Ten thousand words of practical information | How to build a user life cycle?

I didn’t originally plan to write this article about the user lifecycle model, thinking that everyone should understand it, but I found out these days:

1) Many people confuse the customer lifecycle model with other models such as AARRR . These two are not the same thing.

2) The introduction of what is the user life cycle and the popularization of some of the concepts, such as what is the growth period;

3) Just like the RFM model in the previous article, I hope that everyone can not only understand it, but also practice it through technical methods.

So, I wrote this article.

1. What is the product life cycle ?

Before introducing the user life cycle, you need to understand the product life cycle. The life cycle of a product is expressed in professional terms as PLC ( product life cycle). We have seen many pictures of product life cycles on the Internet, and most of them show a "sleeping S shape".

According to the above figure, for any product, using time as the reference dimension and the number of users or income level as the measurement standard, we can divide the life cycle of a product into four stages: start-up stage, growth stage, maturity stage, and decline stage. It can be understood as: from product design to cessation of operation of this product, the entire cycle is the product life cycle.

How do you determine which stage of its life cycle a product is currently in?

1. Downloads

Obtain product download data for a certain period of time or history through KuChuan and App Annie , and make judgments based on the trend of historical download volumes (if historical download data is available, it can be fitted with the “sleeping S-shape”). Take the trend chart of the download volume of an app obtained from KuChuan over the past year as an example:

In the past year, the number of users of this app has increased from 20 million to nearly 50 million, and the angle between the tangent line at any point in the trend graph and the horizontal axis of the coordinate system is relatively large, indicating that the product is in a stage of rapid user growth and may be in the growth stage.

2. Baidu Index

Enter the name of the product you want to view in Baidu Index, as well as the name of the competitor that has always been a leader in the industry. By looking at the value of the Baidu Index and comparing it with the Baidu Index of the leading products, you can basically determine the stage the product is in.

Taking the above picture as an example, the average Baidu Index of this product in the past six months is about 2000. Some of the high points of the index are basically user searches brought about by the product's vigorous promotion or by seizing industry hotspots. As a relatively mature product, the Baidu Index should be able to stabilize at 4000-6000. Therefore, it can be basically judged that this product has not yet reached maturity.

Baidu Index is also useful in another way: sometimes, for unknown reasons, the retention rate of your own products jumps sharply, or you want to better control the development of the product. Baidu Index, as a channel for public opinion monitoring, can be of great help. Of course, in addition to the Baidu Index, 360, Sogou , Alibaba , etc. also have corresponding free services.

3. Cumulative downloads

You can use KuChuan to obtain the cumulative download volume of a product, compare it with the total market volume of the industry and the total number of potential users to determine the stage the product is in.

4. Version iteration

Based on the frequency of product version iterations, we can basically determine the stage a product is in. Products in the development stage are more agile and have a faster iteration frequency.

What is the purpose of clarifying which stage of the product life cycle it is in? From an operational perspective, after obtaining the product life cycle, using the following steps for analysis can help determine the core operational strategy and direction for the current and future periods of time. Next:

a. Use mind mapping to sort out product structure;

b. Based on the product structure, sort out the current core business logic of the product;

c. Based on the product structure, sort out possible typical user usage paths;

d. Combine industry, market, product and other information to sort out the product business logic;

e. Based on the common problems that may be encountered in the stage of the product life cycle determined above, combine the above 1/2/3/4 to make operational plans.

This article mainly talks about the user life cycle and only briefly mentions the product life cycle.

2. What is the user life cycle?

The user life cycle, in simple terms, is the entire process from the time the user first comes into contact with the product to the time they leave the product.

According to the above figure, the user's life cycle is divided into: introduction period, growth period, maturity period, dormancy period, and churn period.

The introduction period can be understood as converting potential users in the market traffic into users of our own products. We call this stage the customer acquisition stage. After users enter the product, we need to find ways to make users active and enter the growth and maturity stages. Users in the growth and mature stages are the core users of the product and are the most valuable loyal users. So, we call it the appreciation stage; from the maturity stage, users will enter a dormant and churn state, which we call the retention stage.

We explain the user life cycle, the behavior of users in different life cycles, and the key tasks that operations focus on at different stages:

1) Customer acquisition area: corresponds to the introduction period. The corresponding user behavior is to transform traffic into users. The core task of operation is to attract new customers and promote the activity of new users.

