Example analysis: “Super user” data operation rules!

The article takes NetEase Yanxuan as an example to describe the data operation rules of "super users". Let's take a look at it together~

Total product value = active user scale ╳ single user value – abnormal user loss

This article shares three points:

1. Three things to do for user growth;
2. User growth indicator system;
3. Strategic tools for user growth.

1. Three things to know about user growth

1. Identify real growth

Suppose there are two products: Product 1 did not have large-scale promotion during the Spring Festival, and its users declined slightly, but returned to a stable growth state in March and April. Product 2 usually maintains a stable growth rate, but it suddenly expanded 10 times during the Spring Festival.

What is your first reaction when you see these two sets of data?

If you have determined that your product has not had any large-scale communication events to trigger growth, we recommend that you first pay attention to whether these users are real users. Calculate the scale of user growth through the account system, and then see whether there are account associations between these users and whether they have unified profit-seeking behavior. If so, you are likely not getting real users, but wool-gathering parties, etc. Only real users who continue to use the product can create value, while those users who demand extreme benefits often bring losses to the company.

2. Expansion of active user base

When we have a strong moat and can obtain real data, we still need to care about two things: the expansion of the scale of active users and the growth of the value of individual users. Taking Yanxuan as an example, the monthly GMV is actually contributed by users who made payments that month. Therefore, we must ensure that the number of new users added each month + the number of recovered lost users is greater than the number of lost users in order to ensure the continuous expansion of the active pie.

3. Growth in the value of a single user

User growth relies on a strict management system to identify the survival and retention of users on the platform. Furthermore, the growth of individual user value requires hierarchical management to meet users' different needs for products.

Take Yanxuan as an example: we sell a lot of four-piece sets, and we also have cultural and creative products to meet the needs of students. The purchase amount of home users will be higher than that of cultural and creative product users, but users are changing. Only by identifying and satisfying the user needs at the current stage can we ensure that we can meet the user needs at all levels one by one.

We believe that the essence of user growth is to increase the total value of the product. First, we must ensure the expansion of active users and the increase in the value of individual users, and then subtract the losses caused by abnormal users.

2. User Growth Indicator System

How to use data to drive growth?

From a data perspective, by building an indicator system for user lifecycle management, you can achieve large-scale user management at different stages. Especially for C-end products, the user group you face is very large, and you need to increase the radius of user management by a single operator in order to manage various types of users more efficiently.

The core of the indicator system is to make large-scale data management controllable and decomposable, and to make indicators driveable and influenceable. The construction of the indicator system is to attract new customers and promote activation.

Therefore, we will split the performance by terminal, such as: mini-programs, media channels, and the way of delivering the user's interests and the attributes of the user itself.

When users enter the product, a large amount of data will be generated, and we can obtain many interactive behaviors between users and the platform. At the e-commerce level, indicator construction will revolve around the three indicators R, F, and M, which evaluate the value of a single user, including R (the time since the last payment), the number of payments within the F period, and the payment amount within the M period (reflecting the user's purchasing power).

In addition to RFM, there is another very important indicator A, the interaction between users and products. In the current e-commerce environment, it not only carries user needs, but more importantly, creates user needs, allowing users to generate more needs after they come and lock in their choice of you. For example: the very popular internet celebrities on Taobao, and the sales models of Toutiao and Douyin, all deepen feelings through continuous interaction, and ultimately allow users to choose and recognize the platform.

Yanxuan has a dedicated physical channel to carry user content and generate demand through positive interactions. Therefore, we will build R, F, M, and A around improving the value of individual users. By combining this system, different customer groups can be constructed, for example: R is particularly high, F is not high, and both R and F are high. In this way, users can be structurally divided and the structure of active user scale can be managed. The potential groups in the user scale structure can be identified, and pain points such as churn can be found, so that group management can be done.

However, if we only focus on the data within the product, the knowledge we gain comes entirely from the product's own behavioral performance. Through various analysis, we find that all indicators have declined, and we cannot find the reason. We need external macro indicators to achieve a comprehensive evaluation.

