Lifetime value (LTV) shows the revenue generated by a new audience via subscriptions,
in-app payments, in-app impressions, and any custom payments.
LTV prediction allows you to forecast probable revenue, evaluate the ROI, and determine in advance the rate of success for the chosen monetization model.
With such a forecast, you can easily identify inefficient channels and skillfully allocate the budget.
How it works
myTracker makes a forecast using machine learning models.
Its prediction mechanism is flexible and
can adjust based on the current stage of the app's life cycle,
amount of historical data available,
and traffic volume (daily install count).
You can see a forecast the very next day after users install your app,
and watch it become more accurate by the day, and fixed on ninth day.
LTV can be predicted for periods of 30, 60, 90, or 180 days after app installation.
A forecast can show your revenue both as a whole and for the individual revenue type:
- In-app payments. LTV prediction is only possible for verified payments.
- Subscriptions. LTV prediction is only possible for verified payments.
- In-app ads. LTV prediction is only possible for revenue from ironSource ads (ironSource is a monetization partner).
- Custom revenue. Revenue loaded via the S2S API, for example, offline purchases or WeChat payments.
ironSource delay can lag for a few days, which might reduce the forecast accuracy in the first days after users installed your app.
For more details, please see the section on ironSource integration
Features and restrictions
Predictive models show zero LTV for cohorts where users didn’t make any payments
or see any in-app ads in the first eight days after app installation.
The actual, LTV in this case might be greater than zero,
because some users start generating revenue after eight-day mark (which is especially likely for apps with deferred monetization).
The first group models can partially address this issue by using statistical indices, but the forecast accuracy will be reduced.
An LTV will be overstated if temporary discounts or promos were active in the app
when the model was learning,
with this unusual spike in payments leading to
an overly optimistic prediction.
Having increased traffic volume at the time of learning may also cause prediction errors.
It can be a result of being featured in the App Store or Google Play, ad campaigns, PR or other external activities. On the one hand, such an increase may lead to a higher number of installs and a reduced paying audience.
But on the other hand, it may drive up views of in-app ads.
Significant prediction errors might occur if there is a marked change in the monetization scheme, prices, availability of merchandise, etc., as the previous predictions did not reflect it, and the new ones would need time to learn. The re-learning will take from 8 to 180 days, depending on the scope of the changes.
- The accuracy will be reduced if you have disabled the ads for the first few days to retain the audience.
- It is more likely that our predictive models will somewhat understate the LTV rather than overstate it in order to minimize the cost of error.
How to use
- Make sure that the myTracker SDK has been integrated into your app.
- Activate payment verification.
myTracker will check each payment, keep track of subscriptions, and let you exclude fraud and developers’ test transactions from stats.
- If you work with ironSource, connect myTracker account to your ironSource account. Wait for the ad data to roll in.
- In the Constructor, select a report period.
The forecast will be made for the cohort that installed the app during the selected period.
- Add LTV Prediction metrics to your report.
LTV can be forecast for periods of 30, 60, 90, or 180 days after app installation.
One report can show a forecast for different revenue types, a day-by-day chart, and
a summary table for various cohorts, including breakdowns by Partner or Traffic source.
You can get reports with LTV forecast using Export API