How to use statistical tools to evaluate the quality of users in the channel

One. Define the underlying metrics:

Add a new user. This is what we often call activating users. It makes no sense for users to only use it, so when evaluating the quality of channel users, you can't pay attention to the quantity, but also take into account the amount of use, and the user can only be called your real user when you apply it and start using it!

Active users. An active user is a user who has opened the app at least once in the selected time period, and of course, a new user on the same day is an active user on the same day.

Composition of active users.

The report provides an indicator of the composition of active users, which clearly reflects the proportion of new and existing users among active users. Not only can you see the proportion of new users, but you can also get the specific day of the old users among the active users.

Duration of use. That is, the distribution of the amount of time a user launches and uses the app at a time.

Retention. As the name suggests, it refers to "how many users stayed". Here's an example:

If 100 new users are added on September 1, 50 of those 100 people launch the app again on September 2, and 55 users use the app again on September 3, then the next day retention of the new users on September 1 is 50%, and the retention is 55% two days later. User retention and retention rate reflect the quality and ability of your app to retain users.

The basic indicators are introduced here, referring to the statistical analysis of "Zhuge ", I think that user activity, user retention and custom events are generally useful, at least in the early stage of product promotion, these indicators are the most important. Other common indicators such as device model, network type, geographical distribution and other information also have a reference role in judging the quality of channels, so let's talk about how to use these indicators to evaluate the quality of channel users.

Two. Evaluate channel user quality with foundational metrics.

Retention analytics. In app promotion, channel promoters are usually asked to provide next-day and seven-day retention data, which also causes some promoters to only swipe the next-day and seven-day retention rates when they commit fraud. In my work, I encountered a developer's application, the next day retention and seven days retention were maintained to 30%-40%, but once I checked the three-day retention and four-day retention, the data was miserable, and all of them were "0".

Such data is clearly not generated by real users.

Duration of use. Let me teach you another trick, there must be a difference between the data generated by real user behavior and the usage time of fake data. If it is found through the data that there is a significant difference between the usage time of a certain channel and other channels, then the authenticity of the data of this channel is questionable.

Terminal properties. The terminal attributes I am talking about here include device model, network and carrier, and geographic distribution.

The specific statistical function can be directly implemented by using the third-party platform OpenInstall, and the installation registration, active retention, and various behavior data of users can be counted in real time under multiple channels.