Tips for building a community operation model!

Before writing this article, I searched for some community articles, most of which focused on various specific group gameplay and solution sharing.

How big is the impact of these gameplays and plans on community indicators? Is it applicable to my project? How to judge how to choose?

This article does not discuss solutions, but returns to the essence of operations. Guided by data, we review the ideas behind building community operation models in specific projects.

The data appearing below are all simulated data.

1. Project Background

In addition to traditional lightweight knowledge-based paid SKUs, a certain platform hopes to expand into educational courses.

After acquiring the copyright of a well-known foreign IP, this new education course project was launched, and the internal team developed an online English course for K12 users. The course is based on the WeChat mini-program, presented in the form of animation, and has functions such as interactive games, listening and reading, etc.

After the product was launched and completed internal testing, it began to be put into operation and commercialization. I entered the industry during this period. We built and implemented a community conversion model, and within half a year, relied on the community model to grow into a project with monthly revenue of tens of millions.

2. Implementation Review

MVP testing, initially building business paths and operational conversion plans (increasing conversion rate by 4 times and shortening conversion cycle by 1 time through social media conversion).

Data analysis drives operational growth.

• Identify key behaviors, typical user paths, and growth points.
• Through the user life cycle model, a data-based and measurable community operation model is established. The number of users has increased 10 times. Through life cycle model and user behavior data analysis, the conversion rate indicator is broken down into 5 sub-indicators, and the sub-indicator data and corresponding strategies are clarified, making the indicator controllable, monitorable and scalable.
• As the user base continues to increase, we will segment users and match them with corresponding operating mechanisms (when the user base grows exponentially, we will find segmentation methods and divide different community operating strategies).
• Unfinished part: After the user base is scaled up, automatically match the corresponding operation strategy according to user segmentation tags or behavior tags to make the operation productized and institutionalized (a direction that can be done but has not yet been realized).

Optimize the operational growth model (improve the logic of full data, collaborate with product, teaching and research teams, and improve the overall North Star indicator).

• Disassemble the entire chain, sort out each business link, and improve and optimize the data of the segmented links.
• By factoring, we can break down the indicators into sufficiently detailed forms and improve each sub-indicator.

Phase 1: Small-scale MVP testing to determine the business model

Business model exploration:

Taking into account the heavy-duty education course marketing, that is, the conversion payment of low-frequency, high-price products, the general marketing method is to purchase after multiple experiences. That is, users need more experience and value the experience before the first purchase, so adding communities as an experience link promotes conversion.

Business path: trial class exposure - sign up for trial class - join the community - experience course - first payment - study course - pay again.

Preliminary in-group operation plan, in-group user path: join the community - consult and answer questions - experience learning - lecture learning - paid purchase.

Determine feasibility based on data feedback:

To determine whether MVP is feasible, there are two data comparison dimensions:

Horizontal comparison: comparison of average data of similar products in the industry.

Vertical comparison: comparison of historical social data of this product and comparison of other business model data of this product.

Since this is the first time we use community operation methods and there is no past data internally, we chose the existing business model data of the product for comparison and compared the final order volume under the same traffic exposure.

1) Original business model: traffic operation

Like other SKUs on the site, traffic conversion is achieved through operations within the APP.

More detailed user data, such as bounce rate, dwell time, trial listening status and other reference data, are used to adjust the optimization direction, such as adjusting the course title, details page, user reviews, etc.

2) Test business model: community training camp model

Within its own platform, users are directed to WeChat communities to achieve experience conversion.

There are also more detailed user path data and user behavior reference data, which can be used as a reference for optimization direction, such as the ratio of users who successfully receive coupons and register, the ratio of users who successfully call to jump to the APP, the ratio of users who successfully call to add WeChat accounts on the WeChat terminal, the number of users who are active in speaking after joining the group, etc.

3) Both business solutions have their own advantages and disadvantages and corresponding optimization space

The first model has a smoother conversion path, but it requires a high level of direct payment conversion capabilities on the course page; the second model lowers the user's decision-making threshold, but it results in greater traffic loss and requires more operating manpower costs.

After actual testing, we calculated the number of first-time orders that were finally completed under the two models with the same amount of resource exposure, and compared and analyzed the feasibility of the new solution.

