Data analytics has been gradually applied to all areas of work/life. The essence of data analytics is to see the deep-seated patterns and mechanisms from the complicated data, so as to make predictions on what has not happened! So, learn about 3 common data analysis scenarios, i.e., describe the current situation, analyze the causes, and predict the future, to deepen your understanding of data analysis.
Data analytics has been gradually applied to all areas of work/life. The essence of data analytics is to see the deep-seated patterns and mechanisms from the complicated data, so as to make predictions about what has not happened!
So, learn about the 3 common data analysis scenarios - describing the current situation, analyzing the causes, and predicting the future - to deepen your understanding of data analytics.
✔ Describe the status quo: dismantle the business logically and systematically through descriptive analysis, and assess the state of the business as a whole with reasonable metrics.
✔ Analyze the causes: through diagnostic analysis, analyze the reasons behind abnormal fluctuations in the business and propose solution strategies.
✔ Predicting the future: predicting the future of the business through predictive analytics, based on existing data, in context.
Describing the status quo: a descriptive analysis
Descriptive analysis requires a more macroscopic and deeper understanding of the entire business, using a systematic framework and reasonable indicators to assess the business status, clearly determine the current state of the business and locate the causes of business fluctuations in the data.
Generally, daily and weekly reports, or descriptive analytical reports for a particular piece of business are deposited on the data product to be updated automatically, as the business side needs to keep an eye on the relevant data frequently. The purpose is as follows:
①Enhance work efficiency: Reduce repetitive workload by automating regular refreshing of descriptive analysis reports through data products.
② Improve Leverage Efficiency: By dismantling the logic of descriptive analytics reports, more people can understand the business state and know how to improve it.
▼ Analytical Methods:
·····
▼Typical Problem: The boss wants to understand the traffic conversion of the enterprise's online product A through overall analysis.
wrap-up
Descriptive analytics reports are in most cases deposited on data products to reduce duplication of work for data analysts and increase productivity.
It can be organized in the following order:
(1) Descriptive analysis report for the object: according to the scope of responsibility, to determine the descriptive analysis report to show which business
(2) The display order of the business: it can be displayed according to the level of the business, or according to the order of the process between the business, etc.
(3) Evaluation of specific operations: Developed according to a total-substructure, with 3 levels, i.e., evaluation of indicators, performance of indicators, and form of presentation.
① Determine the core monitoring metrics of the business and the monitoring metrics of the disassembled sub-modules.
② Assess the performance of indicators using comparative analysis to reflect changes in operations
Targeted comparisons: progress in accomplishing goals versus time schedule
Horizontal comparison: comparison between different objects
Vertical comparison: comparison of different dimensions of the same object (different time periods, same ring, before and after activities, etc.)
③ Reasonable forms of presentation: trend graphs, color gradients, etc.
Analysis of causes: diagnostic analysis
Based on the business logic, for the business to identify the causes that cause it to get better or worse during operation, we need to identify them through diagnostic analysis and combine them with business research to address why they occur. Also known as Root Cause Analysis, it is used to identify the root causes of business problems and find appropriate solutions to prevent these problems from occurring in the future.
▼ General ideas for diagnostic analysis:
Identify the problem->Define the problem->Deconstruct the problem->Find the cause->Propose the solution->Execute on the ground->Feedback Iterate->until the business problem is solved
All of them need to know the business very well and be sensitive to the data to be able to determine whether the fluctuation of the indicators is abnormal or not. Commonly used methods to determine whether the fluctuation of indicators is abnormal include the box-and-line diagram method, the Six Sigma principle, etc. The principle of these methods is the same, that is, to define the range of normal fluctuations, and then determine the outliers, only the methods and criteria used are different.
▼ Analytical Methods:
·····
▼ Typical Problem: Business asks data analysts to help analyze the reasons for a recent sales decline at retail store A, which is operated offline.
wrap-up
1. The recommended structural order of the report of the diagnostic analysis:
Title Page - Table of Contents Page - Conclusion - Strategies - Subargument Argumentation Process - End Page.
(1) The reason for placing the conclusions up front is to give the business side a macroscopic view of the whole analysis right at the beginning.
(2) Place the strategy after the conclusion and give a solution to the problem after summarizing the conclusions of the analysis.
(3) Follow the strategy with a detailed description of the analysis process and findings.
2. notes on the 3 parts of the conclusion, strategy, and subargumentation process:
(1) Conclusions: summarize the conclusions into 3 to 5 concisely, too many conclusions will make the human brain less effective in remembering them.
(2) Strategy: write a clear strategy plan, landing plan and benefit evaluation.
(3) The process of subargumentation
① The title of the PPT is a summary of the content of the entire page of the PPT.
② Content structure of PPT: elaboration of subarguments + data argumentation (graphs/tables)
Predicting the future: predictive analytics
Whether it is a large enterprise to set strategic goals, or as small as the project manager to make business decisions, we need to predict the future development of the business to assist in the judgment, which is within the scope of predictive analytics work.
Forecasting refers to making predictions about the future based on known information (historical data, subsequent resource inputs, etc.) and assuming that the trend of things will extend into the future. Possible special impact events are not considered, but some room for adjustment is left at the end. The two main scenarios are as follows:
① Top-down: the boss first determines the final goal to be achieved, and then disassembles downward the milestones to be accomplished and the resources needed in the estimation process.
② Bottom-up: Estimate the goals of each module of the business first, and then summarize them upward to get the goals that can be achieved by the business as a whole.
▼ Analytical Methods:
For the framework of predictive analytics, we usually break it down based on formulas in two common forms.
(1) Based on mathematical relationships
E.g. GMV = number of paying users × average amount paid.
(2) Based on business logic
e.g. GMV during an operational campaign = GMV before resources are invested × Lift Factor.
Commonly used analytical methods include moving average, correlation analysis, and various interpolation methods.
·····
▼Typical question: E-commerce companies do a lot of marketing activities to stimulate users to spend money on their platforms. Recently, our company is planning for the "Double 11" campaign, and now the product department and operation department want us to estimate how much GMV will be achieved on the "Double 11" day this year.
wrap-up
In most cases, the reports of predictive analytics are presented in the form of Excel sheets. When writing a predictive analytics report, there are several things to keep in mind:
(1) It is important to show the measured process data and the final result data separately, and try not to mix the two together.
(2) Be sure to keep the formula for the process to facilitate subsequent adjustments to the data.
(3) It is advisable to use a separate worksheet to document the logic of calculations between data and the caliber of indicators.
The process of measurement can be shown in the logical order of data calculation to reduce the cost of understanding low, and strive to make the predictive analytical report structure is clear, logical, and data prediction of the interpretability of the strong.
Finally, I would like to introduce Arigato BI - NetEase's enterprise classIntelligent Big Data Agile Analytics Platform
Digital BI Product Manual - Free Download:/doo5t5