Trial-to-paid for a mobile app: the rate that says the paywall promised too much
Trial-to-paid is the share of trials that become a payment. If it falls while trials rise, the paywall promised too much.
Article summary
- Trial-to-paid is the share of started trials that become a payment. The paywall view and the trial rate have a different base.
- In the example in the article, 100 trials and 40 payments make 40%. 140 trials and 42 payments make 30%. The 40 added trials bring 2 payments, or 5%.
- We iterate on the trial rate. We keep a variant only after retained revenue.
- At Ainoha, a client of ours, 9% of downloads start a trial, and 3.2% of those trials move to paid. The 9% is not the trial-to-paid.
- At Parentis, a client of ours, the move from trial to paid takes 34%, also marked +34%, on better trials. The number of trials is not published. Conversion at the price moment takes 18%.
Summarize this article with
Trial-to-paid for a mobile app is the share of trials that become a payment. The trial rate says how many people start. This one says how many of them end up paying.
What is trial-to-paid?
Trial-to-paid measures the move from trial to paid. Take the trials started over a period, count the ones that lead to a payment, and divide. The word "conversion", on its own, does not say what you are talking about: three questions follow one another, and each has its own base.
How many people see the offer: that is the paywall view. How many start a trial: that is the trial rate, calculated on installs or on paywall views. How many of those trials pay: that is trial-to-paid, and its base is the trial already opened, not the install. Write the base next to the percentage, or the three rates get mixed together.
firstapp reads it on subscription apps that are already live, after the trial rate and before revenue. The reading order is on Growth and on the Method page.
A rising trial looks like a win. If those people pay less often, the result is bad for the app. Trial-to-paid is the figure that says so, even before you look at ARPU.
Newsletter
Once a week.
Last issue
We opened 10 apps to see how they sell a subscription.
+534 people read us each week
How do you calculate trial-to-paid?
The formula:
Trial-to-paid = payments from trials / trials started
The denominator is the trial started. Not the install, not the paywall view. A trial counts when the person opens it, not when they have only seen the button. A payment counts when it is collected. If it is refunded, it leaves the numerator: money returned is not a move to paid.
We also wait for the trial to end before we read it. Someone whose trial is still open has not chosen. A three-day trial is therefore read earlier than a two-week trial.
💡 An example, not a measurement. The two rows are the same kind of app. This is not a firstapp figure.
| Trials | Payments | Trial-to-paid | |
|---|---|---|---|
| Control | 100 | 40 | 40% |
| Variant | 140 | 42 | 30% |
Trials go up 40%. Payments, though, move from 40 to 42. Trial-to-paid falls from 40% to 30%.
To see why, look at what the variant added, in the same example. Forty more trials, and only two more payments. Those forty trials pay at 2 out of 40, or 5%, while the first hundred were paying at 40%. The average falls because the extra trials barely pay.
Why trial-to-paid falls when trials rise
The trial rate is the fastest signal, and it is also the easiest to inflate. A longer duration, a larger button, a stronger promise, a price you only show after the trial: more people start.
Those extra people generally do not pay like the first ones. The first ones already had a reason to pay. The next ones opened the trial because it was easy. When it is time to take out the card, the promise does not hold, or the price arrives too late, and they cancel.
The denominator got bigger. The numerator, barely. The rate falls, and it is not a calculation bug: the paywall let in trials it does not convert. It promised too much.
We still iterate on the trial rate, because it answers quickly. We do not keep the variant on that signal. If the one that wins on trials loses on trial-to-paid, the offer promised too much. The reading that settles it next is retained revenue at thirty days, per install. The full order is on the Method page.
Ainoha and Parentis: the base next to the percentage
On Ainoha, a client of ours, there was no paywall. We put one in, in several steps, after onboarding. It is the first one.
9% of downloads open a trial. 3.2% of those trials move to paid. The 9% starts from the download: it is a trial rate. The 3.2% starts from the trial: it is trial-to-paid. Both figures are right, and they do not answer the same question. Putting them in the same "conversion rate" mixes the start and the payment.
On Parentis, a client of ours, the facility and the child's space come before the price. Four screens lead to the offer: the trial, what it includes, cancellation, then €39.99 per year. The trial lasts three days.
