Realized LTV of a mobile app: reading a test once the money is collected
Realized LTV is the money already collected on a cohort, at a date. A projection can move without a single euro coming in.
Article summary
- Realized LTV is the money already collected on a cohort. Projected LTV changes when you change the assumed lifetime, without a single euro coming in.
- On a firstapp test read in RevenueCat: realized LTV +17.6%, LTV per paying customer +31%, MRR +10.5%. These are gaps, not levels. We do not reconstruct the starting amount.
- The three figures do not answer the same question. You do not keep the largest one.
- At Mukiz, a client of ours, an iOS subscriber brings in three times more after three months. Revenue is up 98% over the same three months, with the team's ads. The ×3 comes from the offer.
- The formula and the market median are in the LTV and CAC article. For most tests, the decision is made on revenue at 30 days per install.
Summarize this article with
A mobile app's realized LTV is the money already collected on a cohort, at a date. A projection says what the user would be worth if the lifetime you assumed held. Realized LTV says what actually came in.
What is realized LTV?
LTV, lifetime value, is a user's value over time. It is written two ways, and they do not say the same thing.
Projected LTV starts from a monthly revenue and a lifetime you assume. You stretch that lifetime, the figure goes up, and not a single euro has come in. Realized LTV starts from the cohort: you add up what it has paid up to the reading date, refunds removed.
The projection formula, and the market median, are in LTV and CAC. This article does not paste them again. It says how to read a test once the money has been collected, after revenue at 30 days per install.
firstapp uses it on subscription apps that are already live, when you can wait. It is the last step on the Method page, after trial-to-paid and revenue at thirty days.
Newsletter
Once a week.
Last issue
We opened 10 apps to see how they sell a subscription.
+534 people read us each week
Why the projection does not decide a test
A projection is useful for forming an idea before you have the history. It does not settle a variant. Two teams can announce two LTVs on the same app, simply because they did not assume the same lifetime.
Realized LTV is not negotiated that way. At the stated date, the account is the account. It can still move later, if people stay and pay again. It does not move because you changed an assumption in a spreadsheet.
So you write the date next to the figure. A realized LTV at three months is not a realized LTV at one year. The name alone is not enough, in the same way the denominator is not enough for ARPU.
How to read the test published on the Method page
The Method page shows a control against a firstapp variant, read in RevenueCat. Three gaps, on the same test.
| Reading | Gap |
|---|---|
| Realized LTV | +17.6% |
| LTV per paying customer | +31% |
| MRR | +10.5% |
These are gaps, not levels. The starting amount is not published, and we do not reconstruct it.
All three lines go up. They do not answer the same question. Realized LTV says what the test collected on its base. LTV per paying customer says what a person who pays brings in. MRR says whether the subscription book follows. +31% is not "more true" than +17.6%. Taking the largest one is choosing the question after you have seen the answer.
The gap between +31% and +17.6% is not closed by inventing the base. The page publishes them side by side, with MRR at +10.5%. We leave them side by side.
Price is judged on value, not on the rate
A price increase can lower conversion and raise revenue. The rate, alone, would say the test lost. The money would say the opposite. On the Method page, the caption of the first price test is this: LTV per customer. The first price test is the tipping point. No level in euros is published on that figure.
So you do not declare a winning price on the trial rate, or on trial-to-paid alone. You declare it on revenue, then on value over time. The pricing frame is in the pricing strategy.
Mukiz: three times, over three months
On Mukiz, a client of ours, the trial, the daily pass, and the monthly plan come out. The annual and the weekly remain. As many people subscribe as before.
After three months, an iOS subscriber brings in three times more. Over the same three months, revenue is up 98%. The case separates the two: the ×3 comes from the offer, and it is the team's ads that brought the people in.
This ×3 is a realized LTV, per iOS subscriber, at three months. It is not a projection, and it is not revenue per install. Reading the +98% alone attributes the gain to the ads. Reading the ×3 alone forgets that the base, alongside, was brought in by the ads. Both figures are true. How not to confuse them with an ARPU is in the article on ARPU.
When to read realized LTV
For most apps, you iterate on the trial rate, and you declare the new baseline after revenue retained at thirty days, per install. Realized LTV is the step after. It serves apps that can wait, and tests whose question is value over time: a price, a plan, an offer that collects late.
You do not hold a test from this week while waiting for twelve months of realized LTV. Thirty days have already said whether the money was there. Further out, you check that it holds. If realized LTV contradicts revenue at 30 days, you do not keep the variant on the grounds that a projection looked good.
You reread by channel and by country. You set aside what is not the test: a season, a spend, a store feature, another price changed alongside.
The mistakes that distort realized LTV
Deciding on the projection. It moves when you change the assumed lifetime. Realized LTV moves when the money comes in. The formula is in the other article. It does not keep a variant.
Taking the prettiest gap from the same test. +31% per paying customer and +17.6% realized LTV are the same test, read in RevenueCat. Keeping +31% and staying silent on the rest is changing the question.
Reading Mukiz's ×3 as revenue per install. It is what an iOS subscriber brings in, at three months. The +98% is revenue, with the team's ads.
Comparing a realized LTV at three months and a realized LTV at one year. The date is part of the figure. Without it, two LTVs do not line up.
FAQ
What is a mobile app's realized LTV?
It is the money already collected on a cohort, at a date, once refunds are removed. Projected LTV, by contrast, multiplies a revenue by an assumed lifetime. Changing that lifetime changes the figure, without a single euro coming in.
Where is the formula, and the benchmark?
In the article LTV and CAC. This one does not paste them again. It says how to read a gap once the money has been collected.
What does the test read in RevenueCat say?
Control against a firstapp variant. Realized LTV +17.6%. LTV per paying customer +31%. MRR follows, +10.5%. These are gaps. The starting amount is not published. You do not pick the largest of the three.
What is the difference from revenue at 30 days per install?
Thirty days decide the test, for most apps. Realized LTV is the reading after, when you can wait. If it contradicts revenue at 30 days, you do not keep the variant on a projection.
Is Mukiz's ×3 a projection?
No. Mukiz is a client of ours: after three months, an iOS subscriber brings in three times more. It is realized, per subscriber, not per install. Revenue is up 98% over the same three months, with the team's ads. The ×3 comes from the offer.
Has a price that lowers conversion lost?
Not if you only look at the rate. A price increase can lower conversion and raise revenue. You judge revenue, then LTV per customer. On the Method page, the first price test is the tipping point. No level in euros is published on that figure.