2) Appreciation zone: corresponds to the growth stage and the mature stage. The corresponding user behavior is to be active in the product, contribute to the product, and continue to stay in the product. The core work of the operation is to promote user activity, conversion/payment, and create retention.

3) Churn zone: corresponds to the dormant period and churn period. The corresponding user behavior is to leave the product and stop using it. The core task of the operation is to appease the sunken and lost users or to transfer to new products (if any).

Assuming you already know where in the lifecycle different users of your product are currently, what can you do? This question is similar to the previous introduction to the product life cycle. After we analyzed a lot of content and finally defined the life cycle, it must have a corresponding promoting effect on operations.

We can think about it like this:

Our most desired goal is: "Put our own ideas into users' heads and put users' money into our own pockets." The ideas here can be getting users to accept our product rules, getting users to like our activities, and getting users to be sticky to our products; users' money can be users' time, users' content, users' interactions, and of course users' money. Putting our ideas into users' heads and putting users' money into our pockets are undoubtedly the two most difficult things in the world.

Therefore, if you do these two things well, you have done a good job in user operations . This process can be understood as a user management process. If you want to manage users well, you must pay attention to two dimensions:

1. Every user can be monitored and evaluated;
2. We can influence every user and drive every user.

Therefore, user operation can be understood as "user life cycle management", the essence of which is:

1. How to monitor and evaluate the average single user value of a product;
2. How to drive the improvement of user value?

For question 1, it can be defined through the RFM model or the AARRR model . Generally speaking, different stratification points can be found through data based on user behavior. Next, evaluate its value for different levels of users. This article uses the life cycle model to define user value. In fact, it is to find out which stage of the user's life cycle is in. The following will explain in detail.

Here, we add a more intuitive way to monitor and evaluate user value based on the following two dimensions:

1. Frequency of use
2. Is there direct consumption behavior?

In the first quadrant, users pay directly and use the product frequently. For example, for Ele.me and Mobike , user value is mainly reflected in frequency of use, duration of use, and income.

In the second quadrant, users pay directly and use the product with low frequency. For example, for the travel product Ctrip , users may not open it every day, but when they have a need, they will pay directly. User value is mainly reflected in income.

In the third quadrant , users do not pay directly and the frequency of use is low. Some people say that products in this quadrant have no value and meaning. Sometimes I feel that if I were asked to operate such a product, it would be really terrible.

But as an operator, we cannot give up first. I think the user value of this quadrant is reflected in income, which is different from the second quadrant. Taking Fangduoduo as an example, it is difficult for users to buy a million-level house through an App without viewing the house. Therefore, rebates, discounts, real estate agent ratings, etc. all affect whether customers can pay through long-term guidance when they have needs.

In the fourth quadrant , users do not pay directly but have a high frequency of use, such as Maimai and Toutiao . The main value of users is reflected in the frequency of visits and usage time. Toutiao previously announced that the average user usage time reached more than 60 minutes, which is user value for them.

Insert some off-topic remarks:

People often ask me why the logic of the article should be like this at this point? For example, as mentioned above, why is it necessary to monitor and evaluate user value through frequency of use and whether there is direct consumption behavior?

Friends who ask such questions, I believe you often visit various websites and have read countless useful articles, but you are just reading them. When you think about it, you have basically forgotten them. Everyone still lacks thinking. Here, I would like to recommend to you the RIA rules used in the field of book analysis. This method is also applicable to article reading.

First, you read the article, then you think about it, and finally, you combine the content of the article with your own experience. If you complete the above steps, it means that you may find that many articles are not necessarily written correctly. If you think that all the articles you have read are good, it would be very dangerous.

Let’s go back to the two dimensions of monitoring and evaluating user value mentioned above. For any product, no matter whether it is high-frequency or low-frequency, whether it is in the start-up stage or the mature stage, no matter whether the business model is cash flow or traffic, its core indicator must be that users want to open the App. If they don’t open it, they cannot be considered as its own users. Its core indicator must be whether they consume. This consumption can be direct payment, product diversion and conversion, or even advertising click revenue. This is also the premise for a commercial society and a commercial company to be able to support a group of people.

