How to manage active users?

User operation is a relatively lengthy process. We need to manage and activate users, as well as finally convert them. And today’s “fine-grained operation” is also a way for operators to better activate and convert them. The author of this article shares his analysis of active users. Let’s take a look together.

"How can we fine-tune the operation of active users? What is the difference between user stratification and user grouping? We have shouted the slogan thousands of times, but when can we achieve fine-tune the operation?"

After new users experience the core value of a product, they will spend more time and energy on the product and gradually become active users of the product. Active users are users who recognize the value of the product and are willing to pay for or endorse the product. They not only contribute their own value, but are also critical to the brand building of the product. Therefore, the importance of active user operations is self-evident.

The operation of active users is a huge topic. Today I will only try to explore this issue from some specific angles. Although I cannot fully explain this topic in one go, I hope to provide some ideas and thoughts for everyone.

01 What is refined operation?

In the past, the market was more concerned about how to acquire customers on a large scale and at low cost. As the demographic dividend gradually disappears, the cost of acquiring customers is getting higher and higher. Now, more and more people are paying attention to how to enhance the value of individual users, spend money where it counts, let different users enjoy different services, let users feel the warmth, and give the product soul.

Thus, "refined operation" was born. When talking about user operation, we cannot escape the word "refined". It seems to have become the basic ethics of operators. When communicating with others (chui), you would be embarrassed to say that you are an operator if you lack these words. But what exactly is refined operation? How should it be implemented?

The so-called refinement, first is accuracy, the second is segmentation, the two complement each other and are indispensable. If you want to achieve accuracy, you must segment. You will never be able to retain users if you try to do everything at once. The most ideal situation is to have a thousand faces for a thousand people. But how to segment users? Here are two very commonly used methods: user stratification and user grouping.

02 User stratification vs user segmentation

User stratification vs. user grouping seem similar, but there are obvious differences in positioning and goals. User stratification is a division based on a general direction, which is what core goal you want users to work towards, while user grouping divides them into finer granularity to facilitate targeted operations and improve results. The two complement each other.

The layers in user stratification are hierarchical levels. For example, we start from the time when users register and use our products to become our new users, to becoming active users, to frequently active users or paying loyal users, and then to later when other competing products emerge or the product functions themselves no longer meet the needs, the users begin to become silent and eventually churn away. This life cycle is also a level, as shown in the figure. With this stratification, we can know the composition structure of current users more clearly and whether the user growth in each life cycle is healthy.

Is this enough? We know that the 80/20 rule exists in many fields, that is, 20% of people contribute 80% of the revenue. As for loyal users, some of them are civilians with low per capita consumption, and some are wealthy sponsors who spend money like water. In such cases, we need to further refine the loyal users and divide them into more detailed groups.

For example, a new credit guidance has been introduced to the product recently. The desire is to see if this helps retain new users, or an operational campaign is launched to see if the core indicators have improved. At this time, the users need to be further segmented, resulting in grouping. Grouping is a further segmentation of the layers, which makes it easier to conduct precise operational actions against the users.

Commonly used methods for user grouping include the familiar RFM, Kmeans based on data mining, and so on. The former uses the most recent consumption time, consumption frequency and consumption amount to measure user value, and groups users into high-value users, general-value users, important retention users, etc. However, the establishment of the RFM model requires expert experience, which means that the selection of indicators and the determination of the thresholds of each indicator must be based on business sense, rather than being decided on a whim.

Kmeans mainly uses data mining to find users with similar characteristics, so that birds of a feather flock together. After users are clustered, targeted operations can be carried out by analyzing the characteristics of each group.

03 User stratification application case

Next, we will implement the theory of user segmentation through a case. The case is fictional only for the purpose of illustrating the problem. First, we assume that the changing trend of the number of active users is as shown in the figure below. At first glance, the number of active users per month is continuing to grow, which seems good.

However, we must be wary of the illusion given to us by vanity indicators. We can put in the cumulative number of users, that is, the cumulative number of users up to the current time, and divide the number of active users by the cumulative number of users to get the user activity, which represents the proportion of active users to the total. At first glance, it seems that the proportion is gradually decreasing.

We can continue to subdivide and calculate the number of new users based on the cumulative number of users, and find that a large proportion of active users are new users.

