Guide · Subscription apps
Trial length is a trade, not a best practice.
Short trials convert a higher share of fewer trials. Long trials do the opposite. Which is right depends entirely on how quickly your product proves itself, which is a question about the product rather than about marketing.
What actually changes when you change trial length
| Shorter trial (3 days) | Longer trial (7–14 days) | |
|---|---|---|
| Trial starts | Fewer – feels like less to gain | More – lower perceived commitment |
| Trial-to-paid rate | Higher – less time to forget | Lower – more time to drift |
| Cash timing | Faster, better for payback | Slower by the difference |
| Risk | User never reaches value in time | User gets what they needed and leaves |
| Best for | Products with immediate, visible value | Products needing setup or habit |
Both columns can be correct. What is never correct is choosing between them by copying a competitor, because their answer encodes their time to value, not yours.
The question that decides it
How long after first open does a typical user experience the thing they downloaded the app for? Your trial should be slightly longer than that, and no longer.
A photo editor delivers in minutes: three days is generous. A meditation app is selling a habit that takes a week to feel: three days converts people who have not yet felt anything. A B2B-ish productivity tool with data import and configuration may genuinely need fourteen.
You can measure this rather than guess. Look at when your converting users hit the activation event, historically, and set trial length just past the point where most of them get there.
Payment gating changes everything, so test them together
Requiring card details up front raises trial-to-paid sharply and cuts trial starts sharply. It is a bigger lever than trial length and it interacts with it: a short trial with card details and a long trial without are two coherent products. Combinations that mix the properties usually get the worst of both.
Test the pair as a unit – variant A being three days with a card, variant B being seven without – and judge on paying subscribers per thousand installs rather than on either intermediate rate.
Reminders that are not a countdown
The platform sends a renewal warning and there is nothing you can do about it. Your own communication is more useful earlier and aimed elsewhere: a nudge on day one towards whatever the activation moment is, not a note on day six saying you are about to be charged.
The best-performing trial reminders reference something the person has not done yet. The worst remind them that a payment is coming, which is an invitation to cancel.
Questions people actually ask
- Is three days or seven days better?
- Whichever is slightly longer than your product's time to first real value. Shorter trials convert a higher share of a smaller number of trials; longer trials do the reverse. The winner is decided by how fast your app proves itself, not by a category norm.
- What about a fourteen-day trial?
- Rare in consumer subscriptions and usually a mistake. Long trials attract people who intend to use the free period and leave, and the reminder cadence has to work hard to keep anyone engaged to the end. Fourteen days makes sense for tools with a genuine setup burden.
- Should I require payment details up front?
- It raises trial-to-paid substantially and lowers trial starts substantially. The question is which produces more paying subscribers per thousand installs, and that is testable in about three weeks at reasonable volume.
- Do trial reminders help?
- The system reminder before renewal is required and does not need your help. What helps is a reminder that arrives while the trial is still useful, tied to something the person has not done yet – not a countdown to being charged.
- Can I run different trial lengths by source?
- Technically yes with remote configuration, and it is worth trying. High-intent search traffic often converts fine on a short trial where broad social traffic needs longer to understand the product.
See what a few points of trial-to-paid is worth.
It is the input the acquisition model is most sensitive to. The calculator shows exactly how much your bidding ceiling moves when it improves.