1-- concepts and definitions

If the analysis thinking is reflected in a structured thinking, then the data analytical thinking (referred to as data thinking) is based on structured data analysis relies on the way.
Is different from "I think", "what used to be", "others how to" these intuitive experience of the analogy of the way of thinking, thinking data is data-driven, based on rigorous analysis, statistics and prove to guide specific application and operation.

First, we must have a comprehensive and objective understanding of the thing itself.

Dialectical thinking to understand and look at things, things break down and conduct an integrated, comprehensive and objective data speak, while reducing subjective biased view,
Combined with the external environment, behavioral characteristics timeline, core dimensions, overall analyte or dynamic show
Focus on things external interactions, internal structure and cohesion, analysis of internal and external environmental factors related things

Then, to determine the key attributes of the thing itself, dimension, and analysis and evaluation system.
Each of the key indicators of mutual restraint and promote inter-force characteristics to analyze things.
Any evaluation index value derived, there must be something inherent in itself data and operational support mechanisms, that is, the mapping data will be analyzed to something specific on the available scientific evaluation system.

1.1 - The purpose of data analysis

Clarity of purpose data analysis itself, Begin with the End.
Data analysis is to be able to quantify a way to analyze business problems and draw conclusions.
"Insight obtained from the historical data into executable decision or recommendation process, IT technology, management science and statistics are combined to solve practical problems."

1.2 - general procedure for data analysis

Operational guidance data, data-driven business.
Data analysis is not a result, but the process, in this process is the need for feedback and continuous improvement.

1.3 - the role of data analysis

Three main functions for data analysis, mainly: situation analysis, cause analysis and predictive analytics.
When what kind of data analysis carried out, we need to be determined according to our needs and purposes.

The clearer the purpose of data analysis, analysis of the more valuable.
After a clear purpose, carding ideas, frame structures analysis, the analysis object into several different analysis elements, and then determine the specific analysis and analysis of indicators for each analysis point;
Finally, to ensure that the frame analysis system (systematic, i.e. before any analysis, the analysis of any such logical connections between each analysis point having a), so convincing results.

1.4 - knowledge data (Data Common Sense)

Be sensitive to changes in the data.
Here the "change" is not from a change in the value itself, but more from the "current data" is different from the usual knowledge and judgment.
This kind of "normal cognition and judgment", in fact, is not only a long-term observation and analysis of data formed by the "feel" is your "experience and experience" or even "The current structure of knowledge and ability tree."
Able to recognize the source of "change" can avoid falling into the "subjective judgment" trap.

We want to cultivate awareness of data sensitivity and the ability to deviate from the data, which need to maintain sustained observation and analysis, as well as enough patience and curiosity.
Service data corresponding to the business process, the business process implies business requirements, traffic demand from the actual operation, from a practical application to understand the generation and flow of data helps to maintain accurate "direction" data analysis without departing from the "track."

View concerned about data form the habit of clear and reasonable value Meaning
Keep in mind the core data, indicators, reports and rankings, repeated projections rationality
Detached and structured data, unified storage, easy access and analysis

2 - frame structures

Quantification is to unify the cognitive and ensure that the path can be traced back, can be copied, avoid the "I feel", "I guess" and other subjective judgment.
Back path may be copied: by quantization result, many methods can be optimized to find the cause and may be copied.
To do quantify, need to do three things: establish a system to quantify, quantify clear focus and ensure data accuracy.

2.1 - Create quantification system

If you can not measure, you can not effectively grow and improve.
Uniform standards need to be defined and evaluated this standard is the target.
Specific indicators value can be avoided, "I think 'cause cognitive trap," fuzzy false, "the description would people astray.

2.2 - a clear focus on quantification

Each stage should clear the current business focus; quantization stage system based on business needs, and focus on ways to quantify changes.
This also means: there are more and more indicators of the details of the monitoring and promotion efforts.
Each stage, we need to determine whether the current focus according to different business situations, no dead so as to establish a monitoring system of analysis around the focus.

2.3 - index design method

Isolated play no index value of the data necessary to establish the index system structured.
Different businesses have different forms of index system, there is no universal template.
Indicators can be broken down and disassembled, indicators should be chosen depending on the circumstances,

Accurate and understandable criteria, a set of statistical methodology and results of operations.
It refers to the ability to accurately measure accurately meet the purpose, easy to understand visual display is an indicator of good or bad algorithm, and the algorithm can also be easy to understand indicators.
Accurately it is necessary to guarantee.

Some commonly used statistical tools for the design of indicators

Business Profile: mean, median, mode
Business difference: variance, standard deviation
Business Distribution: Frequency

Some Tips

Ratio indicators: easy to focus on the actual results
Associated indicators: to prevent the side effects of a single "one-sidedness targets"
Prevent bad indicators: error indicator, vanity index (too far from the core business objectives), complex indicators (data fluctuations, difficult to analyze the reasons)

Dictionary Index dimension (Dimension Dictionary)

Dictionary Index dimension, (Bus Matrix), to a certain extent to solve the problem for ill-defined indicators or non-uniform.
Can uniformly manage some critical and common indicators.

