Building a community operation data system from 0 to 1

Community operation is the product of the Internet traffic dividend. We can see that no matter what kind of Internet work you are engaged in, you will almost always use the community. Since community operation is so widely used, how can we make it more systematic? The author shared his method of building a community operation data system from 0 to 1. Let’s take a look.

1. What is community operation?

From a macro perspective, community operation is the product of the gradual disappearance of the Internet's "traffic dividend". All companies have gradually shifted from the wildly growing "land grabbing movement" in 2015 to the "intensive cultivation" of vertical traffic mining. The definition given by Baidu Encyclopedia is: Community operation refers to connecting group members with certain ties so that members have common goals and continuous mutual interactions, and group members have common group consciousness and norms.

For a simple understanding, see the following figure:

Community operation can be understood as the establishment of connections between operators and users through communities (corporate WeChat, WeChat groups, etc.). Operators provide positive feedback to users by inputting product concepts, product function introductions, product promotions, etc., thereby guiding users to become loyal users of the product and then convert to paying users.

2. Why do we need to do community operation?

At present, the community operation model has been perfected in the entire ecological chain of some large companies, but it is a relatively new element for small and medium-sized companies. If decision makers can seize this wave of new operation models, they can improve their own ecological chain and bring huge benefits to the company.

3. Why should we build a community operation data system?

In the entire community operation ecosystem, if there is no relatively complete or convenient and fast query data system available for community operation, the community operation personnel will not be able to quickly, effectively or even accurately judge the conversion effect of users in each link of the entire community operation, and therefore will not be able to efficiently optimize and improve the operation strategy. If this continues for a long time, they will lose the offensive advantage in this field.

At the same time, many of the tasks performed by the community operation team and individual team members cannot be accurately measured and evaluated, which greatly limits the imagination space for the development of community operations.

Finally, without current status data, we cannot set a reasonable North Star metric to effectively guide our operational plans.

Therefore, building a community operation data system can solve four problems:

1. Develop a North Star Metric;
2. Quantify operational results;
3. Improve operational efficiency;
4. Evaluate member performance.

4. How to build a community operation data system?

The hypothetical scenario of this article is to build a community operation data system for a toc software that provides multiple financial management tools. The specific business logic is: users who have purchased any paid product in this software will be directed to the mini program within the terminal (in the product). In the dialog box of the mini program, users will be guided to add the operator's WeChat through the QR code on the page. At this time, the operator and the user have established a connection. Through operation, users can become aware of the brand and maintain stickiness, and then achieve conversion again, such as payment.

We build a community operation data system in four steps.

1. Sort out business logic

Because our product positioning is to provide users with financial management tools and products, the target users of our products are those who have investment and financial management needs. For the community business, our target users are those who intend to pay but have not paid or have already paid and we hope to pay twice or multiple times. Based on the above split, our business logic is as follows:

2. Build an indicator system

Based on the business logic funnel, starting from the result indicators and process indicators, combined with the data dimensions, we can build the following community operation indicator system:

3. Data access

Why do I single this out here? This is because the data that enters WeChat from the mini program is stored on the WeChat platform. We need to transfer the data back to our own database through the interface provided by WeChat.

The user IDs on the WeChat side will include openid and unionid, and there may be a one-to-many or many-to-one situation. It is necessary to determine the scope of operations in advance with the operations staff to prevent inconsistent data expectations on both sides after the data goes online.

4. Build an analysis framework

After the previous steps, we now have the data in hand and have built an indicator system according to the business logic. Next, we need to build an analysis framework according to the business logic.

For example, in the first two sections of the business logic, we will use channel analysis to grasp the quality of the channel, and in all funnel conversion processes, we need to use churn analysis.

Through churn analysis, we can identify which type of people are churned out? At what stage does the loss occur? It is then relatively easy to formulate corresponding strategies.

Here we introduce several common analysis methods, and their application scenarios are not limited to community analysis.

(1) Competitive product analysis

Competitive product analysis generally looks at the various types of apps installed on users' phones. These apps can usually be divided into two categories: competing products and non-competitive products. For this type of data, we usually make a correlation chart between whether users install or not and retention, that is, among users who use our products:

• The retention rate of users who installed product A is lower than that of users who did not install product A. We can assume that product A is our competitor and has taken away our users.
• The retention rate of users who installed product A is higher than that of users who did not install product A. This means that product A cannot meet the needs of users and our product is more competitive.