Based on the frequency of visits and whether there is direct consumption behavior, you can combine your actual business situation to find out the indicators that you care more about. For example, the frequency of visits is a primary dimension, and you are more concerned about secondary indicators such as visit duration and number of video views.

User value can not only be monitored and evaluated roughly, but the value of each user can even be calculated. However, in actual work, there is not much work involving this module, and when there is such a demand, it is mostly completed by colleagues such as data and development. Therefore, our article is based on the operational perspective, and this is enough.

For question 2, there are two directions to drive user value improvement:

1. Improve the value of individual users;
2. Extend user life cycle.

Ultimately, our research on the user lifecycle model evolved into the essence of user lifecycle management: how to increase the value of individual users and extend the user lifecycle. Of course, essential to this process is data-driven , guidance on typical user usage paths, and refined operations.

There are two issues that need to be emphasized:

1) Do all users go through the full user lifecycle?

no. From the perspective of the entire product, because users are distributed at different stages, they may go through a complete life cycle, that is, the introduction period, growth period, maturity period, dormancy period, and churn period mentioned above, for example: registration-login-active-payment-sharing-churn.

However, as individual users, they may be lost after the introduction period. For example, after registering and logging in, due to inadequate onboarding, users simply give up using the app and enter the loss period.

2) Do all products need to manage the user lifecycle?

no. From the perspective of the product life cycle, for products in the start-up phase, due to insufficient resources and insufficient user scale, user life cycle management is generally not required. Judging from the intensity of product demand and market supply, the stronger the demand and the scarcer the supply, the less you need to consider user lifecycle management.

For example, 12306. No matter you buy train tickets on Ctrip, Qunar or Fliggy, they are ultimately purchased through 12306. The market supply is tight and the demand is high. Users have no other choice but to choose other means of transportation . Therefore, such products do not require user life cycle management. For example, there are airlines such as China Southern Airlines, Air China and China Eastern Airlines. The prices, services, meals, discounts, activities, etc. of these airlines are likely to affect users' choices. Supply exceeds demand, so such products generally require user life cycle management.

3. How to build a user life cycle model

Through the foreshadowing above, we have learned that the essence of building a user life cycle model is to find out which stage of the user's life cycle is in, whether it is the introduction stage or the mature stage, based on the user's behavior. Knowing the stage the user is in will facilitate our further operations.

How to build a user life cycle model? The general steps for building a user life cycle model are as follows:

1. Business logic sorting;
2. Find the key features that affect user retention /consumption;
3. Define user behavior at each stage.

Let’s take Maimai as an example:

a. Maimai’s core business logic is as follows:

b. Key features that may affect user retention/consumption

According to the core business logic diagram, for the user side, it is divided into producers and consumers. From the perspective of consumers, the key functions that affect consumer retention and consumption may be: submitting resumes, consuming content, establishing social relationships, and purchasing memberships.

c. Define user behavior at each stage

Let’s first look at the template for user behavior definitions commonly used in the industry:

Based on the above template and the key driving functions that we defined that affect user retention/consumption, we can make the following definitions:

4. How to improve the value of individual users around the user life cycle?

To improve the value of individual users, you can think about which users’ value needs to be improved? Is it necessary during the introduction period? I don't think it is necessary, as this stage is in the customer acquisition zone; is it necessary during the customer loss period? I don't think it is necessary. The main purpose at this stage is to recall users and extend their life cycle. Therefore, we mainly focus on improving the individual value of users in the growth and mature stages, which is reflected in user behaviors: activity, conversion, payment, and retention.

Through the previous part, we have defined users based on their behavior. Taking Maimai as an example, we define: users who follow XX industry contacts within a certain period of time are active users in the growth stage. So, what is XX? For example, when we find through data collection that users who follow connections in 10 industries are in the growth stage, we can then guide users who have not reached the growth stage to pay attention to connections.

