Before carrying out a data analysis work, the first thing to do is not to head in to get a bunch of data, pile up a bunch of data, but first think clearly about what the overall analysis framework, what data analysis methods, data analysis methods, which data you go to the organization, to guide the subsequent development of the entire data work. The following is an expansion:A: Comparative analysis of comparative analysis, is the most basic, most commonly used in data analysis, is also one of the most practical analysis methods. The method mainly refers to the comparison of two or more objects of the data indicators, to describe the quantitative differences in the comparison of objects, so as to derive the business at different stages of the trend of change and the law of the more common comparative analysis: from the time trend of the ring, year-on-year, fixed-base comparisons, from the space on the comparison of the A/B test, the comparison of similar space, the comparison of the advanced space from a specific standard with the target value, Assumptions, the average value of the comparison of an example: conversion / active indicators, today and yesterday for ring comparison, this Monday and last Monday for week-on-week comparison, a company and the industry average comparison ..... Second: segmentation analysis "no segmentation without analysis", this is a common saying when we segmentation analysis. It is clear that segmentation analysis is a very important means of breaking down step by step, is in the process of constantly asking why segmentation analysis is generally two kinds of: (1) step by step breakdown, from coarse to fine, from shallow to deep, and gradually talk about the process of line segmentation. For example: sales decline or rise, the first split to the country, province, city/region, stores, compare and observe which region becomes larger brought; traffic quality deterioration, the first split to paid, free, paid split to the application market, social media ... The traffic quality deteriorates, first split to paid, free, paid split to the application market, social media ..., and then split to GooglePlay, AppStore, Facebook, Twitter, Snapchat and so on. Summarized in one sentence: segmentation is gradually refined step by step down drilling, dismantling. (2) cross-segmentation cross analysis, is based on the vertical analysis and horizontal analysis, from the cross, three-dimensional point of view, from shallow to deep, from low to high level of a method of analysis, it makes up for the independent dimensions of the analysis can not be found in some of the problems. For example:Four quadrants, RFM model three: A/Btest "split bucket is the most scientific", A/BTest is for the same goal to develop two programs, in the same time dimension, respectively, so that the same (similar) user groups randomly use a program to collect the user experience data and business data of each group, and finally according to the significance of the test analysis to assess the superiority of the program. Significance test analysis to assess the better program and formally adopted for example: for example, there are A, B two copy, through a random way to allow users to see, use one of the copy, and then assess the two groups of people bounce, click, use and other data IV: Funnel Analysis Funnel analysis is a set of process data analysis, it can scientifically reflect the state of user behavior and from the beginning to the end of the various phases of the user conversion rate of the situation of the important The analysis model is commonly used in scenarios such as registration and login conversion, browsing and transaction conversion, store sales conversion, etc. By quantifying the conversion rate of each step, the success or failure of a business or product can be measured and optimized and adjusted to the point of For example: an e-commerce APP, from the user to download the APP, access, registration, browsing, transaction, calculate the value of each step and the ratio of the funnel V: Retention Analysis " Users come fast, go fast", retention rate, is to do operations or user growth students must look at an indicator, it is a measure of a piece of business whether the health of the key indicators, do a good job of retention will bring long-term compounding effect, there is no future retention without retention retention analysis, in the field of data operations has a very important position in the common retention indicators are the next day retention, seven-day retention rate, Next week retention rate, next month retention rate, T + N daily/weekly/monthly retention rate, etc., indicating that the target user after a period of time to return to the product or return to the product to complete a certain behavior of the proportion of, for example: there are 100 people installing access to the APP, the next day there are 40 people continue to visit the next day, the next day retention rate is 40%, the seventh day there are 20 people to continue to visit the 7th day retention rate is 20%.... Sixth: correlation analysis correlation analysis, the study of whether there is some kind of dependence between the phenomena so as to find the key impact of business operations and factors. Correlation measurement methods include: scatter plot, correlation coefficient and other correlation analysis, there are mainly the following three types: (1) single correlation: the correlation between two factors is called a single correlation, that is, the study involves only one independent variable and a dependent variable (2) complex correlation: three or more factors of the correlation is called a complex correlation, that is, the study involves two or more independent variables and the dependent variable is related to the study; (3) Partial correlation: in a phenomenon and a variety of phenomena related to the occasion, when other variables are assumed to remain unchanged, the correlation between two of the variables is called partial correlation. For example, the relationship between education and income, study time and academic performance, the number of pages viewed by the user and the number of products purchased? Seven: Cluster Analysis "Things are gathered in groups", cluster analysis is one of the commonly used data analysis methods, the core of which is based on the data before the existence of similarity clustering methods are K-means (K-Means), spectral clustering (SpectralClustering), hierarchical clustering ( HierarchicalClustering), and the clustering method is based on the data before the existence of similarity clustering. HierarchicalClustering), the specifics will not be described. Commonly used in the following two scenarios: (1) user segmentation: based on the similarity of users into different communities, and study the characteristics of each community and do business applications (2) anomaly detection: to find normal and abnormal user data, to identify abnormal behavior, such as: based on the user's registration information, access behavior, transaction information (commodities, amount of money, etc.), through the clustering analysis of the similar population of different populations, and compare in different dimensions and in different ways, and compare in the different dimensions and in different ways, and compare in different dimensions, and compare in different dimensions, and compare in different dimensions, and compare in different dimensions, and compare in different dimensions, and compare in different dimensions, and compare in different dimensions. The new year is a new year, and we will bring you to learn more about new media operation. Everyone remember to pay attention to not lost oh!