News

Latest and greatest

devtodev Update: Use ML for Accurate LTV Predictions

LTV prediction machine learning analytics

We at devtodev have done almost impossible - we have developed our own accurate Machine Learning System (MLS) that allows us to predict the LTV of client apps with an accuracy of up to 90% just three days after the installation!

More accurate forecasting methods allow you to select only those traffic sources that will pay off, or more accurately compare the project’s LTV before and after any changes.

Why LTV Plays a Big Role in Your Project

LTV is a metric that helps you estimate how much revenue you earn from a user before they stop using your app. There are many ways you can apply it to make your project more profitable. For example, you must know it when buying traffic i.e. when you need to understand whether the cost of acquisition is viable or not.

LTV forecasting machine learning
Screenshot from devtodev demo

The issue with LTV is that you can’t calculate it in advance since you can’t know how a user or a cohort of users will behave in the future. Therefore, this metric can only be predicted. But at the same time, predicting with simple mathematical functions produces high error values. So we at devtodev have solved this issue and started applying Machine Learning (ML) to make LTV predictions more accurate.

LTV forecasting intervals
Screenshot from devtodev demo

How We Apply ML

ML is an artificial intelligence technology that enables computing systems to identify patterns in data and continuously improve the outcomes of those patterns through “learning”.

By using data on payments, custom actions, properties, and user activities, we can predict LTV just three days after the user has installed your app! Then, on the seventh and tenth days, when we have more data on the users, we recalculate the forecast to make it more accurate.

LTV forecasting table data
Screenshot from devtodev demo

How To Make Your Forecast Even More Accurate

You'll get the best results if devtodev has already accumulated enough data on your project, there is a sufficient number of users, custom events, and the payment event are already integrated. If all these conditions are met, then, at the moment, your forecast accuracy will be 80-90%!

You can find the report in the Forecast -> LTV forecasting section at devtodev. Here you can also break LTV data down by traffic source (channel, campaign, channel type [paid, organic]) and by country.

Hurry up to try out new LTV forecasts for your projects! It’s completely FREE!

devtodev DEMO

Have feedback? Then please contact us via email or Facebook