Data operation: How to analyze data on APP?

A friend told me that an Internet company he used to work for had a very weak ability to resist risks. The entire operations department was all in on adding new customers, with no consideration for indicators such as retention and activity.

In 2017, the number of new users coming from the app market through ASO alone reached an average of 30,000 per day, not including other channels. However, the retention rate is particularly low, and the 7-day active retention rate can only be maintained at around 10%.

Later, a new product manager came to the company. This product manager saw the company’s problems and gradually improved the data system of the entire company. Later, the operational data indicator system gradually became clear, and the company's user growth also entered a healthy growth state, which was much more sustainable than the new profits generated by all in at the time.

He lamented that if data analysis is done well, sustainable profit growth can be achieved, and he deeply felt the importance of data analysis.

I completely agree with his point of view that data analysis has great value potential.

Today, based on my many years of experience in APP data analysis, I will explain to you some ideas of APP data analysis. Remember, we only talk about ideas, not practical operations . I hope it will be helpful to some partners who are interested in APP data analysis.

Monitoring of daily data operation indicators

Daily data operation indicators, such as the number of downloading users, the number of new users, the number of active users, the number of paying users, etc., are the most basic and fundamental data in operations and are the core indicators that big bosses pay the most attention to.

These indicators have high requirements for data accuracy and timeliness, so once you enter a new company or take over a new project, your first task is to sort out these data.

In addition, many indicators in the operation indicator system are derived from these basic indicators. If the data quality of these basic indicators is not up to standard, other derived indicators will also be biased, and the deviation results will be greater than the basic indicators due to the superposition of errors of multiple basic indicators.

How to ensure the data quality of basic indicators?

The design of the user ID logic is critical . For user statistics, the quality of the data is directly determined by the design logic of the user ID.

Therefore, when you obtain these basic data, you need to have a clear understanding of the ID logic behind the statistics. For e-commerce and social apps, because this type of app has a powerful membership system, it will play a good complementary role in accurately identifying a user.

Channel Analysis

For an APP in its rising or declining stage, the operation team will try to find as many channels as possible to attract traffic and attract the attention of new users.

There are many channels on the Internet, usually bidding channels (Baidu, Sogou, app stores), SEO channels (Baidu, Sogou), new media channels (WeChat official accounts, Weibo, Douyin), alliance advertising channels (Baidu Alliance, Alimama), mobile payment channels (Toutiao, Tencent Guangdiantong), free channels (QQ groups, WeChat groups, Tieba, Q&A platforms, app stores), live broadcast platforms (Huya Live, Inke), etc.

There are so many channels, so monitoring and analyzing channel effects is very helpful in reducing customer acquisition costs and improving the ROI of channel promotion.

Channel analysis is nothing more than monitoring the quality of each channel, which one is more effective, and which one has a cheaper unit price. Of course, we also need to monitor the subsequent performance of users in each different channel and score the users of each channel. We must let the boss know clearly which channels are worth investing in and which channels are rubbish; which channels need to increase investment and which channels should be abandoned.

If the operation team has sufficient resources, it can also conduct comparative analysis on user quality between different mobile phone models, different operating systems, and different regions. In short, new users are sliced ​​in different dimensions to monitor user performance in different dimensions.

Of course, in channel analysis, there are two important issues that require urgent attention from marketers and data analysts, namely channel fraud and channel attribution. Channel fraud and channel attribution are both very complex research topics. I will write something separately about these two topics later, so I will not go into detail here.

Active user analysis

One product cannot satisfy all users. You cannot have your cake and eat it too. The reason why users become active users is that your product must have met certain user needs. Studying active users well will help us improve the most core functional points, so the behavior of this group of people is more worthy of study .

Therefore, active users (or core users) are the most valuable resources of the APP. We must pay close attention to the dynamics of the APP's active users and listen to their voices.

For active user analysis, we can focus on indicators such as DAU, WAU, MAU, number of launches, usage time, DAU/WAU, DAU/MAU, etc. WAU and MAU reflect the total scale of active users, number of launches and usage time reflect the stickiness of active users, and DAU/WAU and DAU/MAU reflect the activity of active users.

In the analysis of active users, indicators that reflect stickiness and activity are worthy of detailed study. For example, take the usage time indicator. This indicator is the time users spend on the APP within a natural period of time. The biggest function of this indicator is to evaluate user activity and user stickiness.

If the user usage time is very ideal, it means that the user has a high degree of recognition and rigid demand for the APP, and vice versa.

On the other hand, think about how much time a normal user was expected to spend on your app when you designed it. After it goes online, is the actual time users spend the same as you expected?

If there is a big deviation here, it means that the user's perception of the APP is different from what you imagined at the time. At this time, you need to think about how to adjust your product to cater to user perception.

User portrait analysis

User portrait is actually the labeling of user information. Such as gender, age, mobile phone model, network model, occupation income, interest preferences, etc. The core work of user portrait analysis is to label users according to labeling rules formulated by humans, so that the information in the labels can be quickly read out, and finally the labels can be extracted and aggregated to form user portraits.

There are two main application scenarios of user portraits: user feature analysis and user segmentation.

User feature analysis is a continuous and in-depth insight into the user attributes of a specific user group, which makes the portrait of the user group gradually clearer and helps companies understand who they are? What are the behavioral characteristics? What are the preferences? What are the underlying needs and behavioral preferences? After gaining insight into these characteristics, targeted analysis can be performed for subsequent user groups.

