In the operation of Internet products , daily active data must be reviewed every day. How can it be combined with user behavior to help business growth? The daily active users have increased or decreased. How can we quickly identify the cause of the abnormal fluctuation in daily active users? These are things that require daily observation, thinking, and comprehensive analysis. Especially when daily active users are declining, it often takes a long time to locate them.

The author of this article introduces in detail a method to quickly locate daily activity fluctuations for everyone's reference and learning. I hope to help you quickly identify the reasons for fluctuations in daily active users, and focus on the main problems of current business to iterate and improve them in a targeted manner.

1. Composition of daily active users

New and active users are a common way to distinguish daily active users, mainly to measure the data performance of new users and overall daily active users. Because active users include new users, when the daily active users fluctuate, we will have a headache about which user group has problems.

From 2019 to 2020, the actual daily active users showed an overall steady downward trend, and business growth entered a bottleneck period. Our ideal daily active users should be rising smoothly amidst ups and downs, so something must have happened here and we need to find it out layer by layer (bugs in data services, data reporting, and data statistics are not considered here).

As shown in the figure, we have broken down daily active users into three levels. The first level is the first breakdown according to the user's life cycle, namely new users, new customer retention, old customer retention, and lost and returning users; the second level needs to consider what we should focus on first if we break down each life cycle stage; the third level must be associated with the user's core behavior, and indicators related to core values ​​can truly reflect user activity and the satisfaction of user needs.

Daily Activity Analysis Level 1

At the first level, if we want to accurately locate which user groups affect the fluctuations in daily active users, we need to ensure that the four levels are completely independent and do not interfere with each other, which involves the definition of these four user groups.

1. New users: The number of new users added on that day
2. New customer retention: The number of users who were newly added yesterday and activated the next day
3. Old customer retention: the number of old users who were active yesterday and activated the next day
4. Lost and returned users: users who were inactive yesterday but are active today

After the user groups are split and automated, we can identify the issues of new users, new customer retention, old customer retention, or lost and returning users from the daily fluctuations in daily active users, and then quickly break them down into the hands of students responsible for new or old customer operations.

Daily Activity Disassembly Level 2

The first step is to identify who has the problem, and the second step is to solve the problem of how to do it.

(1) New users and new customer retention

In line with the strategies of new promotions and new operations on the operational side, a secondary analysis of the new additions is conducted from the channel level to locate which promotion channel has problems and the next-day retention rate performance of each channel.

There is an important issue here that is easy to overlook. If the next-day retention is done well, it will have a great impact on the product's old customers.

The user retention rate curve follows the law of the power function. The higher the retention rate on the next day, the higher the retention rate thereafter. The easiest way to start with the highest return on investment is to increase the retention rate of new users. Once the retention rate of new users increases, the next-day retention rate of subsequent active old users will naturally increase.

(2) Retention of old customers

Because users are divided into daily levels, the old customers here have the problem of user freshness. We can divide old customers into active old customers, silent old customers, and lost old customers.

1. Active old customers: users who have opened the app in the past 7 days
2. Silent old customers: users who have not opened the app in the past 7 days (7 days < last active time < 30 days)
3. Lost old customers: users who have not opened the app in the past 30 days

The next-day retention of active old customers is related to the user's continued activity ability, and the subsequent retention ability of silent and lost old customers after they come back, all of which determine the subsequent changes in weekly and monthly person-days. Increasing the frequency of old customers is the most important goal of old customer operations, so it is recommended to establish a cohort analysis based on user freshness, and observe the subsequent retention performance of various user groups on the next day, 3 days, 7 days, 15 days, and 30 days.

(3) Lost and returning users

If the lost and returned users are split on a daily basis, they account for the largest proportion of daily active users and are also the group that most needs refined operations. According to the user return cycle, they can be further divided into active return, silent return and lost return.

As shown above, the next-day retention rate of each part of the lost and returned users and the retention of old customers complement each other and affect the active frequency of weekly person-days, monthly person-days, etc. The content operation side should pay close attention to the trend of data changes and keep up with the strategy. Observe whether the user activity retention data improves with each strategy change, or whether it is just a temporary shot in the arm.

In addition, the user's return method also reflects the user's true intention of returning. Whether they opened the app on their own due to real demand, or returned through personalized or automated Push methods, or were externally triggered by other third-party cooperation, we need to treat them differently.

Dismantling the third level of daily activity

The third layer returns to the most core user behavior that the product provides to users. Daily activity is actually a vanity indicator. Indicators related to core values ​​are the ones that can truly reflect user activity. If the core behavior improves, there is no need to worry about the growth of daily activity. Focus on increasing the proportion of users using core behaviors, improving the engagement of core behavior users, and closely monitor the changing trends of high, medium, and low frequency users of core functions and users without core behaviors.

One thing to note here is that we can classify a core function into high, medium, low, and none according to the frequency of user usage. This way we can better quantify the changes in user usage of core functions, because a product may have more than one core function, and it is impossible for users to use only one core function. There will be overlap, and after the overlap, the specific values ​​that affect the fluctuations cannot be quantified simply from the data, so it is necessary to split it according to the frequency of users using the core behavior.

In summary, in order to promote growth, efforts must be made in four aspects: increasing new users, improving new user retention, activating old users, and recalling lost users.

2. Identifying the Cause of Abnormal Daily Activity Fluctuations

After confirming the accuracy of the data, extend the time range to observe the daily activity fluctuation trend, observe whether there are any abnormalities in the data, and eliminate daily cyclical fluctuations. If there is a sharp drop or rise or a continuous downward trend that deviates from the daily cyclical fluctuation trend (this can be implemented simultaneously with the abnormal alarm strategy), these nodes need to be closely monitored to investigate the specific reasons.

At this time, a quantitative indicator is needed to quantify the impact of a certain factor on the daily activity fluctuation, which can be calculated using the impact coefficient indicator:

Impact coefficient = (today's volume - yesterday's volume) / (today's total volume - yesterday's total volume)

The larger the result, the greater the impact of this factor on daily activity fluctuations, and it needs to be paid attention to and optimized first.

The amount of lost traffic continues to decrease, and there was a sharp drop on May 17. Let's continue to look down~

The silent, active, and churned user groups are all declining, but the decline in silent users is the largest. Among silent returning users, is it the number of users who open the App on their own that has decreased, or is it the result of Push or external awakening? Different reflow methods reveal different problems, just like peeling an onion, just peel off the layers...

3. Looking for opportunities from the long-term daily activity fluctuation trend

Extend the time range of the daily activity fluctuation data of each layer to observe the trend changes and see whether the product is developing in a good or bad direction.

Due to the limitation of space, I will not elaborate here. I may elaborate on the points later, so please stay tuned~

Author: Polaris

Source: Polaris

Related reading:

Product Operation: Competitive Product Analysis Methodology

Analysis of Internet Financial Product Operation Strategy