Learning notes - 7 weeks as data analyst "in the first week: data analysis of thinking"

I. Three core thinking in data analysis

Structure: Structure your analytical thinking:

1. Organize arguments
2. Progress and disassemble arguments
3. Complete and supplement the argument

Core argument: Find the top of the pyramid, it can be a hypothesis, a problem, a prediction, a cause
Structural disassembly: Top down, disassemble the core argument into component arguments, and show causal or dependent relationship between the upper and lower.
MECE: mutually independent and completely exhaustive. Avoid overlapping and repetition between arguments, and the sub-arguments should be as perfect as possible.
Verification: Both core arguments and sub-arguments should be quantifiable and speak with data. They must be verifiable.

Structure is analytical thinking, but it is not enough data, and inevitably has the disadvantage of divergence.

Formula :
The top and bottom are calculated for each other, and all structures can be quantified
The left and right are related, the least indivisible

Addition can be used to superimpose different types of services.
Subtraction is often used to calculate the logical relationship between services.
Multiplication and division are various proportions or ratios.

Using structured thinking + formulaic disassembly, the final analysis argument is obtained. Many times it is a phenomenon. Data is the embodiment of a certain result, but it does not represent the reason.

Commercialization: Fitting the business
Structured thinking (smoothing the way of thinking) - structured data (making it digitized) - structured business data (landing, fittingbusiness)

Second, seven thinking skills of data analysis

1. Quadrant Method:
   Core: It is a strategy-driven thinking
   Application: Wide range of applications, strategic analysis, product analysis, market analysis, customer management, user management, etc.
   Advantages:Intuitive, clear, manual division of data.The division results can be directly applied to policies.
   Note: Quadrants can be divided by median, average, or experience.
   

2. Multidimensional method
   User statistics dimension: gender, age...
   User behavior dimension: registered users, user preferences, user interests, user streaming...
   Consumption dimension: consumption amount, consumption frequency, consumption level...
   Product dimension: product category, product brand, product attribute...
   

3. Assumption method
   The company conducted a marketing campaign, and the overall sales data on the APP increased by 20% compared with last week. Due to statistical errors, the details could not be obtained. The problem now is that the sales itself may beRaised because of the festival, how can you prove that the event is valid or invalid.
   
   You are a self-operated e-commerce data analyst. Will the income change after the price of goods is raised? What do you do?
   
   Core: Hypothesis is a kind of thought-driven thinking.
   Advantages: When there is no intuitive data or clues to analyze, inference is made in a hypothesis-first manner, which is an argumentative process.
   Application: It is more of a way of thinking, assuming-verifying-judging.
   Notice: You can assume not only the premise, but also the probability or proportion. Everything can be assumed, as long as it is self-consistent.

4. Exponential method
   (1) Linear weighting
   (2) Inverse proportional
   (3) log

Core: is a goal-driven thinking.
Application: Different from the hypothesis method, which lacks valid data, the index method cannot make use of the data and process it into usable.
Advantages: Strong goal driving force, intuitive, concise and effective.It has a certain guiding effect on the business.Once the index is established, it is not easy to change frequently.

5. Two-eight method
   (1) In the data, 20% of the variables will directly produce 80% of the effect, and data analysis should focus on these 20%.
   (2) It is a very good habit to keep paying attention to TopN's data, especially in some industries.
   (3) Although there are many indicators, some indicators are often more valuable. The 28 rule can not only analyze data, but also manage data.

Core: The 28 method is a kind of thinking that only focuses on the key points.
Application: exists in almost all fields, so there is no limit to this kind of analytical thinking.
Advantages: It is closely related to the business and more closely related to KPIs.It can achieve good results with almost the least experience, and the price is very good.
Notice: When conditions permit, data analysis still cannot give up the overall situation, otherwise it will make the thinking narrow.

6. Comparison method
   
   Competitor comparison
   Category Comparison
   Feature and Attribute Comparison
   Time YoY MoM
   Conversion Comparison
   Before and After Change Comparison

Core: Contrast method is a way of thinking about mining data rules
Application: Contrast is more of a habit, and it is the tip of the horn in data analysis. A qualified analysis must use N comparisons.
Advantages: Contrast method can find a lot of rules between data, it can be combined with any thinking skills, such as multi-dimensional comparison, quadrant comparison, hypothesis comparison and so on.
Notice: When conditions permit, data analysis still cannot give up the overall situation, otherwise it will make the thinking narrow.

7. Funnel Method
   Core: It is a process-based way of thinking.
   Application: Can be used for changes and processes.
   Advantages: A single funnel analysis is useless, and a conversion rate of 20% doesn't mean anything.It should be combined with other analytical thinking, such as multi-dimensional, such as contrast.
   Note: There is no single conversion rate.