most of the value they are being asked to optimize. Meta’s longest standard click window is seven days. A fourteen-day trial converts to paid after that window has closed, so from Meta’s point of view the trial start is the last thing that ever happened. Second, the true result of a month of spend is not known at the end of that month. It is known over the following quarter, one billing cycle at a time.
None of that is a reason to stop using the platforms’ optimization. It is a reason to be deliberate about which event they optimize toward, what value it carries, and how you judge the outcome later from your own records.
The metrics that matter, in order
Every metric below is useful. The mistake is letting an early one stand in for a later one. Read them in this order, and never let the first two decide a budget on their own.
| Metric | What it tells you | When you know it |
|---|---|---|
| Cost per trial or install | How efficiently the ads buy attention that acts | Same day |
| Trial-to-paid rate | Whether that attention was the right attention | End of the trial |
| Cost per paying subscriber | What a customer actually cost | End of the trial |
| Month-two and month-three retention | Whether the product kept the promise the ad made | One to three billing cycles later |
| Payback months | How long before a subscriber has earned back their cost | Projected early, confirmed over months |
| Twelve-month contribution per subscriber | Whether acquisition is building a business | A year later, estimated in between |
The trial-to-paid rate is the hinge. It converts a cheap number into an honest one. Cost per trial divided by trial-to-paid rate is cost per paying subscriber, and that single division is the most important calculation in subscription acquisition. A $10 trial that converts at 20% and an $18 trial that converts at 40% are not a cheap channel and an expensive one. They are a $50 subscriber and a $45 subscriber.
Trial design is a media decision
Free trial or paid trial, card required or not, seven days or fourteen, monthly or annual by default. These look like product and pricing decisions, and they are. They are also the single largest influence on what your ad account will learn to buy.
A no-card trial produces more trial starts at a lower cost, and the platforms will love it, because their job is to find more of the event you asked for. A card-required trial produces fewer starts at a higher cost and usually a much higher trial-to-paid rate. Which one wins depends entirely on the division above, which is why the choice cannot be made from the ad account alone. The arithmetic, and what to optimize toward in each case, is in free trial or paid: what your ads should optimize toward.
What to tell the platforms
The platforms optimize toward the event you send and the value attached to it. For a subscription, that makes the signal design the most consequential setup work in the account.
- Pick an optimization event early enough that the platform sees it inside its attribution window, and meaningful enough that it predicts payment. Trial start is the usual choice. An activation step, such as completing onboarding, is often better when volume allows.
- Attach an expected value. If trials historically convert at 30% and a paying subscriber is worth about $75 in contribution over the first year, a trial start is worth about $22.50. Segment the value where you can: an annual-plan trial and a monthly-plan trial are not worth the same.
- Send the event from your server as well as the browser or app, with deduplication, so it survives ad blockers, browser limits, and app privacy frameworks. For Meta that is the Conversions API. For Google it is enhanced conversions and offline conversion import.
- Send the later truth back where the platform can use it. Google accepts conversions imported after the fact and adjustments to values already reported. When a trial converts, refunds, or cancels, the value can be restated.
- Keep the definitions stable. Changing the optimization event or its value mid-flight resets what the system has learned and breaks every comparison with the weeks before.
Apps add a layer. On iOS, installs and in-app events reach the ad networks through Apple’s privacy framework, with delays and coarse values rather than user-level detail. Web-to-app flows, where the subscription starts on your site before the download, keep more of the signal in your own hands and are worth considering for exactly that reason.
Where each channel tends to fit
There is no universal channel mix, but the roles are fairly consistent across consumer subscriptions.
- Meta and TikTok create demand. Most people did not know they wanted a meditation app, a language course, or a photo editor until they saw someone use one. These channels live or die on creative that shows the product doing its job, and they need the most careful signal design because their windows are short.
- Google Search harvests demand. Brand search, category searches such as “budget app”, and competitor comparisons are high intent and usually convert to paid at a better rate. Search volume caps how far this can scale, which is why it rarely carries a subscription business alone.
- App store search, including Apple’s own search ads, captures people already in the store looking for a solution. It is close to the install and often efficient, and it is judged on the same trial-to-paid discipline as everything else.
- YouTube and Demand Gen sit between the two, useful when the product needs a demonstration longer than a short video allows.
Creative that sells a subscription
A subscription ad makes a promise the product has to keep every month. Creative that overpromises produces cheap trials and poor retention, which is the worst combination because it looks good for two weeks. The most reliable subscription creative shows the product in use, names the specific job it does, and is honest about what happens when the trial ends.
That last point matters more than it seems. People who were surprised by a charge ask for refunds, dispute payments, and leave reviews. Being clear about price in the ad lowers the trial rate a little and raises the trial-to-paid rate, and it is the second number that pays the bills.
