Data Operations: How to define “user retention” and “activity”?

1. Active Users

Let’s first define active users: apps that can be detected by statistical platforms - in simple terms, they are living apps.

"Active users" can be said to be the most relevant data indicator for a product, because only when it is alive does it mean that the App is really being used, and only such a status can create value for the company to which the App belongs.

There are three important concepts about active users: daily/monthly active, active/passive active, and new/old users.

1. Daily/Monthly Active Users

The only difference between "daily active users" and "monthly active users" is that the time units used are different. Therefore, this section focuses on daily active users.

To add, the abbreviations for daily active users and monthly active users are DAU and MAU, abbreviated in English as DAU and MAU.

Daily active users refer to apps that can be monitored by the data platform every day, as shown in the following figure:

As you can see, the daily activity of the sample app fluctuated greatly in the first few days, but after a period of decline, it slowly recovered and eventually stabilized.

1. Active/Passive Active

Before explaining active users and passive active users, let’s take an example:

On the first day, suppose 100 people installed and opened the sample app. After the first day, 50 people uninstalled the sample app.

The next day, among the remaining 50 users, 5 turned off their phones, 5 used background cleaning programs to clear the sample app in the background, and the remaining 40 received push notifications from the sample app, of which 30 opened the sample app and used it.

Although you may feel a little dizzy after reading the above text, it is an excellent case for explaining "active users" and "passive active users".

Active users are referred to as main users, which refers to those apps that have been opened.

Therefore, the main activity of the sample App on the first day was 100, and on the second day it was 30. Similarly, these users are also the most valuable.

Passive active users, also known as active users, are: all active users including "active active users" - as long as the App is not killed by the system, it is passively active. Therefore, as long as users can still receive notifications, they are counted as passive active users.

So, the first day of the sample app has 100 passive active users, and the second day has 40 passive active users.

1. New/Old Users

In fact, in the previous case, you also found that new users are also classified as active users.

This is because the new users also opened (used) the sample APP, so they meet the definition of active users. So how do we distinguish between new and old active users?

There is no unified standard for the classification of new and old users. Currently, there are two commonly used classification criteria:

1. New users within 24 hours: Users within 24 hours from the time the sample app is opened for the first time are considered new users.
2. New users on natural days: Based on natural days, all users added on the first day are new users, even users added at 23:59:59 on the first day are counted as new users.

There is no difference between the two categories, and each product is counted according to its own specific circumstances.

Generally speaking, "24-hour new users" is more accurate, but it is relatively difficult to count; on the contrary, "natural daily new users" is easier to count, but the accuracy is compromised.

The last point that needs to be emphasized is that "active users" and "new users" are two relatively independent concepts - even if there are no new users, active users can still exist independently (old users).

This is very different from the "uninstall" mentioned earlier and the "retention" to be discussed next, because uninstallation and retention only have meaning if new users are added.

2. Retention

After talking about "new", "uninstall" and "active", many concepts related to "retention" will be much easier to understand. The reason is: "retention" is inextricably linked to the above three concepts.

Although the word "retention" is easy to understand literally, in actual application, if even one of the above three concepts is not understood, it is very likely that you will be confused in this chapter. So, with this assumption in mind, please read the following explanation carefully.

Retention refers to the number of new users who have not uninstalled within a specified period of time.

Therefore, the relationship between new additions, uninstalls, and retention can be expressed as follows:

Add = Uninstall + Retain

In "Introduction to Macro Data Indicators (Part 1)", it is mentioned that uninstallation is a very intuitive indicator to measure whether a product is good or bad, so it is very meaningful.

However, since the number of users who uninstall is very difficult to count, the true retention is also difficult to count according to the above formula. However, retention is not rarely mentioned just because it is difficult to be truly counted. On the contrary, it may be one of the most frequently appearing data indicators in product data.

The reason is: in actual statistical analysis, we will use "active users" instead of "retained users".

Before explaining the specific reasons, we need to introduce a new concept - silent users, which refers to those apps that have not been uninstalled and cannot send events to the statistics platform.

For example: Apps killed by the phone's cleaning program; Apps in the phone when the phone is turned off...

Therefore, the retention of users can be expressed by the following formula:

Retained users = active users + silent users

At this point, the reason for using active users instead of retained users is very clear: the data statistics platform can only count living users!

Finally, let me summarize the retained users used in actual work in one sentence: Among a group of new users, the apps that can still be counted by the data platform within a specified time range are called retained users.

In the following text description, unless otherwise specified, "retained users" refers to apps that can be counted by the data statistics platform (excluding silent users).

1. Survival rate (retention activity rate)

Usually we use the "retention rate" indicator to indicate the quality of retention. Commonly used retention indicators are: 24-hour retention, next-day retention rate, 7-day retention rate, 15-day retention rate, and 30-day retention rate.

The key points here are: 24-hour retention rate and next-day retention rate

The 24-hour retention rate refers to the percentage of new users that can still be counted by the data statistics platform 24 hours after the first opening of the sample App. For example, if there are 100 new users on the first day, and among these 100 users, 30 can still be counted by the statistics platform 24 hours after installation, then the 24-hour retention rate of this group of new users is 30%.