The following are the business data of the two business plans (the data in this article are all assumptions):

Option 2:

After obtaining the data, there are three directions for consideration and prediction:

• Data Floating Space: While collecting key business data, you also need to pay attention to process data related to operations. This is an optimization action that operations can take during the testing cycle, which can qualitatively estimate the room for data improvement. The key CTR/conversion rate data of Plan 1 can also refer to the average or optimal value of other projects within the company. Plan 2 needs to consider the upper limit of traffic loss when users jump at each link and the possible decline in community conversion rate.
• Comparison of MVP data results: Finally, we obtained the data results of the two plans within one cycle. The first-time order volume of Plan 2 is twice that of Plan 1. With the floating space, Plan 2 shows an advantage. We preliminarily judge that Plan 2 is feasible.
• Prediction and deduction: If you have experience from your peers or do research on competitive products, you can do one more step of deduction.

Calculated based on a single month's revenue, assuming that the number of new users joining the community is 100 people per day.

First purchase GMV = 100 new community traffic * 30 days * first purchase conversion rate 0.2 * first purchase unit price 100 yuan = 60,000

Repeat purchase GMV = first-time purchase users 600 * repeat purchase conversion rate 0.2 * repeat purchase amount 2000 yuan = 240,000

Cost = personnel cost + traffic cost

Each person can manage 2,000 users per month, assuming the salary cost of two people is 20,000 yuan.

ROI = GMV / cost = (GMV of first purchase 6W + GMV of repeat purchase 24W) / (staff cost 2W + traffic cost 3000 people * traffic cost of a single community x)

Conclusion: The cost of new traffic for a single community is x93 yuan, and the ROI is 1. The customer acquisition cost of 93 yuan is achievable for both on-site and off-site user acquisition in the future.

Within one month of MVP testing, the community operation conversion model was finally established, which actually increased the conversion rate by 4 times and shortened the conversion cycle by 2 times, creating an opportunity for the project to grow on a larger scale.

Phase 2: Initial growth phase - data-driven, building a user operation model

This phase is the beginning of formal operations, while collecting and analyzing large amounts of data to find growth areas. And establish measurable segmentation indicators around the user life cycle model.

1) Data collection and analysis

Three types of data are mainly collected (for analysis purposes):

• User behavior data - find out the key user behaviors (key behaviors that affect key indicators).
• User path data - user usage data at each stage of the product (summarizing typical user paths and finding general rules in the process of improving user value).
• Basic user information - potential user portrait (① basic judgment on abnormal data; ② user segmentation to a certain extent; ③ distinguish channels or judge channel quality), paying user portrait (if there are commonalities, it can provide direction for the acquisition and operation of accurate users).

2) Find the key behavioral factors that affect key indicators

After users enter the community, user behaviors that may affect the conversion rate may include: speaking and asking questions, opening courses, completing courses, participating in group activities, receiving coupons, etc.

Then, taking into account the time points, users' daily course opening rate, completion rate, daily speaking frequency, lecture participation rate, and coupon collection rate are all data that need to be closely monitored.

Taking into account the data sample size and statistical difficulty, we first conduct quantitative analysis on the opening rate, learning rate, coupon collection rate, etc. of the mini program courses, and label users based on their speaking frequency in the community, lecture participation, and other behaviors to make qualitative judgments.

Taking learning as an example, can the learning rate affect the purchasing decision of users who get the course for free and only have 3-5 days to study? How big is the impact value and is it enough to be a critical factor?

The backend can pull out the batch of courses that started on February 2, 2020, and there are 200 class students. The course learning data of these students are as follows:

The table contains basic user information, the last opening time and course learning status. There are 4 parts in each class, where 1, 2, 3, and 4 refer to the number of parts in the course that the student has learned, and 0 means that a part has not been learned.

This background data is counted according to the learning status of the course and has no time dimension. It is necessary to pull out the learning status and payment status of these students within the defined community operation time. Do cross comparison.

To understand the relationship between learning rate and payment rate, it is necessary to ignore the user’s basic information first.

Data processing: Define that within a time period, if a student has studied one link of the course, he or she is considered to have studied the course. The original data is processed, 0 means that the student has not studied the course once, and 1 means that the student has studied at least one link. Get the learning rate of a class.

Zooming in on the granularity, looking at the learning rates and payment rates of multiple classes, and comparing the learning rates and conversion rates on a class-by-class basis, there is a positive correlation, but the extent of the correlation is not clear.

Back to Table 1, we extract the learning status of 10,000 trial class users in the past month.

