How to conduct big data analysis and capture the fickle hearts of users?

I originally thought that "homebodies" were heavy users of game and video apps, but I didn't expect that they would spend a lot of time on learning and fitness apps. Especially for learning apps, not only did the usage time increase first, but the total usage time also ranked in the top three. The results of this report are very different from what everyone thinks, and it made me start to think about whether the user base I am operating has also changed?

In an era where user stock is king, as an operations dog, only by truly understanding users and keeping up with the times to complete KPIs can you avoid being led by the dog. At this time, data analysis becomes our secret code. If we practice it well, we can plan "star" content that will gain both good reputation and traffic . But there are thousands of books, which ones do I really need? How should we use the data correctly?

There is so much data, what should I do if I want all of it?

User data is massive, and it is impractical to analyze all of it, so it is necessary to classify the data from different dimensions. In my opinion, it can be divided into two major categories: basic data and personalized data.

Basic data is data that every APP operator needs to have a clear understanding of, such as the gender ratio of users, age composition, user activity, etc. These data are the basis for operational work. If you don’t understand these data yet, please stop what you are doing and do the new employee training again.

Personalized data is targeted data, which is extracted and screened according to different user scenarios or operational needs. Take the daily activity operation of APP users as an example:

In the early planning, the user's group portrait can guide the planning direction of the activity, and the user's needs determine the goal of the activity; by understanding the user's interests, the content and presentation of the activity can be determined; by understanding the consistency of user behavior , the timing of the activity promotion can be determined.

During operations, through detailed event statistics and customized tracking points, we can further analyze user behavior in activities, understand the data conversion status of each link in the entire activity, and then optimize the activities and adjust the investment in activities based on the data feedback.

At the end of the event, you can analyze the number of new users, active users, retention users , and even uninstall users to evaluate the effectiveness of the entire event and provide valuable data comparison references for the next event.

Therefore, as refined operations become more and more important, the statistics, analysis and application of personalized data are the core capabilities of data operations and will also become the key to successful operations.

Operations require a long-term commitment. How can we capture the fickle hearts of users?

Users are fickle, and if we don’t know what they want, how can we expect to have a long-lasting relationship with them? The data reflects the results of a single dimension. How to combine this data into a true portrait of the user, analyze it in an integrated manner, and truly understand the user will test the operations staff's ability to apply data.

• User data requires a multi-dimensional combination (picture from the Internet) -

First of all, the data that constitutes the user portrait can be divided into attribute data, behavior data and scenario data.

• Attribute data reflects the objective attributes of users , that is, data that will not change over a long period of time, such as gender, age group, consumption level, etc.
• Behavioral data reflects the user’s recent behavior, such as the applications the user has recently liked and the scenes the user has recently visited.
• Scene data reflects the user’s real-time scene. By using LBS geo-fencing technology, the user's current scene is determined based on the user's geographic location.

When these three data are used organically, hundreds of user tags can be formed, which can truly visualize the thousands of faces of each user and make it easier for operators to carry out refined user operations . Here I would like to recommend the user analysis tool “Personal Image” that I often use for Getui. Personal Image can help me analyze users' online and offline behavior data, and form a very complete and accurate user portrait through dozens of attribute tags and hundreds of interest tags on the Personal Image platform.

-User label system for “personal image” –

These rich user tags can help me find the target user group more accurately. For example, when promoting a movie , accurate data operations are very helpful for distribution strategies. Users who like to watch "Gang Rinpoche" have certain common characteristics, such as heavy users of movie apps like to write movie reviews or prefer to use literary apps. Then we can use data analysis to tap into this group of literary and artistic users, interact with them, and use them to drive a larger audience market.

The concept we want to focus on here is the user's recent behavior data. It can reflect the user's growth cycle, the user's interest shift, etc., and is particularly important for content operations . For example, travel apps can use users’ recent behavior data to understand the travel scenes users have visited recently and avoid duplicate recommendations; they can also understand users’ recent behavior preferences and recommend suitable travel content from the perspective of users’ interests.

No comparison, no harm. Let the data speak the truth?

Mining the content of data is a technical job. For operations, the most basic data analysis is data comparison. Only with comparison can the truth (shang) be revealed (hai). There are two types of data that operators need to carefully analyze: one is the APP's own data, that is, the data generated by users when using the APP, such as browsing data of pages within the APP, consumption data, etc.; the other is APP's external data, such as industry public data, research data, etc.

In the analysis of the APP’s own data, we can make “fancy” comparisons by adding time points, link points, comparing data, and other methods.

Taking marketing activities as an example, we should not only look at the final sales data, but also establish points throughout the entire marketing process and count the conversion status of each link. For example, the opening of the marketing campaign page, the clicking of the product introduction page, the clicking of adding to the shopping cart, etc. There will be conversions and losses in every link of the entire marketing campaign, but the key question that operators really need to ask is at which link the most users are lost.

• Pay attention to the progress and transformation of events at each stage –

It is difficult for many companies to conduct comparative analysis of external data independently. They often lack large-scale data coverage and industry trend comparisons. At this time, it is necessary to seek help from third-party data service providers.

It is understood that some third-party big data service providers that are at the top of the industry are able to help companies conduct more comprehensive data analysis through the massive data accumulated over the years and powerful data analysis capabilities. A few days ago, I was attracted by the application data statistics and analysis product "Number". What attracts me most about the number is that it can provide unique data analysis services such as industry comparison and uninstall analysis, which is very valuable for optimizing operations.

The industry comparison index can help operators understand the overall development of the market, the industry competitiveness of APPs, and the development stage of their own APPs, and guide the operators' decision-making.

The application scenarios of uninstall user analysis are more targeted: 1. Compare customer acquisition and churn data to assist in determining the product life cycle; 2. Analyze the uninstall rate of users from various source channels and optimize advertising strategies; 3. Combine customized embedding points to deeply explore the characteristics of uninstall users and analyze the reasons for uninstallation; 4. During the event, correlate and analyze uninstall data to evaluate the negative impact of the event on users.

• “Number” of uninstall user flow display –

Fully interpreting the data and exploring the value behind the data can provide more objective feedback for operational work and effectively avoid human cognitive bias.

To sum up, under the trend of refined operations, we increasingly need to "recognize" the true appearance of users, and the rational and effective use of data has become a skill that must be acquired and upgraded. Only by using the right methods can we understand users more deeply and provide new ideas for operations.

Author: Xia Operation , authorized to publish by Qinggua Media .

Source: Shrimp Operation