A clear definition and interpretation of indicators, requiring makers must have a thorough understanding of the business and has a very high abstraction.
The industry has been looking for a method to quantify and promotion within the organization, the thing is a multiplier.

2.4 - Indicators distinction

"Good indicator"

Not all indicators are valid, the core driver is the need to focus on indicators of "good indicators."
In simple terms, the core driver indicators and organizations associated with the development of the entire operations teams, product development teams as well as a unified team in whom goal is to focus on the direction within a period / stage.
The core business of driving different indicators are not the same.
"Good indicator" should be on the effective rate or the proportion of base, easy to measure and compare.

"Bad indicator"

Vanity Index: There is no practical significance, can whitewash job performance
Posterior indicators: timeliness poor, in fact, only in the description of the event has occurred, it is difficult to recover costs and make up for losses through measures
The complexity of the indicators: data analysis fall into "a bunch of hidden variable targets" trap, unable to start

2.5 - create the correct index structure

According to "index design method" to establish a system of indicators around the business.
The core index is determined according to the service feature core, disassembled at a different angle on the basis of core indicators, then slowly added indicators of other services.

Pyramid structure and analytical thinking, like dismantling process in accordance with the methodology of the pyramid "step by step dismantling, do not leak (MECE)", indicators show a tree structure, is to build the core business processes as thinking, structure-guided .
From the perspective of the process to build a framework of indicators can include comprehensive user-related data, there are no omissions.
Indicators listed principles: the need for a core driver indicators. Remove the vanity indicators appropriate for exclusion, do not add index and add index.
If too much out dismantling or business complementary indicators can learn the concept of "field" data warehouse to manage these indicators.

3 - Data Accuracy

Ensure the accuracy of the method

Take credible source: multi-source cross-validation, should be particularly careful when using new sources
Confirmation processing methods: processing algorithm and index definitions
Double Check: middleweight, computational logic and common sense business

Double Check Tips

Middleweight Check: boundary values, each has its own data about the scope of
Computational logic Check: Total, median, average, etc.
Business knowledge Check: figure out the business scope in accordance with other commonly used digital

Data products already have mature data quality management; involves data sources, index calculation and presentation of data and other aspects of monitoring.

4 - communication and feedback

4.1 - stand in the perspective of the business side

Only solve business problems in order to create value analysis, including personal values and corporate values.
"Worry they consider, to its desired": precise understanding of the needs of the other party.

4.2 - main link

Communicate well: Determine what you want to analyze the business side, put forward a more reasonable way of professional measurement and analysis, clearly the meaning behind the data should, while making node synchronization, should a road go black
Conclusion simple: in the elaborate analysis results, remember conclusion first, layer by layer to explain, and then provide arguments. The argument map> table> Text
Can provide information and suggestions floor: From the professional point of view, from the known to the unknown boundary analysis border, unknown to the other party to provide information and advice available landing
Seek feedback: continuous improvement trigger point

5 - Method manner

5.1 - the difference between data analysis methodology and data analysis method

Data analysis methodology mainly from a macro perspective guidance on how to analyze data, like a pre-planned analysis of data to guide the conduct post-data analysis work.

The use of dimensional analysis data
Use knowledge of statistics, such as data distribution hypothesis testing
Using machine learning

Data analysis mainly from the microscopic point of guidance on how to analyze data, means that a particular method of analysis, such comparative analysis, cross analysis, correlation analysis, regression analysis.

5.2 - Dimension Analysis

Dimensions are parameters that describe the object, in particular the analysis, it can be considered a point of view of things.
With the dimension, it is possible by a combination of different dimensions, forming a data model, a multidimensional data cube.
Data model complex data in a structured and orderly organized.
Data model can be viewed from different angles and aspects of data, thus increasing the flexibility of the analysis, to meet the different needs of analysis, this process is called OLAP (online analytical processing).

5.3 - metrics and dimensions What is the difference?

Dimension is the angle of looking at things and explain, index is a measure of data.
Dimension is a larger scale, not just data, such as the time dimension and urban dimension, we can not be represented by indicators and indicators (retention rate, bounce rate, time spent, etc.) but they can become dimensions.
Established by business and selected indicators, the index as a dimension, using the dimensions analysis, the popular understanding: Dimension> index.
In general, the indicators can be used as a dimension.

6 - Data analysis Why-What-How

7-- "bullying"

Treat drug dose aside talk toxicity; calculating the return, the amount set aside to talk about the cycle;
reading list, put aside the difficulty of talking about the number of pages; the ability to assess, aside talk about life experiences;
and so, logical confusion, rogue extremely!

Data - Data thinking - on