Advantages: clear, simple, intuitive, the results of the division can be directly applied to the strategy (live data)

Application: more common, strategic analysis, product analysis, market analysis, customer management, user management, product management

Note: the division of the quadrant can be based on the median, or by the average or different empirical dimensions.

Natural factors dimension: customer's gender, age, region, etc.

Customer's social characteristics dimension: income, occupation, education, etc.

Behavioral characteristics dimension: preferences, interests, number of purchases, frequency of purchases;

Consumption dimension: amount of money consumed, frequency of consumption, level of consumption;

Commodity dimension: commodity category, commodity brands, commodity attributes;

Multi-dimensional analysis of the different dimensions of the composition of the cube, usually available visualization tools to analyze the content of the multi-dimensional display (such as Tableau).

Application: better when dealing with large amounts of data and more dimensional data, but pay attention to the Simpson's paradox (the way to avoid Simpson's paradox can be used to drill down);

(Simpson's paradox tends to occur when the differences are large within the group, and can be avoided when the differences between the groups are large).

For example, what to analyze in the absence of breakdowns?

The company ran a marketing campaign over the holidays, and the sales data on the app was up 20% overall from last week. Now for some reason the breakdown data is not available. May I ask how to prove whether the campaign was effective or ineffective when the sales itself could have gone up due to the holidays?

Assuming the campaign is valid - there will be a certain number of users who will buy, if this one is proved, then there is a reason to believe that the campaign is valid - what observable behavior will happen when users buy through the campaign? - - Assuming that some users will comment on the message, then the words mentioning the campaign can be counted - When users mention the campaign marketing campaign, then how much is actually effective? 10%? 20%? - - Assuming that the behavior of users involved in the activity has not changed, then through the historical data of the percentage of user comments, and then reverse the number of people to buy, that is, by controlling variables and manually set a ratio to reverse.

You are a data analyst of self-owned e-commerce, and now you want to think about whether there will be a change in revenue after the commodity price increase.

Linear weighting: the proportion of weights for different metrics

Inverse proportion: y=k/x, the effect of convergence (can be combined with linear weighting)

Log: make the data smaller, but the data is greater than 1. Individual values won't have a big impact on the data

(the three are basically used in conjunction)

It is also called the law of two or eight, Pareto's Law

Focus on 20% of the data, TopN, grab the core, the energy used in more important places (the enterprise's resources are limited); focus on the core indicators, of course, also take into account the overall situation;

Good data indicators, must be the proportion or ratio; good data analysis, must be used in comparison ( Lean Data Analysis )

The data analysis of the data, must be used in comparison ( Lean Data Analysis )

The data analysis of the data, must be used in comparison (). )

Because the lone number is not evidence, a single number is meaningless, to be compared with other data to be meaningful, generally using the percentage, absolute value, contrast data;

Usually there are competitors to compare, category comparison, characteristics and attributes of the comparison, time year-on-year comparison, conversion comparison, before and after the change in the comparison;

Contrast method is a way of thinking about mining data patterns, and can be combined with other methods, multi-dimensional comparison, quadrant comparison and hypothesis comparison

Usually the conversion rate, is a way of thinking about the flow of words, but a single funnel analysis is not very useful, and other analytical methods should be combined with other analytical methods (multi-dimensional method, comparison method)

The analysis of the perspective is generally divided into four categories: comparative perspective, correlation perspective, classification perspective and descriptive perspective. perspective, categorization perspective and descriptive perspective.

The value of the analysis can be enhanced by using a combination of analytical perspectives on the premise of the previous analytical framework and thinking skills.

Usually there are frequency statistics, mean analysis, normalization, conversion rate analysis, PSM model and attribution analysis;

ANOVA, crosstabulation, comparison of means, factor analysis, correspondence analysis, brand and awareness map analysis, regression analysis, linear programming, etc.;

SWOT analysis, Porter's Five Forces (consumers, suppliers, direct competitors, substitutes, potential entrants), clustering, and clustering. potential entrants), cluster analysis, matrix analysis, Graveyard model, KANO model, etc.;

Concentration trend (weighted average calculation of the score), out of the trend (coefficient of variation method to determine the weight);

The above analysis methods and models are basically a combination of multiple perspectives, categorization is only categorized into the most important analytical perspective in the actual work of the application needs to be carried out in a comprehensive manner. In the actual work of the application of the need for comprehensive assessment