For example: evaluating the brand's coverage or growth space, which is the brand awareness of the Yanxuan platform; if we are more concerned about whether users have the ability to spread and the willingness to convert, we will pay attention to users' satisfaction with the product and the frequency of searching for products, and understand our position in the users' minds. In this way, we can use auxiliary data-level indicators to evaluate user intentions and obtain results that are closer to the actual situation of users.

Focusing on growth, data analysts will design data strategy products to make it easier for business students to achieve growth. This is also the concept of data-driven management.

User growth strategy tools

In the stage of attracting new customers, customer acquisition channel management will be involved. When users enter the platform and generate a large amount of interactive behaviors and data, it is necessary to build a user label system and provide supporting CRM closed-loop services for refined operations. The refined management of Yanxuan has people on one end and goods on the other end, so in order to achieve a fine match between people and goods, the coordination of the product selection system is required.

We have a four-quadrant indicator system, including channel customer acquisition capabilities, cash flow contribution capabilities, the scale of new customers and channel user quality assessment, to comprehensively measure channel performance.

Through the application of thinking indicators, in practice, algorithm models will be used to build a channel scoring system to describe channel characteristics. When new channel resources come in, we can quickly test to determine which type of channel this is on our platform, draw a benchmark line, better identify its characteristics, and implement channel KA management at the same time.

For example: For high-quality channels, conduct deeper user cooperation and embed the registration entrance into the product. You can also put materials to recall lost users in the channel. Users who have lost users through Yanxuan may also visit this channel again, so it is valuable in both attracting new users and recalling lost users.

Through the channel scoring system, we can achieve control over the channel, that is, improve efficiency and control costs, so that the business side can spend money reasonably.

Once the platform has a large amount of product cooperation data, it can build a labeling system that reflects its own product characteristics. In the design of the labeling system, we first collect basic data, and then there will be a basic cleaning work.

It is important to note that the cost of managing large amounts of tags is very high. We recommend adding a data cleansing layer to ensure that business logic and basic integration are clear after large amounts of data are imported. Various ETL personnel at the upper level will form a unified indicator system when processing data, and the resulting labeling system will be more valuable.

There are two types of labeling systems: statistical labels and predictive labels.

This is pretty much the same across all platforms. What we are trying to do is to identify purchase intentions, for example: how to convert users who have visited for a long time but have not placed an order? This is a very typical scenario in e-commerce. We find users with purchasing intentions on the platform and demonstrate their intentions through their interactive behaviors. With these tags, we can find products and interest points that match them and make the final targeted contact, which often has a better conversion effect.

In summary, the work of data analysts in this part is to profile, quantify, and target:

1. It is to regularly review the overall user situation from the perspective of analysts, so that the business side can understand who the users are;
2. It quantitatively describes the user scale of different groups to help business colleagues better manage groups.
3. It is to lock in the target customer group and achieve business goals through the flexible use of labels.

Once the target customer group is locked in, a closed-loop user operation practice is completed.

Going further, we will design an MVP operation model to achieve closed-loop data services. This includes locking in users and finding matching operational strategies based on user characteristics, such as: what kind of interest points to choose and how to deliver them. Then monitor the operation strategy and build a set of long-term available indicator system to make it reusable, because the closed loop is not completed in one go, it requires repeated experience accumulation to verify the optimal effect, and finally form a user operation strategy system through continuous testing.

In order to achieve this strategic system, the previous closed loop must be put into operation quickly. Therefore, we built a CRM system so that MVP experiments can be implemented on products in the future and operations can be better configured, including communication methods, service methods, and label and user management capabilities. In this way, the MVP model will realize the iteration of users + methods + gameplay + copywriting + products in flexible scenarios, and ultimately form a strategic and systematic solution.

Back to Yanxuan itself, all growth centered around users is inseparable from products. So back to the product, we are more focused on matching people and goods. Combine products, user needs and promotion scenarios, for example: choose classic products to attract new customers, to achieve growth in user comprehensive value. In practice, we have also found that a high-quality product is far more effective than a lot of gameplay measures.

Source: Umeng Global Data