The next-day retention rate refers to the number of new users on the first day who can still be counted by the data statistics platform on the second day.

For example: there are 100 new users on the first day. If there are only 50 users left after 24:00 on the first day, then the retention rate on the next day is 50%.

If you look at it in general, you might think that there is not much difference between "24-hour retention rate" and "next-day retention rate". But if you think about it carefully, the difference between the two is very big.

Next, let’s continue with the examples:

Assume that for the sample app, there is no new user during the time period of 00:00:00 — 23:59:58 today, and 100 users are added instantly at 23:59:59.

Among this group of users, uninstallation will begin at 00:00:01 tomorrow, and all of them will be uninstalled at 23:59:58 tomorrow. In other words, at 23:59:58 tomorrow, none of today's 100 users will remain.

So, among these 100 users, the next-day retention rate is 100%, while the 24-hour retention rate is 0.

Because the time interval for the next-day retention rate is a natural day, while the time interval for the 24-hour retention rate is a real 24 hours.

As for "weekly retention", "15-day retention" and "30-day retention", their definitions are the same as "next-day retention", except that the time interval has changed from natural days to natural 7 days, natural 15 days, natural 30 days...

1. Retention coefficient

As mentioned in the “Uninstall” section: It is currently very difficult for domestic apps to obtain a core indicator such as “uninstall rate”, so in most cases, relevant practitioners can only judge the quality of a product by “retention rate”.

Here is a diagram of a common product retention rate:

As you can see, the product's retention rate dropped sharply in the first few days, and then there was basically no significant change in the following days. (Of course, the specific reasons will not be explained.)

The area of ​​the blue part can represent the accumulated retained users over these days.

Don’t underestimate the term “cumulative retained users”. This indicator directly determines the life or death of the sample app. Because only when more users stay can the sample app make more money. On the contrary, if the cumulative retained users are fewer, then the company to which the sample app belongs may not be able to survive the Internet's winter for many more days.

Therefore, the sooner you know the cumulative retained users, the more conducive it will be to make timely adjustments and win more living space.

However, as shown in the figure above: if we use the retention rate on the first day, we can predict the cumulative retained users in the next few days or even more than ten days. That is a significant thing for any App, and the retention coefficient can just solve this problem.

The so-called "retention coefficient" refers to: integrating the retention rates of existing apps to obtain a relatively stable integral formula.

Taking the above picture as an example, through Excel's prediction function, we can deduce that the "retention coefficient" of the current sample App is

y = 0.0001×4 – 0.0037×3 + 0.0478×2 – 0.2526x + 0.4826

Therefore, with the retention coefficient, we only need to know the retention rate on the first day to roughly predict the retention of the sample app in the next few days.

1. Retention survival rate

The retention survival rate refers to the proportion of active users among retained users (specifically users who have not uninstalled) among a group of new users. It is mainly used to measure the survival ability of the sample App.

Simply put, every app hopes to be used by users all the time, and it is a blessing even if it can operate secretly in the background. Because as long as you are alive, there is hope!

However, the cruel fact is: now mobile phones have increasingly strict management of apps. Once the sample app is put into the background, the possibility of its survival becomes extremely low! In this case, "How to ensure the survival rate of the sample app?" becomes a very headache.

There are two common keep-alive methods:

1. Improve the survival rate of the sample App itself. Even if the mobile phone has very strict management of background Apps, the sample App itself can resist being killed by the mobile phone system through technical means.
2. Wake up through other active apps. For example, if the user has 3 apps installed on his phone, and App1, App2, and App3 are all killed, if the user opens App1 at this time, App1 will secretly start App2 and App3. In this way, as long as App1 is alive, even if App2 or App3 is killed by the system again, they can be awakened through App1.

As for the pros and cons of the two options, we will not comment on them here. However, the retention survival rate is the key indicator to measure the survival plan. Before giving the formula for the "retention survival rate", let us recall again the relationship between new additions, uninstalls, retention, and activity:

New users = retained users + uninstalled users

Retained users = active users + silent users

The survival rate can be expressed as follows:

Retention survival rate = active / (newly added - uninstalled)

Don't underestimate the ability to survive, because the chances of many applications being opened are very low. Think about your phone. Are there many apps gathering dust all year round, such as calculator?

If an app has no chance of survival after being installed, it is actually no different from being uninstalled. Therefore, for these low-frequency apps, all they can do is try their best to survive. Life is not easy, and so is App!

3. Summary

I have finally concluded the "Product Macro Data Indicators". I originally wanted to streamline this article, but after the revision, I didn't expect that it would have more than 2,000 words. But at least I have said everything that should be said.

For students who have never been exposed to product data analysis, these articles will be tiring to read. Because each article involves many concepts, and all the opinions and statements are just my own opinions, there are great limitations and even errors, but for those who just want to get started, it is enough~.

Source: MING's Great Voyage