Read by cohort, not by calendar month
A monthly revenue chart mixes subscribers who joined last week with subscribers who joined two years ago. It cannot tell you whether this month’s acquisition was good. The unit that can is the cohort: everyone who started in the same week, followed forward one billing cycle at a time.
Group cohorts by acquisition channel and campaign where your records allow it. Then compare them at the same age. How many of the March Meta cohort were still paying at month three, against the March search cohort, against February? A channel whose cohorts retain well can afford a higher cost per subscriber than one whose cohorts churn early, even when the day-one numbers look identical.
Cohorts also catch the failure that monthly reporting hides best: a campaign that scaled by finding people who try everything and keep nothing. Its trial numbers climb, its month-two retention falls, and the blended monthly chart looks fine until the churn arrives.
Payback is the budget question
CAC payback is the number of months before a subscriber’s cumulative contribution covers what it cost to acquire them. It turns acquisition into a cash question the business can plan around. A company with plenty of runway can accept a long payback. A company funding growth from revenue needs a short one, whatever the eventual lifetime value.
Working it out takes three things: the cost per paying subscriber, the contribution each subscriber brings per month after fees and variable costs, and the retention curve that says how many are still paying at each month. The full method, with a worked table, is in subscription payback: what you can afford per subscriber. The short version is that small improvements to early retention move payback more than most bid changes ever will.
The four ledgers, for a subscription
The same discipline that applies to an ecommerce store applies here, with the billing system in the role of the store. Read these side by side, and never average them.
| Ledger | What it records | What it is good for |
|---|---|---|
| Ad platforms | Installs, trials, and purchases they believe they caused | Comparing campaigns within one platform |
| Product analytics | What people did after they arrived | Activation and early engagement |
| First-party tracking | The click, the session, and the signup on your own domain | Which channel started which subscription |
| Billing system | Who paid, who renewed, who refunded, who canceled | What the business actually earned |
The billing system, whether that is your payment processor or the app stores’ own reporting, is the anchor. Every other ledger describes intent or activity. Only billing describes money. When the platforms report more subscriptions than billing recorded, the gap is overlap between platforms and trials that never converted. When the gap suddenly changes shape, check tracking before you touch the budget. The same reasoning is set out in full for ecommerce in why Meta, Google, GA4, and Shopify report different revenue.
What to ask whoever runs your acquisition
- What is our cost per paying subscriber by channel, not our cost per trial?
- What event are the platforms optimizing toward, and what value does it carry?
- How do the last three months of cohorts compare at month two and month three, by channel?
- What is our payback in months, and what would it take to shorten it by one?
- When platform-reported subscriptions and billing disagree, which one moved?
If those questions get answered with trial counts and platform ROAS, the account is being run as if it were a store. It is not a store. It is a set of promises, and the only honest measure of acquisition is how many of them are kept.
Where to start this month
- Pull cost per paying subscriber by channel for the last three complete months, from billing rather than the ad accounts.
- Write down the event each platform optimizes toward and the value it carries. If nobody can say, that is the first fix.
- Build a simple monthly cohort table by channel, even if it only covers six months.
- Set a payback period with whoever owns the cash, and derive a maximum cost per subscriber from it.
- Change one thing, such as the optimization event or the trial design, and read the result on the cohorts that started after the change.
Should I optimize subscription ads for installs, trials, or purchases?
Optimize for the earliest event that reliably predicts payment and still arrives inside the platform attribution window, usually a trial start or an activation step, and attach an expected value to it. Optimizing for installs alone tends to buy people who never start a trial. Optimizing only for the first payment often leaves the platform with too little signal, too late.
What is a good trial-to-paid conversion rate?
It depends heavily on trial design. Card-required trials convert at much higher rates than no-card trials, and category and price matter too. Rather than chase a benchmark, compare your own rate by channel and campaign, and judge it together with cost per trial through the cost per paying subscriber.
How long should a subscription payback period be?
As long as the business can fund. A company with ample cash can accept a long payback if cohorts retain. A company funding growth from revenue usually needs payback within a few months. Work it out from your own contribution margin and retention curve rather than a rule of thumb.
Why do Meta and my billing system disagree on subscriptions?
Meta counts conversions it can attribute inside its window, including trials and purchases other channels also claim, and it cannot see renewals, refunds, or cancellations. Billing records who actually paid. The gap is normal. A sudden change in the gap usually points to a tracking or event configuration problem rather than a change in demand.
Written by Sam Nouri, founder, adsrunner. If this resonated and you want to apply it to your own account, you can book a strategy call or run a free audit.
How we research, source figures, and handle corrections: editorial policy.