The move from trial to paid takes 34%. The case also marks it +34%, and it says where the gain comes from: better trials. The number of trials is not published. We therefore do not reconstruct a before and an after from these two writings.
Conversion at the price moment takes 18%, across the four screens. It is not the same figure. The 18% starts from the people who see the offer. The 34% starts from the trials. Reading the +34% as "more people saw the price" misses the case: the trials that open are better, they are not more numerous.
Ainoha publishes a level, with the base written next to it. Parentis publishes the passage, and says the gain comes from the quality of the trials. We do not put 3.2% and 34% in the same ranking.
When to read trial-to-paid
We read it once the trial is over, on the cohort that has had time to pay or cancel. If we cut in the middle, we count people who have not chosen yet as if they were going to pay.
The test window is written before launch. On the Method page, we cover at least two weeks when the decision is about the move from trial to paid: the week and the weekend do not bring the same people. We stop on the planned date. Cutting on the day it "is significant" manufactures false winners.
Trial length changes the calendar inside that window. Parentis's three-day trial is read earlier than a two-week trial, because the cohort has already chosen. It also changes the rate. The benchmarks by trial length, by model, and by category are in the paywall conversion benchmarks. This article does not paste them again: it says what to do with the figure once you have it.
We then reread by channel and by country. A trial-to-paid that holds overall can fall on the traffic that carries the weight. And we set aside what is not the test: a season, a spend change, a store feature.
What trial-to-paid does not decide
It does not say the money. Trials that pay rarely can still bring in money if the price is high, and bring in nothing if the price is low. The reading that follows is retained revenue at thirty days, per install. The article on ARPU explains why this figure changes depending on whether you divide by an install or by a paying user.
It also does not say that you need a trial. On Mukiz, a client of ours, the three-day trial comes out, along with the daily pass and the monthly plan. As many people subscribe as before. What an iOS subscriber brings in afterward is not a trial-to-paid: it is read in the ARPU article, then in the one on realized LTV.
The mistakes that distort trial-to-paid
Keeping a variant because trials go up. The trial rate is for iterating. Trial-to-paid says whether the offer held through to payment. The two are read together.
Confusing it with paywall conversion. Conversion at the price moment starts from the people who see the offer. Trial-to-paid starts from the trials. On Parentis, a client of ours, 18% and 34% do not measure the same thing.
Reading before the trial ends. Cancellations often arrive in the first days. A rate read too early counts trials that are still open as if they were going to pay.
Changing the denominator from one report to the next. Installed to paid, trial to paid, paywall view to trial: three rates. The word "conversion" is not enough. Write the base next to the percentage.
Ranking two apps on the percentage alone. Ainoha publishes 3.2% of trials that pay, and 9% of downloads that open a trial. Parentis publishes a passage at 34%, also marked +34%. Without the base, the ranking means nothing.
FAQ
What is trial-to-paid for a mobile app?
Trial-to-paid is the share of started trials that become a payment. It says whether the trial held through to the money. It does not say how many people started the trial.
Is a rising trial a win?
It is when the payments follow. It is not when trial-to-paid falls: the extra trials often pay less than the first ones, the denominator grows, and the numerator barely does. We iterate on the trial rate. We keep the variant only after retained revenue.
What is the difference with paywall conversion?
Paywall conversion starts from the people who see the offer. Trial-to-paid starts from trials already opened. On Parentis, a client of ours, conversion at the price moment takes 18%, and the move from trial to paid takes 34%. Both figures are right. They do not answer the same question.
When should you read trial-to-paid?
Once the trial is over, on the cohort that has had time to pay or cancel. The test window lasts at least two weeks when the decision is about this passage, and it is written before launch. A three-day trial is read earlier, inside that window, than a two-week trial.
Does trial-to-paid say whether the app made money?
No. It says whether the trials paid, not how much. The next reading is retained revenue at thirty days, per install. A high rate on a small price can bring in less than a lower rate on a high price.
How do you calculate trial-to-paid?
Divide the payments from trials by the trials started. Trial-to-paid = payments from trials / trials started. A trial counts when it is opened. A refunded payment leaves the count. In the example in the article, not a firstapp measurement, 40 payments on 100 trials make 40%, and 42 payments on 140 trials make 30%. The 40 added trials bring only 2 payments, or 5%.