Therefore, we need to analyze according to the following steps to find the user-defined data nodes and complete user guidance by formulating operation strategies at the corresponding nodes:

1. Sorting out user behavior paths;

2. Data definition/data collection;

3. Find the focus through data;

4. Complete user guidance.

1. User behavior path analysis

Let’s still take Maimai as an example to sort out the typical user behavior paths of Maimai. Assume that in Maimai, there are two typical usage paths for a user from the introduction stage to the mature stage:

a. User registration → Improve personal homepage → Upload resume/improve online resume → Search for positions → Submit resume → Add recruiter as friend → Purchase membership → Add more connections

b. User registration → Improve personal homepage → Add industry contacts → View updates → Publish updates → Purchase membership → Add more contacts

2. Data definition/data collection

What data do we need? Generally, to solve the problem of increasing the value of users from stage A to stage B (of course, the value of users in stage B must be higher than that in stage A), we need to first filter out users in stage B, and then obtain three types of data from users in stage B:

a. Data related to typical user usage paths , providing data requirements around the typical usage paths of users from stage A to stage B;

b. Basic user data , such as the user's gender, occupation, age, region, hobbies, family status, etc.;

c. User behavior data : determine which modules within the product may affect the user's subsequent behavior, and then pull out the user data who has used the module. For example, after registering on Weibo, if you immediately follow 5 other users, your activity and retention will be better.

In addition to obtaining the above three types of data from users, you can also refer to user channel source data, business data, etc. for data analysis .

3. Find the focus of operations through data analysis

Several key ideas for data analysis:

a. Which path is better for users to go from stage A to stage B?

b. From stage A to stage B, what characteristics do most users meet?

c. From stage A to stage B, did most users engage in some of the same behaviors?

d. From stage A to stage B, is it affected by different channel sources?

We combine several key ideas of data definition/data collection and data analysis. Taking Maimai as an example, we still assume that:

a. Through the data analysis of 100,000 Maimai users, we found that 20,000 users completed Path 1 and 80,000 users completed Path 2. Therefore, we can preliminarily guess that Path 2 is more optimal for users to move from Stage A to Stage B.

b. We looked at what characteristics most users have and analyzed basic data. We found that among mature users, the male to female ratio is 7:3, with 70% of men aged 22-30 and 75% working in the Internet industry.

c. We found that for users who have submitted resumes three times, the membership purchase rate is 89%, and the average purchase rate of on-site membership is 21%; for users who have posted 10 messages, the proportion of purchasing other user services is 70%, and the average purchase rate of on-site services is 11%.

4. Complete user onboarding

Based on the above-mentioned potential focus points found through the data, we design corresponding strategies to guide users. It is necessary to emphasize the difference between strategy and means: if it is a set of mechanisms and rules, it is a strategy; if it is a one-time push to users on the spur of the moment, it is not. When formulating strategies, many partners write a lot of activity plans and contact methods, which are actually specific and feasible means supported by the strategies.

Take the above-mentioned force points as an example:

a. Path 2 is better. Our core strategy is to strengthen the guidance of users at each node of Path 2. The specific means of guidance can be to recommend more accurate users and high-quality dynamics, etc.;

b. For those who have submitted resumes three times, the membership purchase rate is higher and it is easier to reach the mature stage. Our core strategy has two aspects. On the one hand, we will vigorously introduce B-end recruitment, and on the other hand, we will try to make more accurate recommendations to C-end job seekers . Specific means can be to increase B-end entry benefits, provide 1V1 services to job seekers, etc.;

c . There are many male Internet practitioners. Our core strategy is to strengthen the acquisition and operation of users with similar attributes, which will make it easier to improve the overall revenue of the product.

(The above analysis of Maimai is all based on assumptions, especially the numerical part. It is mainly to illustrate the practical operation of the user life cycle model. If there is anything wrong, please correct me in time.)

5. How to establish a good user churn warning mechanism to extend the user life cycle?

As mentioned above, there are two directions to drive user value improvement:

1. Improve the value of individual users;

2. Extend the user life cycle.

The methods for improving the value of individual users around the user life cycle have been explained in detail. The next step is to focus on how to extend the user life cycle so that users stay in the product as long as possible and contribute value to the product.

How can we extend the user life cycle ?

1. Prevent user churn and keep users in the product;
2. Recall lost users.

When it comes to prevention, people usually think of mechanism and automation. When the system determines that a user has churn characteristics, it will automatically reach out to the user in a timely manner and retain the user as much as possible. Therefore, an automated mechanism needs to be designed here.

An automated contact system can also be designed for lost users, but the actual situation is that the number of lost users is huge, the reasons for loss are unknown, and some users may even lose users right after registration, making them of low value. In addition, the cost of contact methods such as SMS and phone calls is high. Lost users are generally screened manually and then contacted manually. I believe many partners have done this.

