Three steps to build a cyclic process of user operation data

Process-based and refined thinking are the basic thinking that every operator must possess. If process-based thinking is the operator's qualitative thinking about the operational goals, then data-based thinking is the quantitative description of the path and effect of achieving this goal. It implements your work ideas on specific data indicators to measure your work results and goal achievement.

The necessity of establishing data-based user operations lies in, firstly, quantitatively measuring the value of your work, and secondly, it is the basis for achieving refined operations. For example, the user stratification classification and user portraits based on data mentioned later are the prerequisites for refined operations.

Data-driven user operation utilizes the idea of ​​user operation, combines it with the idea of ​​data analysis , uses business to guide data, and uses data to drive business, to achieve refined operation of users. This is the core idea of ​​data-driven user operation. The cyclical process of user operation dataization is as follows: user data collection - building a user data-based operation indicator system - data-driven operation.

1. User Data Collection

The collection of user data mainly includes user basic data, user behavior data and user traffic data.

• User basic data refers to the user's static data, including gender, age, region, job, etc. This type of data describes who the user is and is mainly achieved by filling in basic information.
• User behavior data is a collection of a series of user operations on the product, including which user completed which type of operation at which point in time, in which place, and in what way. It includes user browsing, purchasing, content contribution, invitation dissemination, social networking and other behaviors. This type of data describes what the user has done and is mainly achieved through data embedding.
• User traffic data is the source of users and is generated based on the web pages visited by users, including devices, operators, ports, time, etc. This type of data describes where the users come from. However, the current traffic data statistics mainly come from third-party tools such as GA and Baidu Statistics, and cannot be recorded in the database, which means that they cannot be one-to-one corresponded with the basic user data and behavior data mentioned above.

The above data are all raw data obtained from products or third-party tools. To achieve operational goals, it is necessary to do data mining and data analysis based on the raw data, and build a data-based operational indicator system based on operational goals and paths.

2. Build a user data-based operation indicator system

If you can’t describe your business with metrics, then you can’t grow it effectively. So what you need to do in this step is to indicatorize your business. Data indicators are not constant. They rely on the business process or functional process of your product and are closely related to the goals and the path to achieve them.

The purpose of user operation is to maximize user value. If you are an e-commerce product, then your purpose is to get users to pay for products. If you are a community product, then your purpose is to get users to contribute and disseminate content. However, the realization of product goals and user value is a gradual process and also a process of dynamic evolution. Some potential users register to become active users, some turn from active to lost users, and some return from lost to active users.

In the above picture, orange represents the dynamic evolution of user status, and red represents the operation goal. Following the operational idea of ​​goal-path-effect, data analysis is to break down your goals and express them in specific data indicators as core evaluation indicators, use data to monitor the path to achieve the goals to evaluate the results, and compare with the core evaluation indicators set up at the beginning to judge, verify, correct and optimize the work path to achieve better and faster results. Based on this idea, we construct the following data-based operation indicator system, each of which contains a series of related indicators. The construction of the indicator system is achieved through data processing and processing based on the user data collected in the first part.

1. In the stage of attracting new users from potential users to registered users, we need to analyze the channels for attracting new users and the promotion strategies adopted in each channel, evaluate the channel quality through data indicators, and optimize the channel promotion strategy. The data indicators mainly include the number of new users, user acquisition cost, and new user retention rate .

• New users: New users refer to users who launch the app for the first time after installing the app. According to the statistical time span, new users are divided into daily, weekly and monthly. The new user volume indicator is mainly the most basic indicator to measure the effectiveness of marketing promotion channels ;
• User acquisition cost: For paid channels, it reflects the conversion rate of the channel.
• New user retention rate: reflects the quality of new users and their fit with target users . In addition, for mature versions of products, if the user retention rate changes significantly, it means that the user quality has changed significantly, which is most likely caused by changes in the quality of promotion channels.

Channel A: SEM

Channel B: Weibo

2. Promoting activation and retention of registered users and active users is one of the most important tasks of operations personnel. Our daily work on user stratification and classification, user growth incentive system, etc. are all done in this link. In terms of data, the indicator system we can set up includes a system to understand user scale and quality, a system to understand user engagement (depth of usage), and a user portrait system to understand user attributes.

(1) User scale and quality

• Active user indicator: Active users refer to the number of devices that have launched an application (APP) within a certain statistical period. Active users are an indicator of the user scale of an application. Usually, if we only look at one indicator to determine whether a product is successful, then that indicator must be the number of active users. The number of active users can be divided into daily active users (DAU), weekly active users (WAU), and monthly active users (MAU) according to different statistical periods. The statistical periods for different product categories are also different.
• New user indicator: As mentioned earlier, the new user volume indicator is the main indicator for measuring the effectiveness of promotion channels; in addition, the proportion of new users to active users can also be used to measure product health. Pay special attention to retention rate when the ratio is too high.
• User retention rate indicator: User retention rate refers to the proportion of users who still launch the application after a period of time among the number of new users in a certain statistical period. User retention rate can focus on the next day, 7 days, 14 days and 30 days retention rate. The retention rate reflects the quality of users on the one hand, and the attractiveness of the product on the other. When the retention rate is abnormal, the reasons can be found in these two aspects.
• User composition indicators: User composition is an analysis of the composition of active users within the statistical period. Taking weekly active users as an example, weekly active users include returning users this week, users who have been active for n consecutive weeks, loyal users, etc. This helps to understand the health of active users through the structure of new and old users.
• Single user active days indicator: the average number of days each user is active in the app during the statistical period. If the statistical period is relatively long, such as more than one year, then the total active days of each user can basically reflect the number of days the user spent on the APP before churn. This is a very important indicator reflecting user quality, especially user activity.

