Five questions to see if you can make a good data analysis project

This article selects five classic questions in data analysis, tests whether you have the concepts that an excellent data analyst should have through practice, and shares the five key points for making an excellent data analysis project through the questions, which is very helpful. It is suitable for those who are interested in data analysis but don’t know how to start.

Many students often ask: "How to create an excellent data analysis project? Otherwise, I don’t know how to write my resume and year-end summary." I have given detailed answers and follow-ups. Today I will summarize and share the common problems mentioned by everyone.

If you want to do a good job in data analysis projects, you must establish the right concept . Here are 5 test questions to test how likely you are to do a good project.

Question 1 (single choice question)

The indicators for measuring the quality of data analysis projects are:

A. Time, cost, quality

B. Algorithm difficulty, statistical knowledge, mathematical formulas

This topic is the most important concept, which directly determines whether a data analyst can succeed or fail in the current company. Data analysis work has its particularity:

★ It is different from sales and cannot directly generate revenue for the company.

★ It is different from operations and cannot directly increase active retention and payment indicators.

★ It is different from the development of transaction/website/ERP systems, which are necessary supports for the business.

★ It is different from DBA. There is no company without DBA, but there are many companies without full-time analysts.

Data analysis is essentially a highly replaceable auxiliary position. Before the concept of data analysis became popular, many companies hired a programmer who could write SQL to fill this position. It's like the relationship between a scope and a gun. Without a scope, the gun can still shoot. With a scope, the gun can shoot more accurately.

Therefore, although data analysis is supported by algorithms, statistics, and data, companies do not care about these book chapters, but what is the use of it for the business? How useful is it? Like other projects in the enterprise, the most important thing for data analysis projects is to examine time, cost, and quality.

If you are separated from these and just empty-mindedly pursue "my method is so complicated, I am so awesome", then you should go back to school to do scientific research. Scientific research needs to pursue high-end and sophisticated technology, while enterprises pursue: the lower the cost, the better, and the shorter the time, the better, while achieving the goal. So you must choose A for this question.

Many data analysts who have just graduated, changed careers, or just joined prefer to choose B. Choosing B does not mean there is no future.

Because students who choose B will spend a lot of energy studying hard, so although they may not be able to get promoted in one company, they will still be able to pass interviews when they change jobs, so they can also increase their salary by changing jobs once a year. But if you want to make achievements in one company, it is better to choose A.

This question is the most crucial. Once you understand it, the following problems will be easily solved.

Question 2 (sorting question)

The following people have a say in the quality of data analysis projects:

A. Business department leaders

B. Data department leader

C. Business department employees

D. Data department staff (myself)

If you understand the problem thoroughly, it is not difficult at all. The answer is A≥B≥C≥D. The opinions of the leaders are more important than the opinions of the employees. If the leaders of the business department agree, the leaders of the data department will directly agree. If the leaders of the business department do not speak out, then it depends on whether the leaders of the data department agree. It is meaningless to say "I think I am doing a great job". Remember this.

Please note that sometimes the attitudes of business leaders and data leaders may be inconsistent. In this case, the attitude of your direct leader should prevail, and the external department will consider it later. In most companies, your direct leader is the one who determines your performance score, so you must not offend him.

Question 3 (sorting question)

Please rank the following five project results in order of quality:

A. Visualized data products

B. Data model for regular monthly output

C. Report ppt of departmental level and above meetings

D. PPT without group presentation

E. Excel data table

F. Numbers without fixed format

G. Write SQL and tell the business after the number of runs

A complete explanation of this question would require an entire article, but students can use the literal meaning of the words to give a direct answer. The answer is: A=B≥C≥D=E≥F≥G.

The results of data analysis are easily used as chamber pots: it feels good when you use it, but you forget about it after you use it. People even think it is dirty: I just want a number! Why do I have to run so long?

Therefore, when doing data analysis projects, we should strive to output results that are regularly used, productized, and must be seen by everyone. The best way is to use a set of BI, or use a model to optimize and sort the user follow-up list of the business, so that everyone has to use it.

At worst, write a PPT but try to speak publicly at the meeting. At worst, you run a bunch of ad hoc demand sheets and write 2,000 lines of SQL, but there is no official output, and you don’t even know what to write in the performance report at the end of the year.

Question 4 (single choice question)

Today is 12 noon on November 11th. Your leader asked you to give an estimate before leaving get off work to predict how much our performance will be on Double Eleven. What would you do?

A. Go back and start modeling, time series, and XGboost

B. Go back and find out how the operation and promotion expenses are used, and calculate the input-output ratio.

C. Go back and look at the morning data. Take a picture based on the same period last year.

This question is very confusing. Especially after reading the previous question, many students will habitually choose A. The key to this question is not "prediction" but "12 noon to the end of get off work."

Data analysis can be used for modeling, BI, and visualization, but all of these take time. In many cases, business cannot wait and results must be given quickly. At this time, simple and direct methods should be given priority, and combined with data to indicate risks.

Therefore, you need to learn modeling, statistics, and how to quickly and reasonably make decisions. Many newcomers have worked hard to come up with a bunch of models, but their leaders have overturned them with a few words, and even criticized them: "What have you been doing so slowly?" This is the reason. For this question, choose C. It is best to give an answer within 10 minutes. After the leader instructs: "This question is very important, and a more complex and scientific method should be used", consider AB.

Question 5 (Multiple Choice)

What constitutes the cost of data analysis work?

A. Database cost

B. Computer cost

C. Software development cost

D. BI product cost

E. Data collection quality

F. Data cleaning quality

G. Programmer working hours

This question is also very confusing. Before Teacher Chen asked this question, many people had never thought about it: "Does data analysis have a cost?!" "Isn't this something that someone who has read "Machine Learning", "Statistics", and "Master Python in 21 Days" can do by typing on the keyboard?!"

A: Data analysis definitely has costs, and the biggest cost is data quality. Good data leads to good analysis, and bad data leads to bad analysis. Especially for data collection, business processes are full of loopholes, business management is not standardized, data points are not properly embedded before going online, and basic data is dirty and messy, so analysis is useless.

As for software costs and computer costs, they are just a drop in the bucket. Data quality is a fundamental issue for the company. So the answer to this question is ABCDEFG. If you rank them, it is E≥≥F≥G≥A≥D≥B≥C

Wait! Many students will ask: Why is there G, and why is it ranked so high? ! Because the working time of data analysts is very precious.

The huge and complex models in schools, papers, and patents are all created by spending countless hours. Data analysts in ordinary companies are already exhausted by dealing with various data collection, reporting, and Excel every day. They may not even have time to find a partner, let alone create high-tech models.

So you have to consider your work hours, prioritize your daily needs, concentrate on doing things that are productive, and let those messy “I want a number” requests queue up.

After completing the above 5 questions and understanding the ideas of the questions, how to do an excellent data analysis project is just around the corner.

How to do an excellent data analysis project:

1. Formal project establishment, taking business pain points as project goals

2. Consider time urgency and data quality and design appropriate methods

3. Output results that are used regularly, are productized, and must be seen by everyone

4. Prioritize your needs and give yourself time to do high-quality work

5. Use quick analysis methods to deal with simple needs and save energy for complex needs

In order to achieve the best teaching effect, here are the five most frequently asked questions. The purpose is to let everyone remember the five key points to make an excellent data analysis project.

Author: Down-to-earth Teacher Chen

Source: WeChat public account: Down-to-earth Teacher Chen (ID: gh_abf29df6ada8)