User segmentation is the basis of refined operations and has been widely used in data analysis processes in various industries. For example, it can help companies achieve precision marketing by identifying target marketing groups; it can help companies achieve precise push notifications to awaken dormant users or recall lost users; and it can help companies achieve personalized content recommendations for e-commerce or information apps, etc.

Product core function conversion analysis

What is a conversion?

When a user performs an action in the direction of your business value point, a conversion occurs. The business value points here include but are not limited to completing registration, downloading, purchasing and other behaviors. In the analysis field of Internet products and operations, conversion analysis is the most core and critical scenario.

Taking shopping on e-commerce websites as an example, a successful purchase behavior involves multiple links such as searching, browsing, adding to shopping cart, modifying orders, settlement, and payment. Problems in any link may lead to the failure of the user's final purchase behavior. In the context of refined operations, it is very important to conduct conversion analysis.

So, when you want to do conversion analysis, think about what the core function of your product is, and then monitor the conversion rate of this core function. Different industries have corresponding conversion rates. For example, game apps focus more on payment rates, while e-commerce apps focus more on purchase rates.

Conversion rate analysis, you can also compare your products with the industry average to see where your products stand in the industry. In addition, you can also evaluate the pros and cons of different versions of the APP through long-term trend monitoring.

User churn analysis

Recalling lost users is an important part of operations work, and defining lost users is the starting point of user churn analysis. Lost users usually refer to those users who have used a product or service but no longer use it for some reason.

In actual work, the definition of lost users is much more complicated for different product or service business types.

• For example, for e-commerce products, based on the definition of user purchasing behavior, a user is considered lost if they do not purchase again for a certain period of time;
• For example, for content products, based on the definition of user access behavior, a user is considered lost if they have not visited for a certain period of time.
• For example, for video products, based on the definition of user viewing behavior, a user is considered lost if they have not watched for a certain period of time.

Therefore, it is necessary to quantify the key user behaviors based on the product business type to define lost users.

User churn is a process, not a node . Before churned users officially stop using the product, they will show some abnormal behavioral characteristics: a significant decrease in visit frequency, a significant decrease in online time, a significant decrease in interaction frequency, etc.

Therefore, we need to establish a user churn warning mechanism through rules or machine learning modeling, predict the probability of churned users in advance, and support operations to intervene in activities for users with high potential for churn.

If conditions permit, you can compare it with the industry average to make yourself more aware of the position of your product's churn rate in the industry. In addition, we can also create profiles for lost users , which can help us better understand the characteristics of lost users. The more detailed and representative the portrait of lost users is, the higher the recall success rate will be.

However, we know that lost users and profiles of lost users are not enough. We also need to find out where the users are lost , see where the users are lost, and then make targeted changes to the products accordingly.

Once we have clearly defined churned users, understood their profiles, and know which channels they are concentrated in, we will then need to define the path and strategy for user recall.

From the user's perspective, give users a reason to use the product again. Recalling lost users is not the end. We need to maintain and reactivate the recalled lost users to consolidate the recall effect.

User life cycle analysis

What is the life cycle of an APP user?

It refers to the entire development process from the time a user establishes a relationship with an APP to the time they completely break up with the APP. The total value that a user brings to an APP during the entire life cycle is called life cycle value.

During the entire life cycle of an APP user, from the perspective of user value contribution, it can be divided into four different periods, namely, the exploration period, the formation period, the stable period and the decline period. Users in each period bring different values ​​to the APP.

(1) Probation period

At this time, users mainly verify and examine their own needs on the functions and services provided by the APP products. Once users find that the product cannot meet their needs, they will quickly leave.

Therefore, when planning a product, it is necessary to accurately identify the target group and target user needs, and try to avoid large-scale user loss after the product goes online. During this period, users' value contribution is relatively low.

(2) Formation period

When the functions and services of a product can meet the needs of users, users will use the product tentatively, and the user experience of the product will play a decisive role in this process. Especially when there are many homogeneous apps, users will overwhelmingly choose apps with better experience.

During this period, users will truly choose and decide to use the product, and the value created by users will also increase rapidly.

(3) Stable period

Users in this period have the highest loyalty and activity levels. They will use the product frequently, promote the product through word-of-mouth, and attract and recommend more users to choose the product. User value creation during this period will reach its highest level and remain stable for a long time.

(4) Degeneration stage

There are many factors that cause stable users to enter the degradation stage. For example, if the child grows up, he or she will stop using the product.

In short, certain factors that affect user satisfaction may cause users to enter the degeneration stage and then completely leave the product. Once a user enters the degradation period, timely user maintenance should be carried out. At this stage, the value created by users will decrease rapidly.

Summarize

The APP data analysis ideas summarized above are not comprehensive. For example, commonly used analysis ideas such as A/B testing, heat map analysis, form analysis, and path analysis are not included. There are so many APP data analysis ideas. In fact, there are already very mature APP data analysis tools on the market, which provide us with powerful analysis support.

For example, domestic ones include Umeng, MTA, Talkingdata, Sensors Data, Growingio, Zhuge IO, and Shugeek, and foreign ones include GA, Mixpannel, Appsee, etc. Every APP data analysis tool is almost the same except for the basic data analysis dimensions, and each product has its own unique advantages.

Therefore, if you want to choose a third-party data analysis tool, you should choose a data analysis tool that suits you based on your analysis purpose and your company's conditions.

Author: Nobita

Source: Big Bear