(2) User engagement

The user engagement system is an important indicator system for measuring user activity. The definition of activity is different in different products. For example, activity in e-commerce products can be defined as purchases, and activity in community products can be defined as content contribution. Therefore, the following three indicators can evolve differently in different products.

Number of launches = number of purchases = number of content contributions;

Last use = last consumption = last content contribution;

Usage time = consumption amount = content contribution;

Usage interval = purchase frequency = content contribution frequency.

• Number of launches: refers to the number of times a user launches an application within a certain statistical period. When conducting data analysis, on the one hand, we need to pay attention to the total trend of the number of startup times, and on the other hand, we need to pay attention to the average number of startups per person, that is, the ratio of the number of startups to the number of active users in the same statistical period. For example, the average daily number of startups per person is the ratio of the daily number of startups to the number of daily active users, which reflects the average number of startups per user per day.
• Last used: The time from the last time the user used the app to now. Through dimension and distribution analysis, it can also reflect the activity to a certain extent.
• Usage duration: refers to the total duration from the start of the APP to the end of its use within a certain statistical period. The usage time can also be analyzed from the perspectives of average usage time per person (the ratio of total usage time to the number of active users), single usage time (total usage time and number of launches), etc. It is an important indicator for measuring product activity and product quality.
• Usage time interval: The usage time interval refers to the time interval between two adjacent activations by the same user. We usually need to analyze the distribution of usage time intervals, generally counting the distribution of the number of active users of the application within a month. It is also possible to discover user experience problems through the differences in the distribution of usage time intervals in different statistical periods (different time points but the same span).
• Visited pages: The number of visited pages refers to the number of pages that a user accesses at one time. We usually need to analyze the distribution of the number of visited pages, that is, to count the distribution of the number of active users of the number of visited pages of the application within a certain period (such as 1 day, 7 days or 30 days) , such as the number of active users visiting 1-2 pages, the number of active users visiting 3-5 pages, the number of active users visiting 6-9 pages, the number of active users visiting 10-29 pages, the number of active users visiting 30-50 pages, and the number of active users visiting more than 50 pages. At the same time, we can discover user experience problems by looking at the differences in the distribution of visited pages in different statistical periods (but with the same statistical span, such as 7 days).

Among the above user engagement indicators, we can select an indicator that can reflect the main operating goals, such as consumption amount, to build a user level model (user stratification), or we can select multiple related indicators, such as the last consumption time R, consumption frequency F, and consumption amount M to build a commonly used RFM user model.

Its function is to formulate targeted operation strategies or develop a user incentive system for conversion between levels based on the characteristics of users at different levels (user stratification) or in different regions (RFM model) in the constructed model.

Taking the question-and-answer community as an example, the main KPI is the quantity and quality of content, which is reflected in the number of recognitions obtained by the content contributed by users. Through data collection and collation, the distribution of user recognition numbers is as follows. We establish user stratification based on the number of user recognitions.

It can be seen that the distribution is similar to the logarithmic normal distribution. By defining the first, second, and third quartiles as critical values ​​through a similar distribution histogram, users are divided into four levels: ordinary users, content producers, content contributors, and big Vs.

When the number of users is large enough, the user characteristics in each user level also show great differences. For example, in the first layer of content contributors, some people mainly publish articles, with a low frequency and a high number of recognitions per article; some people mainly ask questions, with a high frequency and a low number of recognitions per article. This can be combined with the RFM model to further segment the users in each layer.

For example, some people have less than 3 years of experience, some have more than 5 years of experience, some like social content, and some like e-commerce content. In this way, we can combine the user portraits introduced below to make a more detailed description of the user's attributes and achieve a more refined operation effect.

RFM Model

(3) User Profile

User portraits are created by outlining the user's profile through various data. Any indicator that can define user attributes can be included in the user portrait, including gender, age, education, income, expenditure, occupation, industry, personal interests and hobbies, business interests, social relationships, etc. The more data there is, the clearer the user's profile will be, and the more targeted the corresponding operational strategy will be.

3. Every day we receive various text messages, push notifications, and unfamiliar phone calls on our mobile phones. We often receive advertising emails in our mailboxes, which hit your point more and more accurately, prompting you to launch the app again. If you haven’t used this app for a long time, then this is most likely a recall measure taken by the operator based on data analysis, trying to win back lost users.