Therefore, below we will mainly discuss the prevention of user churn and ultimately design a user churn early warning mechanism.

How to design a churn warning mechanism?

The design steps of the churn warning mechanism are as follows:

• Lost User Definition
• Analyze the signs of churn
• Establish an early warning mechanism
• Complete user onboarding

1. Definition of lost users

Only when we define what kind of users are churned users can we make corresponding churn warnings for non-churned users. When non-churned users exhibit churned user-related behaviors, the system can respond in a timely manner to avoid user churn. To define whether a user has churned, we generally evaluate from two dimensions: user behavior and time.

User Conduct

In short, based on the user's behavior, we consider him to have churned. The user's behavior within the product will definitely rely on the product's functions and services, so user behavior should be comprehensively evaluated based on the basic functions or core functions of a product. Let me give you some product examples:

• The core function of financial products is financial management. Users invest their funds into the target. When users stop this behavior, we speculate that users may be about to leave or have already left. There is also a situation where users come to the App every day to check which financial products have higher yields, but exit after checking and do not invest because there are no targets that meet the user's yield expectations. We call these users "sleeping users" or "dormant users." Users have not left or uninstalled the App, but if they are not retained in time, or if they find other financial products with higher returns, they will become lost users.
• Tool products, such as NetEase Mail, have the core function of sending and receiving emails. However, we may not need to view the user's operations on the core functions to judge the behavior of churned users, because both receiving and sending emails require logging into the mailbox. We can judge user churn by analyzing the user's login behavior.
• For e-commerce products, such as NetEase Yanxuan , after users log in to their accounts, they do not need to log in again every time they open the app. Users may take a look at the recommended items on the homepage and close the app without making any purchases. We can determine user churn based on their payment behavior.
• For content products, such as 36kr , users will read articles in the entrepreneurship section every morning. One day, they suddenly stop visiting. In this case, we need to determine user churn based on their access behavior to 36kr content.

Therefore, the key actions that generally indicate churn behavior are login, visit, and payment. The specific definition of churn behavior can be defined based on the actual situation of your own products.

time

After we select XX behavior as the key churn action for lost users, we need to define the time when this action occurs. Taking NetEase Yanxuan as an example, if a user does not make any payment, does it mean that the user has been lost? This is obviously unscientific. We need to determine how long it takes for a user to not make his or her first or second payment before he or she is considered a lost user.

Here, we need to introduce a concept: returning users .

The Returning Visitor defined in GA ( Google Analytics) means that when a user visits for the first time, an independent Client ID will be generated. When the user visits again, GA detects a new session with the existing Client ID, which is called a Returning Visitor. The returning users of GA correspond to the new users.

The "returning users" we mentioned today refer to users who visit again after losing users.

For example, after analyzing a large amount of data, you find that for a social product, users who have not opened the app within 10 days are defined as lost users. There was a user who hadn't opened the App for 15 days. We thought he had lost the app, but our operations were very strong, and we used various musical instruments and singing skills to invite the user back. This user is what we call a returning user today, that is, a user who comes back after losing the app.

We have a general definition for user churn time: when the user return rate is in the range of 5%-10%, we consider that the user has churned at the time point corresponding to this range and the time thereafter. I found a picture on the Internet as follows:

User return rate = return users / lost users 100%

At the inflection point, the user revisit rate is 5%. The churn period corresponding to this inflection point is 5 weeks. After 5 weeks, the user revisit rate is lower than 5%. Therefore, we believe that the user churn time of this product can be selected as 5 weeks after the user stops performing XX behavior, and the user is considered a churned user.

Based on user behavior and time dimensions, we define lost users as follows:

The above figure shows the relationship between user revisit rate and churn days. We can see that when the user churns for 10 days, the revisit rate drops below 5%. From this, we can define the churned users within the product and clearly identify which users are churned users.

Note: Returning users are not just users who come back to open the App. They should be judged based on the specific product attributes. For financial products, returning users refer to investments; for tool products, returning users refer to logins; for social products, returning users refer to interactions; and for content products, returning users refer to visits.

2. Analyze the signs of loss

The core of this article is to design a churn warning mechanism, that is, to provide churn warning for non-churned users to avoid user churn. Therefore, after we know which users have been lost, we need to conduct research and analysis on them to find out the common points before they lost. If current active users show similar signs of lost users in the future, the loss warning mechanism needs to be triggered.