You can collaborate with BI colleagues to put the data into the corresponding model to find patterns and find aggregate classification methods.

There are 12 learning sessions in total in 3 trial classes. 20% of users completed 1-3 sessions, and the payment rate was 55%. Next was 12% of users who completed 4-8 sessions, and the payment rate was 75%. Even fewer, 8% of users, learned more than 9 sessions, and their payment rate was 80%. If the user does not take any lessons, the payment rate is 6%.

40% of learning users contribute 94% of payments, which clearly shows the correlation between user learning and payment. The payment rate of users who have completed more than 4 steps is higher than that of users who have completed 3 steps. However, because the proportion of users who have completed 4-8 steps is smaller than the total number, the current operation strategy will focus on guiding users to complete at least one course.

Secondly, we see a special data: users’ learning situation is not divided by the number of course sections, but by 1-3, 4-8. In light of actual business conditions, the fourth part of the course is the student follow-up output part, and a large number of learning users are lost during the follow-up part.

According to operational feedback, the reasons for churn may be:

• The person who tried the course was a parent, and he jumped out when the fourth part, which required him to follow along, came up.
• It is somewhat difficult to repeat sentences, and some students do not have the patience and ability to complete it.

The several questions raised here can be further investigated, and these reasons will also be included in the scope of the next step of optimization.

Now that we know the importance of learning rate, we hope to further analyze what factors affect the learning rate.

Same method:

• Based on user base data, are there users of certain age groups, regions, or consumption habits who are more willing to start learning?
• From the user behavior data, whether there is a regularity in the average learning time of users, the distribution of user learning time, the frequency of user learning, etc.
• Judging from the user path data, are users more willing to start learning on the first day of joining the community, on the day when coupons are distributed, or on the last day? Do the user proportions of these different nodes differ significantly? Are their payment rates significantly different? And so on.

If it is not possible to obtain values ​​and conduct quantitative analysis immediately, what should operations do?

Operations can also use qualitative methods to make preliminary verification and judgment.

We want to know on which day our users are more willing to open the course, so we add a sharing session that allows users to actively share their study time and learning results.

Get feedback through methods such as check-in incentives and homework grading. At this time, you can find that the number of users who check in for learning is the highest on the first day of joining the community, the lowest on the second day, and will increase again on the last day. The corresponding strategy to improve the learning rate could be to provide new user guidance on the first day, retention activities on the second day, and learning rewards on the last day.

The above only takes the learning rate as an example. Other influencing factors can also be analyzed in the same way to ultimately find out the key behaviors.

3. Build a user lifecycle management model

Community users have a clear process of joining the community, participating in learning, and leaving the community over time, so the user life cycle model is used to build a community user operation system.

When we were working on the community MVP, we were able to summarize some behaviors that may affect community payment conversions, including asking questions, taking courses, participating in lectures, and receiving coupons.

Just like analyzing the correlation between user learning behavior and payment in the previous section, during the operation process, the user definition can be gradually clarified through more specific user data collection and analysis.

1. Clarify user definitions through analysis of user behavior data

In the previous stage, it has been discovered that as long as users complete one part of a course, there will be a higher payment rate, accompanied by consulting behavior. Completing a course, which means more than 4 parts, is more likely to happen on the second day or the last day. Therefore, users who have opened the course are defined as activated users, and users who have completed a course are retained users, without requiring 2 or 3 courses.

Similarly, for dormant users, we use multiple channels such as SMS push APP push and WeChat push to wake up users, and we find that the wake-up rate of users on the WeChat channel is the highest. After accumulating a large number of users on the WeChat channel, we tested and found that the user awakening rate within one month is higher than that within three months. At the same time, the payment intention is also extremely high, second only to the user payment rate within a 7-day cycle.

2. Develop operational strategies

With a clear definition of users, you can then formulate corresponding policies.

3. Execute the operation plan

The operation strategy is a set of mechanisms and rules, and each strategy will have more specific operation plans based on actual conditions.

For example, for the same new user guidance strategy, what kind of title, what background color of the picture, at what time it pops up, and how many times it pops up can encourage more users to click.

