Big Data Cloud ERA game analysis There are 4 major misunderstandings

In recent years, the popularity of the word big data and the rapid development of http://www.aliyun.com/zixun/aggregation/13568.html "> Large data technology, data analysis and data mining work has been more and more attention of enterprises, Especially in the game industry, more and more data on the analysis of the volume and perspective also blowout, "data-driven refinement operations," quantitative research of player groups "," mobile game Data Analysis System "and so on concepts, can be seen in the game development and operation of the data analysis needs are endless, Correspondingly, the discussion of the data analysis is also expressed. The loss of blindly relying too much on data, or subjective interpretation of data, is incalculable. Below, I will combine my work experience, the following data indicators will deceive and game data analysis of several misunderstandings.

1. No understanding of data definition, blind comparison of data

Peers are willing to pay attention to the industry of various types of game benchmark, such as the next day of the S-class game, S-Class game Arpu,s game Acu/pcu, and so on some common indicators of operation, in my opinion, comparison is one of the value of data, is used to measure the gap between the best measure of product And the data comparison is to establish the same data acquisition method and the method of calculating index. So before comparing data, please understand the benchmark standard and data acquisition method, so as to reflect the significance of data comparison.

2. Over-reliance on analytical methods, indulging in data modeling processes

In college, the author reads the statistics major, university participated in mathematical modeling competition to get a better position, but also done a series of such as BP Neural network, Bayesian decision tree or cluster analysis projects, in the first contact with the game data analysis, very excited, then used a variety of methods to analyze the data. Gradually I found that, in practical work, data analysis is not as rigorous as academic research, it needs to make a quick judgment on the performance of the data, do not need to verify that the sample group conforms to some statistical distribution before each analysis, and may not need to use "high-tech means" such as "artificial neural network" to predict the future user number of the product. Even give the "a>b" conclusion also do not need to do "significant test", the test is more of the understanding of the business ability to grasp. Therefore, in the process of data analysis, do not rely too much on the analysis method, but should attach importance to the game business grasp.

3. Data is objective, do not read data subjectively

For peers who have worked on the frontline for some time, data analysis often goes into a vicious circle, in our data extraction process, we will see some of the data performance, and a variety of phenomena have some of their own understanding of the conclusion, under the guidance of such thinking, there is always a way to use data to verify their own conclusions. In my opinion, the data is objective, the interpretation of data also need to uphold an objective and neutral attitude, we must avoid to interpret a data for their own point of view.

4. Not clear data analysis purposes, fuzzy analysis requirements, analysis is incomplete, should do a 300% analysis report

Clarify the purpose and requirements of the analysis, such as not to mistake the core user research for active user analysis. The manager Liu of the net dragon once shared with me to one such case, product Manager and you propose to do a COC activity data analysis Report, to measure the effect of activities, in general, you will be active in the early, medium and late game macro data out, and then draw a picture of the performance of each stage, and then make a judgment. Then leap with the report to the product manager, so feel the trouble. If you look at it from a data analyst's point of view, such a report is cheap. When others ask for analysis, maybe he has 10 questions, but only give us a description of 3 problems, we can not solve such a simple 3 problems, we should be more neutral and objective from multiple angles to think about such a problem, and then from the product itself, product players, product operations, and so many angles, Comprehensively measure such a problem, identify potential opportunities, and then make a 300% analysis, not 100%.