Similarly, we can divide the cumulative users into new users and old users, and divide the active users into new active users and old active users. Similarly, we can get the activity of new and old users. We found that the activity of old users is lower.

We want to see what happened among old users? We further subdivide active users into two categories: active and inactive users. Active users include new active users and old users. Then we further divide old users into general active users, loyal users and returning users. Inactive users mainly include silent users and lost users.

We found that the main reason why old users are active is that there are few active and loyal users, but there are many new users. This shows that we need to guide and retain new users, and at the same time encourage users to convert into loyal users.

Furthermore, we can segment the users each month and analyze the composition of users at different levels in the same month to determine the health of user growth.

But for greater clarity, we look at the composition of users based on active and inactive users, so that we can see the health status of users at each level more clearly.

Users continue to grow during the product life cycle. In addition to looking at the active composition of users at a certain point in time, we may also need to pay attention to the user's growth path: How many new users become active users every day? How many active users became inactive? How many loyal users have become inactive? How many lost users have been recalled by us, etc. This will help us analyze the user's whereabouts more intuitively, locate the problem more accurately, and take targeted actions.

For example, the growth path of new users of a product in January can be displayed in the form of a Sankey diagram. It can be found that a considerable proportion of users are no longer active in February and have become silent users. It is necessary to reach these users through operational means in a timely manner to prevent them from leaving in March.

Similarly, active users or silent users over a period of time can also be monitored in a similar way to understand the user's whereabouts in a timely manner and intervene in time to prevent user loss.

04 User grouping application case

The above example illustrates the ideas and methods of user segmentation. The following example introduces the application of user segmentation. There are some commonly used methods for user segmentation. For example, the empirical RFM model can be used to evaluate users from different dimensions and then divide them into users of different values ​​for operation. Or, clustering algorithms of big data mining can be used to mine similar features of a large number of users to achieve the goal of grouping people by their own like.

These methods are already mature and many people are already familiar with them, so I will not go into details here. Today I would like to introduce to you another important grouping method - cohort analysis. The so-called cohort analysis is to further segment stratified users, group users in the same life cycle, and see the effect of similar grouping.

Generally speaking, the same cohort needs to meet the following requirements: be in the same life cycle, for example, the users being studied are all new users, or users with common behaviors. This way, we can see the changing trends over time within the group, and we can see the effects by comparing different groups. It is generally used to measure the effects before and after of product or operational optimization plans. For example, we launched a new feature in February, which resulted in significantly better new user retention in March and April than in January and February. By comparing the new user retention in January and February with the new user retention in March and April after the iteration, we found that the optimization plan was effective.

Let’s use a case to illustrate the specific application of cohort analysis. Suppose we get the sales data of a certain store and find that although the sales volume and number of customers continue to grow every month, the customer’s ARPU continues to decline. Is the customer’s purchasing power gradually weakening?

To explore the reason, we first stratified our customers into new and old users, and then conducted cohort analysis on both new and old users.

We first conducted a cohort analysis on the ARPU of new users from January to April, that is, we took the new users of each month as a cohort, and studied the changes in ARPU of different cohorts in the first month and thereafter. We found that with the passage of time, the ARPU of the first month of new users from January to April is constantly increasing, indicating that the purchasing power of new users is constantly increasing, which means that the purchasing power of old users is likely to have declined.

Similarly, we conducted a cohort analysis on old users and found that over time, the ARPU of old users was gradually decreasing. It was the decline in the purchasing power of old users that led to a decline in the ARPU of overall users.

05 Conclusion

This article attempts to start from the operation of active users, explore how to implement the refined operation of active users, and two important methods to achieve refined operation - user stratification and user grouping. It also gradually demonstrates the detailed steps of applying the two methods through cases. It is hoped that the refined operation, which is praised by everyone but rarely realized, can be implemented by combining theory with cases.

But it is undeniable that the operation of active users does not mean that you can sit back and relax once you have mastered these methods. Users' cognition and needs are constantly changing with the rapid development of the Internet.

We cannot expect to win over users through some fixed methods and routines. All methods and routines are aimed at understanding users as accurately as possible. Continuously providing users with soulful products and high-quality services is the most advanced method and routine to make products last long.

Author: Big Data Analysis and Operation Planet

Source: Big Data Analysis and Operation Planet