The Yandex analytical service has received a new tool. It is called "LTV and Outflow Predicates". The service predicts revenue from the existing audience and identifies users who are going to leave the application.

Let's look at each of the functions in a little more detail.

LTV Predicates

The predictive LTV model evaluates the user base, and then divides it into cohorts "based on the probability of generating revenue for the application (top 5% of LTV users, top 20% or top 50%)."

Knowing which users can potentially bring the greatest LTV, AppMetrica can start looking for an audience similar in behavior in advertising networks.

It will take 24 hours for the service to evaluate the LTV of a new audience. After that, she will give a forecast for her already. Representatives of Yandex assure that its accuracy can increase to 99% over time.

Outflow predicates

During the analysis, this tool divides the active audience into four cohorts according to the probability of outflow: > 95%, 75-95%, 50-75% and < 50%. The initial analysis lasts three weeks.

Based on the information received, the application developer will be able to focus his efforts to retain the audience on those who are at risk, for example, by launching special offers for them.

It is stated that after the analysis, AppMetrica will be able to determine the character of the user immediately at the time of installation. That is, instantly. In other words, this is another way to evaluate incoming traffic.