User Operation | The underlying logic and operation guide of the user churn model!

To do a good job in user operation , the first step is to establish a good user churn model. Only by establishing an accurate user churn model can we better continue the subsequent work.

User operation is like a reservoir with two taps turned on at the same time, where there is inflow and outflow. For any product, user churn is an inevitable phenomenon.

One of the tasks of operations is to accurately predict user churn and retain users with a higher probability of churn through effective operational activities, so that the user churn rate is lower than the user growth rate, which can also ensure upward growth in the number of users.

With quantity, there is a foundation for conversion.

The first step to retain users is to establish a user churn model . Only by establishing an accurate user churn model can subsequent work, such as sorting out user churn nodes and recalling users through various channels, be more effective.

1. The underlying logic of user modeling

First of all, why modeling?

Because there are tens of millions of users using a product, and each person's interests and personalities vary greatly, it is impossible for the product side to provide 1V1 service to everyone.

However, Internet users are very picky. More and more precise push notifications, personalized marketing, and personal privileges are all designed to satisfy everyone's unique tastes. It can be said that in the current Internet environment, precision is the core of whether products and services have the opportunity to connect with users.

Therefore, user modeling is needed, the purpose of which is to clarify user characteristics and achieve the most efficient operation based on the lowest cost and the widest coverage.

Okay, so how do we do it?

There are two entry points for user modeling: user attributes and user behavior.

User attribute characteristics are basic information about users that is difficult to change for the time being, including region, gender, age, education level, social status, etc.

For example, a female college student in a first-tier city and a full-time mother in a fourth-tier city may have very different product usage requirements and information acceptance levels.

The user's behavioral characteristics are even more valuable: Does she like our product? How are you using our products? Is there any obvious preference during use? How often is it used, etc.

With the above two foundations, we can restore the true portrait of this user with a high probability.

User modeling is to separate users with different attributes and behaviors, and then perform differentiated operations based on different goals .

For example: for the activity indicator, the specific application scenarios of user modeling are: launching targeted activity improvement operation strategies for inactive users, launching targeted loyalty enhancement operations for active users, and guiding and driving inactive users' operation strategies.

2. Construction of User Churn Model

When we are modeling user churn, the key point is to classify churned users according to certain attributes or behavioral characteristics, break down the attributes or behavioral characteristics of churned users, and find key indicators for churned users.

It is mainly used in two aspects: recalling lost users and preventing loss of existing active users.

Specific steps:

1. Define churned users

To accurately prevent user churn, the first step is to clearly define churned users. The concept of churned users needs to be defined based on the type, tone, and user portrait of your product.

However, different types of products have different requirements for user activity, so it is impossible to set a unified standard. Here I propose two standards for your reference.

Standard 1: For social products, define lost users by DAU/MAU

Social products have extremely high requirements for user stickiness, so user activity is an important assessment criterion. The DAU/MAU value is a number between 0.03-1. The higher the number, the higher the activity (DAU is the average daily DAU of the month).

If DAU/MAU=1, it means that users come every day, so DAU and MAU are equal, and the lowest line of this value is around 0.03, that is, all users only come one day a month. Users with a value below 0.03 can basically be defined as lost users.

Except for social products that must be used every day, such as WeChat and QQ (WeChat's DAU/MAU ratio has been maintained at around 0.75-0.8 since 2016, and its users have extremely strong stickiness), basically, a DAU/MAU of around 0.3 is considered relatively active, which means that users basically open it once every three days.

Standard 2: For e-commerce products, define lost users based on purchasing activity indicators

The usage scenario of a product determines its basic frequency of use. Not everything has to be used every day to be valuable. At the other end of the spectrum, there are products that are used occasionally but where every interaction is very valuable. For these products, DAU/MAU is not an appropriate metric.

Taobao's activity level is only 0.29, and the average activity level is about three days a week. However, Taobao is an e-commerce application, and it is impossible for users to open and browse it every day. Its purchasing activity level is a more important indicator.

E-commerce apps make profits through user purchases, so churn is usually defined by the level of purchasing activity. If a user only browses but does not buy, the e-commerce company may lose the user.

2. Constructing a user churn model

It is used to build a behavioral model by referring to the behavioral characteristics of users with different frequencies, to break down the behavioral characteristics of lost users, and to find key indicators for lost users.

One of the great conveniences of building a model is that it can clearly show the specific critical value of lost users. We all know that the higher the DAU/MAU value, the better, but what value below is considered lost?

At this time, you can use the chart to judge: when the churn rate reaches a relatively stable trend, it is more reasonable to define the lost users at this time point.

The churn rate of this group of new users reached 40% and reached a stable trend after the 28th day, which proves that the definition of "no visit within 30 days" as churned users is relatively reasonable.

It can also be seen from the figure that the user churn rate is relatively high within two weeks after activation. If you can survive these two weeks, the number of lost users will be greatly reduced.

The next step is to segment the profiles of these churned users, including their behavioral differences from active users, the channels through which they enter the app, the frequency of visits to the app before churn, and their usage behavior in the app (e.g., at which stage they left and churned), so as to infer the reasons for user churn.

For example, by analyzing user behavior, we found that user A visited the app very frequently before churn, 3-5 times a week. However, the pages that he jumped out of the app several times were payment pages. In this case, there is a high possibility that there is a big problem in the payment process.

It may be that payment errors often prompt users, causing them to be annoyed, or it may be that the payment process is complicated, causing users to feel troubled. Bad experience is the main reason for user A’s loss.

Here is another example:

After a product was updated, the user churn rate increased. After analyzing user attributes, it was found that female users accounted for a large proportion of the churned users. This may be because the UI interface of the product was not popular with female users after the revision.

Or if behavioral analysis reveals that a high percentage of newly registered users are lost, that could be because the user guide after the revision was not done well.

3. Find the key points of product retention and recall through various channels

After defining lost users, establishing a user churn model, and finding the reasons for user churn, the next step is to recall users.

Common ones include: SMS, email, site push, WeChat service account, etc.

In this link, the user churn model can also be of great use.

For example: segment by purchase frequency and amount.

For users who have never made a purchase, large coupons, big promotions or super low-priced products can be distributed to attract return visits and become the first new customers.

For users who purchase 1-2 times and have a low average order value, we can accurately push special offers or good products at this average order value level.

For users who purchase three times or more, we can push their preferred brands or categories, and provide additional member-exclusive coupons, etc.

In short, by distinguishing users with different behaviors and attributes, as well as the nodes and reasons for their churn according to the user churn model, operations can be targeted and enhance the effect of user recall.

It is difficult to regain lost users. A more effective approach is: since we already know the characteristics of churned users, when inactive users show the characteristics of churned users, it means that a churn warning has occurred and the corresponding anti-churn strategy needs to be initiated.

The key to user operation work is "targeted". No matter what kind of user model is established, it is necessary to work with the data product team many times based on the characteristics of the product to find a more appropriate way to establish the model. Successful user operations require segmenting users and proposing targeted solutions.

Related reading:

1. User operation: new funnel model for conversion analysis!

2. User operation: How to use B-side operation thinking to increase user growth?

3. Product operation: How to use data analysis to drive product user growth?

4. APP user growth: One model solves 90% of growth problems!

5.How to increase users? Take Pinduoduo and Xiaohongshu as examples

6. Triggering user growth: Is user operation just about attracting new users?

7. User operation: What else can you do to attract new users without fission users?

8. User operation: how can financial products awaken dormant users?

9. User operation: How to make use of private domain traffic?

Author: Fulu Network

Source: Fulu Network