For example: Retention is more difficult in community operations. A common situation is that users are very active after just joining the community, but soon as time goes by, fewer and fewer users open the group. First-time purchase communities also have the same problem. Because of the experience conversion community, we provide two strategies:

1. Give rewards for continuous learning;
2. Preview the next day’s lecture to provide users with value and benefits.

Taking the lecture as an example, the specific operation plan is:

User retention strategy: Community activities-lecture operation plan:

• Lecture format: text lecture.
• Frequency of lectures: Day 3.
• Lecture target: users who have active behavior (filling out forms) and lurking users.
• Lecture topics: topics from the perspective of user pain points, topics from the perspective of course highlights, user interaction consultation, etc.
• Lecture rules: Questionnaires will be distributed in advance on the second day to collect user questions. If less than % of user questions are collected, private chat will be held again 1 hour before the start of the third day. Lecture questions should include some controversial topics from previous sessions. If the amount of interaction is less than %, such questions will be raised for guidance.
• Lecture goal: **% users participate in interaction.

If interested, the operator can design several sets of community lecture programs for different objects, different themes and different purposes.

Such things as what kind of display form and frequency on what channel to attract what kind of users to participate and achieve a specific goal are called operation plan planning.

Through many attempts and optimizations, we found that the lecture format with the best user retention effect was the question-and-answer format. We also summarized 10 topics that could trigger user discussions. Ultimately, this lecture plan was templated and fixed, and we guaranteed an 8% user interaction rate each time (indirect retention statistics).

4. Data-based community user operation model

The construction of the above life cycle model is a top-down analysis of the underlying logic of user operations:

The user definition is clarified (users with which behavioral data are judged to be in which stage of the life cycle), which operational strategies and corresponding plans can be adopted (to what extent the adopted strategies will affect the corresponding user data, and whether the optimized operational plans have improved data indicators).

After many iterations, we finally had a set of optimal operation plans and corresponding segmentation indicators at the moment.

Assume that in this team, the conversion rate of the first-purchase community is the first-level key indicator; the second-level indicator is the core factor that affects the first-level indicator, which converts the user behavior data indicator into a data indicator based on the group.

To achieve secondary indicators, operators can explore more appropriate strategies and plans. For managers, monitoring the completion of secondary segmentation indicators can predict and control the final conversion rate.

Phase 3: Exponential Growth Phase - Implementing User Segmentation Strategy

Based on the excellent ROI performance mentioned above, the product has higher and longer-term resource exposure in the main site, and also has more marketing budget for external placement. The project entered a new stage, with traffic and the number of new community users growing exponentially. However, at this time, the conversion rate plummeted for several consecutive weeks, and there seemed to be problems with the original operating mechanism.

Faced with a sudden and obvious decline in data, several assumptions are put forward:

• Due to the rapid growth of distribution channels and exposure, the accuracy of acquired users has decreased;
• Due to the rapid increase in the number of users, the operating staff was overloaded, the training time for new employees was too short, and the operational work was not done properly.

It is impossible to make a comparative evaluation of channels outside the site. First, pull the basic data of users obtained from the site and compare it with the basic data of previous users, including user age group, regional consumption, activity status, etc. The data results are not much different.

More guidance has also been given to the work efficiency of operations staff. , but old employees also failed to meet the target, and the data has been declining for two consecutive weeks.

When several hypotheses are not valid or have no obvious correlation, we finally return to the method used in the startup project: user interviews and user surveys.

On the one hand, during the operation process, we consult more active users to understand the reasons why they are unwilling to pay. On the other hand, we find directions within the group based on the questions that users actively ask and do a lot of research and testing.

It was finally discovered that the awareness among new users had declined.

As exposure increases, more users are attracted by the broader "enlightenment" and "English", but their understanding of the course IP and the company's brand is extremely low. However, our operating strategy still assumes that users have a cognitive basis for conversion.

After this stage, the community operation cycle made a new strategic adjustment, increasing from 3 days to 5 days. In the first two days, awareness promotion was added to emphasize the brand and teaching philosophy, and conversion began to be involved in the last 3 days.

This method was finally used many times among users acquired through off-site channels. Corresponding communities were established for users pulled in through each new channel, and survey questionnaires were set up in the middle of the operation cycle and at the end of the operation.

Generally, segmentation is done from several aspects: the age of users in this channel, the consumption capacity of users in this channel, the awareness of users in this channel about our courses, and their perception of the experience process.

For the user group with an older age group, add an entry for purchasing high-level courses; for the user group with a higher level of awareness, increase promotional methods such as group buying and person-to-person referrals, etc.

Author: Chen Ergou

Source: Chen Ergou

Related reading:

Perfect Diary’s community operation method!

6000-word detailed explanation of community operation