The loss of a large number of users brings great challenges to our products. Why are users lost? Where did it go after it was lost? How to retain existing users? How to "recall" lost users? This article takes "Research on Lost Users of Mobile QQ Music Player" as an example to talk about my own understanding for your reference.
1. First clarify the relevant terms
Some terms have been defined within the team before. If there are no mistakes, they just need to be understood and used. Some terms need to be defined for the first time when the team comes into contact with them, and only after they are unanimously agreed upon by the team can follow-up work be carried out.
Taking the mobile phone QQ music player as an example, the following terms were defined before product testing:
The definitions of two terms commonly used in churn user research are as follows:
Users: Anyone who has used a product can be called a user. The focus is on what dimensions to use to divide these users. Solving the "dimension" problem can help us conduct sampling research. Generally, it can include age, gender, region, occupation, income Such as demographic attributes and the relationship between people and products, that is, whether the product has been used, length of use, frequency of use and other dimensions to divide;
Lost users: Determine what lost users are and the dimensions of lost users. There are generally two ways to define lost users: objective dimension definition—time, login frequency, number of logins; subjective dimension—users no longer intend to use the product. The dimensional division of lost users should still be considered from two aspects: user attributes and the relationship between users and products. In the project of researching lost users of mobile QQ Music, the dimensions of lost users were divided from the following two aspects: 1. User attributes, including age, gender, education, etc.; 2. The relationship between users and products, including new users/old users , places of use, and other competing products used.
The above definitions will vary depending on the product and project team. For example: the mobile QQreader project team defines lost users as "not using the product for four months", while the mobile QQ Music project team defines "lost users" as "two months". The product has not been used", and the gaming group defined the time span as half a year. Just determine a rough time based on the characteristics of the product (online time, frequency of use, etc.). There is no need to waste time on this issue, because the purpose of the survey is to find out the shortcomings of the product and correct them. For the same product, The reasons for giving up two months ago are not much different from the reasons for giving up four months ago.
Other terms that will be involved include: historical users, returning users, retained users, etc.
2. The overall idea of the research
Who to study? (Determine the sample) → How to study? (Comprehensively find out the reasons for churn) →Are the survey results applicable to all churn users? (Quantitative verification) → Which results are more important? (Determine priorities) → How to improve the product? (Make suggestions).
Let’s talk about each stage in detail below.
1. Determine the research sample
If operations can provide relevant user behavior data and associate it with individual users, the target object can be found by simply filtering and clustering the data.
If there is no relevant data, we can only gradually narrow the scope and then launch a screening questionnaire for the smallest scope.
When doing research on lost users of the mobile QQ music player, since it was impossible to extract the user's QQ number, we could only narrow the scope to symbian and java platforms (at that time, the mobile QQ music player only had these two platform versions), so users of these two platforms were extracted QQ number for questionnaire survey, and gradually narrowed the scope through the screening questions in the questionnaire to find the research samples we wanted.
When doing this work, it is best to take a long-term view. The collected user data can also be used as user resources in other subsequent projects. Therefore, it is important to organize and classify users and establish a user database.
2. Qualitative mining
The purpose of qualitative mining is to comprehensively and in-depth find out all the reasons for user loss. Generally, the list of problems is found through background data analysis, questionnaires, interviews, tests, focus groups, etc. Whether we can accurately and comprehensively identify the problem depends on our selection of research objects and control of the operating process, both of which are often limited by research conditions.
When digging for causes, you should pay attention to three points: 1. The dependent variable and the independent variable must be related; 2. The cause must occur before the effect; 3. It is not a spurious correlation.
Tips: After user researchers determine the reasons for churn, it is best to invite several experts related to the project (can be product managers, interaction designers, product directors, etc.) to hold an expert forum. First, it can verify the reasons for churn you have discovered. Whether it is comprehensive and in place; second, everyone reaches a consensus on the phased results of the research to facilitate future work.
3. Quantitative verification
As long as the effective data exceeds a certain amount, you can conduct quantitative research. It has nothing to do with the research method you use. This "amount" is determined by the product characteristics and the content of your research, and has a lot to do with your sampling method. For QQ Users, because the total number of users is very large, if the sampling method is unreasonable, even if your sample size is 100,000, it will not be representative. After all, this is a process of accumulation of experience. For starters, when you're not sure what your sample size should be, grab what you can afford.
There are three problems solved by quantitative verification:
1. Are the reasons for loss discovered during the qualitative stage correct?
2. Are there any missing reasons for churn discovered during the qualitative stage?
3． The severity of the cause of churn;
It should be noted here that our determination of the severity of the problem may be based on the frequency of the problem, the number of users who responded to the problem, or the results determined by expert discussions, but these are not the order in which the project team solves the problem. , when it comes to the implementation stage, the problems encountered are difficult to predict.
4. Data analysis
We will not go into too much detail in this part about organizing and analyzing the obtained qualitative and quantitative data, because it is difficult to have a universal template and specifications, and we need to think of solutions to specific problems.
Don't be lazy when mining data, think more about the previous correlations of the data and the reasons behind each data. When I was doing research on lost users, the product manager put forward a requirement: Ask users whether they listen to music in the local and online song lists. When the music is playing, they prefer to switch the interface to the playback interface or listen directly on the list page. At that time, for This question has the following two questions in the questionnaire:
When you listen to music on your mobile phone, which way is more useful for selecting songs?
1. When listening to local songs:
A. Purposefully choose a certain song or type of song to listen to
B. Listen casually
C. The two situations are almost the same
2. When listening to online songs:
A. Purposefully choose a certain song or type of song to listen to
B. Listen casually
C. The two situations are almost the same
After the data was recovered, I made statistics, and finally made a habitual analysis of the data. Based on the correlation between the two, the product manager made a decision between "local" and "network" processing. .
5. Propose improvement strategies
Macro: Propose the overall strategy for the future development of the product;
Micro: Put forward suggestions for improvement of detailed experience issues;
Any suggestions must be constructive and should be directly related to the goals of the project team and the KPIs of the project team members.
This is what the project team is most concerned about, and it is also a demonstration of the hard work in the early stage. We must put in more effort. It is said that one minute on stage, ten years of hard work off stage, if this part of the content is not done wonderfully, our "ten years of hard work off stage" will be in vain.
3. General operating procedures
According to the research ideas, the specific operation process is summarized as follows:
Qualitative mining → data analysis → quantitative verification → data analysis → output report → review → follow up on product improvements.
Taking the research on lost users of mobile QQ music player as an example, the operation process I formulated is as follows:
Finally, the biggest insight I have gained during this period of work is: use the mentality of a product manager to do user research. The work requires constant thinking, summarizing, settling, and sharing. This is a continuous cycle process that requires persistence, resistance to loneliness, and Time needs to be cherished like gold...