This stage mainly involves analyzing the reasons for loss and formulating corresponding recall plans, and data indicators are used to measure work effectiveness. It is reflected in the data indicators as a churn and recall system, including churn rate, arrival rate, open rate, open click rate , and return rate.

Churn rate: Churn rate and retention rate are a pair of concepts that complement each other. They are the ratio of users who no longer use the product after a certain statistical period. Both indicators are generally calculated using the same group method, but because the churn rate has a certain lag, the churn rate is usually estimated by querying the retention rate.

• Reach rate: The ratio of push notifications reaching the user’s mobile phone or email address.
• Open rate: The rate at which users open a push notification.
• Open click rate: the ratio of users clicking on content/links after opening the website.
• Return rate: the ratio of the number of returning users to the number of lost users during the statistical period.

Our goal is to bring back lost users, but it cannot be achieved overnight. The following four indicators are progressive and form a conversion funnel. The form of push, push sending time, push title , whether the sender is official, whether the sending target is accurate, whether the actual content is consistent with the title and even the page layout will affect the conversion at each level.

3. Data-driven user operations

Although we have a structured data indicator system, this cannot be considered a complete operating system. Data itself has no value. It only becomes valuable when it is transformed into a strategy. The data indicators we construct are all for decision-making purposes, helping us formulate and optimize operational strategies.

Through data, we not only want to know "what" and "how much", but more importantly, we want to know "why"? This is the key to data-driven business. Data-driven business is reflected in two aspects:

One is to use data to optimize operational strategies . For example, the user retention rate is low, which is related to user quality and product attractiveness. Channel analysis shows that there is no problem with user quality, and user churn analysis shows that the main churn stage is in the initial contact period. This finds the reason, so we optimize product stability, ease of use, and new user guidance.

The second is to verify the operation strategy with data . For example, if you want to launch a new user incentive measure, but are not sure whether it will have better results compared to the original method, the comparative data obtained through reasonable AB testing can provide you with a basis for decision-making.

Data analysis to find causes and operational strategy optimization are iterative processes. Let’s take user churn prevention as an example to illustrate.

The core of preventing user churn is to reduce the user churn rate or extend the user life cycle (when churn is unavoidable). The reasons for user churn vary greatly. Some are that a large number of low-value users are attracted during the promotion process, some are that users are not interested in the product, and some are that the excitement level continues to increase while the interest continues to decrease during use. Only by identifying the reasons for user churn can we launch effective strategies to prevent user churn and recall lost users, and this all requires relying on the use of a data indicator system to explain the problem.

1. Analysis of user churn rate in different channels

• Channel A: SEM

• Channel B: Weibo

• All site users

By analyzing the churn of users from different channels, we can see that the churn rates of users from different channels are significantly different, and are also different from the churn rates of users from the full stack.

In the first week, the churn rate of users who signed up for the product through SEM in Channel A was 34%, while the churn rate of users who signed up for the product through Weibo in Channel B was 54%. Why is the user situation of channel A significantly better than that of channel B? Whether it is active keyword search on channel A or interested click on promotional link on channel B, users' needs are basically the same, otherwise they will be lost during the registration stage.

Further analysis shows that because SEM is a paid promotion and Weibo is free natural traffic, in order to improve the input-output ratio of SEM, users of channel A will have a special landing page introducing the product after entering the product page, while Weibo directly links to the activity page. Users have poor knowledge of the product, resulting in increased user churn.

Operations can add a new user guidance function for users coming through Weibo and continue to observe the data of new users coming from Weibo. Repeat this process to continuously optimize your strategy.

2. Analysis of user churn rate in different life cycles

The user life cycle is divided into the contact adaptation period, exploration growth period, mature stability period, and decline period by using the two indicators of usage duration and usage frequency. The number of users and churn rate in different periods are as follows:

As can be seen from the above table, the churn rate is relatively high during the exploration and adaptation period when users first come into contact with the product, and there is much room for improvement. According to our user operation experience, the reasons for user loss at this stage generally include poor onboarding experience, slow access speed, high learning cost, content mismatch, etc., which will be reflected in the data. We can find the reasons through analysis of the corresponding data, formulate relevant user strategies, continue to observe the data, and repeatedly optimize the strategies.

The above example of preventing user churn is used to illustrate how to use data to drive user operations. The other links are similar. Use the idea of ​​user operation, combine it with the idea of ​​data analysis, select a reasonable data indicator system, accurately analyze the reasons, formulate corresponding strategies, and re-observe the data optimization strategy.

IV. Conclusion

1. The process of building a data-based user operation system is: user data collection - building a user data-based operation indicator system - data-driven operation.
2. Establish a structured data indicator system based on the user's product cycle and digitize user operation work.
3. Data drives the business, formulates corresponding operation strategies through data search and continuously optimizes by continuing to observe the data.

Mobile application product promotion service: APP promotion service Qinggua Media advertising

This article was compiled and published by @章鱼怕黑(Qinggua Media). Please indicate the author information and source when reprinting! Site Map