The retention rate of users who have installed Product A is approximately equal to the retention rate of users who have not installed Product A. This means that there is a large overlap between the user groups of Product A and our products, so we can consider cooperation in exchange for volume.

(2) Key behavior analysis

Key behavior analysis generally looks at the relationship between whether a user performs a certain behavior and the target data (such as retention). This way you can see which behaviors are the “high points” for users to become familiar with the product.

Generally we need to define key behaviors. Users will have hundreds or thousands of behaviors on our products. Business personnel can define the key behaviors of users from the user's perspective based on their own business understanding. Data analysts can classify user behaviors and find out those behaviors that are large in volume and have a significant difference in impact on target data (such as retention) when done or not. As shown below:

The above figure shows the six key behaviors we found with the business, including:

• The horizontal axis is the next-day retention rate of users who have performed the behavior. The closer the bubble is to the right, the higher the next-day retention rate.
• The vertical axis is the next-day retention of users who performed the behavior/the next-day retention of users who did not perform the behavior. The higher the bubble is, the greater the impact of the behavior on retention.
• The width of the bubble represents the number of users, and the larger the bubble, the larger the number of users.

• From the above figure, we can conclude that:

1. Behavior A and behavior C contribute more to retention. Assuming that for behavior A, users need at least five steps to reach behavior A after opening the app, then we can shorten the path to A from the product perspective and allow users to reach behavior A faster.
2. Behavior E is relatively poor in improving retention effects;

Operations personnel should guide new users to acquire behavior A and behavior C as quickly as possible.

(3) Churn analysis

We generally define lost users as users who leave the app on the same day and do not open our app for the next period of time (one month, three months, etc.).

Churn analysis We generally analyze the behavioral differences between churned users and non-churned users before they leave the app. The reason why you need to look at both types of user behavior is that, if you find that 70% of churned users did behavior A before churn, you cannot do the following:

The conclusion is that behavior A leads to a large number of user losses. Because non-churned users may have performed behavior A in large numbers before leaving the app on the same day, but this does not prevent these users from coming to our app again the next day.

In this process, data analysts need to classify the user's behaviors in the last few steps and then draw conclusions from them.

The process is time-consuming. This process not only requires you to have strong inductive ability, but also requires you to have solid SQL ability. Because in this process you may constantly use regular expressions to classify scattered behaviors into several major categories.

After you have obtained some key behaviors before user churn, we need to intervene manually when these behaviors occur, and use real-time push tools to follow up with corresponding strategies after the specified behaviors occur.

(4) Search analysis

Search analysis refers to the analysis of keywords users enter in the search box? Why do I classify search analytics as a separate analysis method? Because for new users, based on their unfamiliarity with the product, even if you have good features, they may miss out on using them because the entry point is too deep or they cannot understand what the feature means. At this time, search becomes a place for users to vent.

Users who have not directly lost will enter their product usage needs into the search box using some keywords. For general products, search traffic is relatively abundant, and there is a lot of information to be mined.

New user search analysis is similar to churn analysis and both require summary.

First of all, we need to extract the keywords searched by users and classify them to see which functions we have but are missed due to weak user perception, resulting in direct loss of users. Users need to be guided to use these functions. In addition, there are some users’ demands for which we do not have corresponding functions to meet. In both cases, we need to slam the product manager’s desk and ask him to put the function online as soon as possible.

5. Provide decision support

Finally, with the analytical framework, what support can we provide to the business? For community operations, we generally provide three types of decision-making support: operational strategy formulation, contribution evaluation and finding growth points.

Among them, the formulation of operational strategies requires algorithms to provide a population library and a material library. In this way, we formulate and consolidate operational strategies through AB testing.

Finding growth points requires us to apply various analysis methods in the analysis framework built in the fourth step to find breakthroughs in the business. For example, on 818, we conducted a successful community event, which attracted a sharp increase in the number of new members. We analyzed and reviewed the data and summarized the reasons for the good performance, and used it as the direction of community fission.

Contribution evaluation refers to the performance appraisal of community operators. We can calculate weights through multiple dimensions such as evaluation of the effectiveness of operational activities, scoring of the quality of operational communities (group activity, life cycle of community members, etc.), scoring of operational means (number of new members attracted, etc.), transaction amount, etc., and finally calculate a comprehensive score for rating.

5. Summary of the establishment of community operation data system

1. Sorting out business logic
2. Building an indicator system
3. Data access
4. Building an analysis framework
5. Providing decision support

At the same time, this process can also be reused to build data systems for other business scenarios.

Author: Dongdian Data

Source: Dongdian Data