The steps to analyze the signs of loss are as follows:

• Identify churned users and analyze their behavior before churn
• Hypothesized factors affecting user churn
• Through interviews, clarify the behavioral path of user churn
• To analyze the behavior of lost users before they churn, data analysis can generally be conducted from the following directions:
• What similar behaviors occurred before the user churned?
• Are users concentrated in a certain channel?
• Are user attributes consistent?
• Before the loss, have there been any changes in products, operations, or markets?

Note: To analyze user behavior before churn, we only need to analyze data from the user introduction period, growth period, and maturity period.

We take Pinduoduo, which is doing relatively well in attracting new users, as an example and make some assumptions based on the data to analyze the behavior of lost users before they churned. We take Pinduoduo's growth-stage users as an example:

Pinduoduo's users in the growth stage account for 30%.

The proportion of loss in each channel during the growth stage:

Pinduoduo has always been known for its good fission effect, but the disadvantage of fission inviting friends is that friends register to use Pinduoduo in order to cooperate with you to open red envelopes, and their loyalty may be low. Therefore, in terms of churn, the churn rate of users brought by friend invitations through this channel reached 30%, followed by users brought by soft-text delivery on WeChat public accounts , with a churn rate of 25%.

Among users in the growth stage, only 10% have made three purchases, and 90% have made only 1-2 purchases. This means they have just entered the growth stage and are lost without further conversion to the mature stage.

Among users in the growth stage, only 5% have purchase amounts reaching 200 yuan, and 95% have purchase amounts ranging from 1 to 200 yuan.

As Pinduoduo's strongest welfare module, inviting friends can bring obvious benefits. However, we found that among users in the growth stage, only 30% of users actively initiate the behavior of inviting friends, and only once, and 70% of users have never initiated any invitation behavior.

Hypothesized factors affecting user churn

Based on the above data chart analysis, we assume that the factors affecting user churn are:

• The user churn rate brought by the friend invitation channel is as high as 30%, indicating that user loyalty is not enough and it is easy to churn

• 90% of users only purchase once or twice, indicating that they have not developed a shopping habit

• 95% of users purchase less than 200 yuan, which may be due to weak demand for products or low trust in the platform.

• 70% of users have not initiated invitations, which may be due to insufficient activity guidance or incentives.

Interviews to identify user churn paths

Find the churned users, conduct direct interviews, determine the user churn path, and clarify the hypothesis content. The interview part belongs to the work of user research, but we can still understand some working methods. If it is convenient to carry out work quickly in some companies with insufficient resource support, here is a brief description of the steps of user interview:

• Collect internal and external information of the product to check whether the external environment has changed. Whether the product has been released, whether any operational activities have been launched, whether there is any negative news, and whether industry supervision has changed. For example, the business development of P2P products is closely related to the strictness of industry supervision.

• Develop a plan. Clarify what the interview objectives are, who the interviewees are, and how the interview process should be designed.

• The plan was implemented. Users can be invited to the company for interviews, information can be collected through online questionnaires, and the churn paths can be analyzed through phone calls or even competitive product analysis .

Someone asked me, if I can analyze the behavior of users before they churn, is it okay without making assumptions or conducting user research? It feels so complicated after adding these two steps. In fact, we can still conduct agile testing according to the MVP principle and solve the analyzed problems in a small-scale, manual manner. If it is verified to be correct, it can be gradually systematized.

3. Design an early warning mechanism

Based on the above understanding of what constitutes a lost user and their behavior before they lost, we can start designing an early warning mechanism. The early warning mechanism takes different forms in different companies, and is mainly designed based on the scale and type of business.

• If it is a startup company or a company with rapidly changing business, you may not need to design an early warning mechanism at all. You only need to provide user churn warning. Through data analysis, the operation partner defines labels for lost users and gives them to the data analyst. The data uses SQL to help us generate a list of these users. We can then simply reach out to the users through channels such as push and SMS.

• If the business has reached a certain scale or is mature and stable, in addition to churn warning, it is best to build an automated warning mechanism. Taking WeChat Reading as an example, there are dozens of tags for user gender, age, hobbies, reading habits, reading progress, etc. If the operator manually defines the tags and gives them to data analysts to export the user list, the workload is too large, the efficiency is extremely low, and it is easy to crash (just kidding...)

• According to the behavioral indicators of users before churn, corresponding weights are assigned to these indicators respectively, and calculations are performed based on these indicators and weights. When the value is higher than XX value, the system defines that the user has a churn risk, and it is either automatically triggered or the list is transferred to the operation for manual recall. Of course, if adding weights is more complicated, you can also make warnings directly based on defined user behavior indicators. Everything can be flexible.

• It is best to implement early warnings in different levels. Taking Pinduoduo as an example, among users in the growth stage, 90% of users only shop 1-2 times before churn, and the churn time is assumed to be after 20 days. Then, if the user's shopping frequency remains at 1-2 times after 10 days, we can make more precise promotion for the user based on the user's purchasing or browsing behavior; if the shopping frequency remains at 1-2 times after 12 days, we can issue discount coupons or product cash back to the user based on the recommendation; if the user has only purchased 1-2 times after 15 days, we can call the manual customer service to recall the user and inform the user that there is an XX yuan coupon in the account, which is valid for 2 days and can be used; if it is still 1-2 times after 18 days, we can increase the frequency of push and SMS to the user and add inspiring copy . If there are still only 1-2 visits after more than 20 days, we believe that the user has been lost and can conduct daily recall according to the standards for lost users.

• The churn warning mechanism that I know of can be presented in two forms:

.

A data collection system, as mentioned above, the system will export a list every day, which will be customized by the operation to reach users.

An automatic reach system is the best, as it can greatly improve efficiency. Only when the hierarchical reach is impossible to recall the users will the list be exported to the operation department for phone calls, user follow-up surveys, etc.

4. Complete user onboarding

By the time the above three steps are completed, a loss warning mechanism has been established, which means that user guidance has been completed. However, formulating a strategy for user guidance is a very important step, so we will discuss it separately.

Since it is an early warning mechanism, the core is how to retain users. Therefore, it is not enough to simply analyze who the lost users are, what behaviors the lost users have before losing, and reach out to them before losing them. How to reach? What is the content reached? What are the channels of reach? What is the form of contact? We still need to do it step by step in a targeted manner.

Still taking Pinduoduo as an example, we formulate corresponding operation strategies for early warning touch actions:

• The user churn rate brought by the friend invitation channel is high, which enhances guidance to these users; and strengthens incentives to these users.

• 90% of users in the growth period only purchase 1-2 times, and users are cultivated by issuing coupons and providing exclusive products.

• 95% of users in the growth period have purchased less than 200 yuan, which will enhance media publicity, enhance users' trust in the platform, and increase users' chances of purchasing large-scale products; recommend more accurate products to users, and come with large-scale coupons to actively help users discover needs and stimulate conversions.

• 70% of users in the growth period have not initiated invitation activities, strengthening guidance on invitation activities, or when users have certain actions continuously in the App, additional incentives for these users to participate in friend invitation activities, such as as long as they participate, they will immediately give a 5 yuan voucher without threshold.

(The above data on Pinduoduo's user composition and different user behaviors are hypotheses, mainly to present the theories in the article through actual cases. If there is any inappropriateness, please point them out in time.)

Here are some common channels of reach:

Email recall

• advantage:
• Lower cost
• The email can convey a lot of content
• New email reminders for emails are in place
• shortcoming:
• Domestic users have lower email usage rates and lower email opening rates;
• The sending settings have certain technical requirements, otherwise they will be easily blocked.

SMS recall

• advantage
• High delivery rate
• shortcoming
• Cost higher than email
• Smartphone software may block text messages
• A single product type

Site letter recall

• advantage
• The cost is relatively low
• High delivery rate
• shortcoming
• Low open rate

Push push

• advantage
• Relatively low cost
• High reach rate
• High user opening rate
• shortcoming
• Users can't receive messages when uninstalling the app
• If you set up interception, you won't receive a message

Telephone follow-up

• advantage
• More humane, high success rate
• shortcoming
• Low efficiency
• High cost
• High requirements for the initiator of the phone, such as speech, pronunciation, etc.

At this point, the content about the construction of the user's life cycle model has come to an end. Although there are cases in the article, if there are no actual operations, these are all in the stage of talking on paper for us. I hope you can really do it with the opportunity. Everyone is welcome to discuss and communicate in the message area.

Source: Chris