Every year at the end of the year, some students sighed: "I have been busy for a year, and I feel that they are all routine data reports, and there is not even a project that can be sold!" So what should I do for high-quality data analysis projects?
01. How to count high quality
If you want to answer this question, you must first clarify: what is a "high-quality" project. In essence, data analysis is a supporting position, and the quality of the work is mainly determined by the department served. If you work in an enterprise, it mainly depends on the evaluation opinions of the management/business department. If it is during the interview, it will be mainly evaluated by the interview HR/employer leader. The key is to find out the needs of the other party and hit the pain points of the other party. ,
Some students often get confused here, thinking: using a linear regression model (complex models will not) / blingbling flashing charts / checking a number sql and writing 2000 lines is considered "high quality", and these things are ignored Whether it is useful to the business or not, the result is naturally a joke. A few days ago, a classmate came to ask in a hurry, saying that they built a lost user prediction model, but the operation said: "What are you doing? I don't know how to use it after predicting it!" Then the project was yellowed... This It is a typical result of behind closed doors.
So how can we do it to solve the business pain points?
02. Identify the core needs
For business, data analysis is a matter of "sneaking into the night with the wind, moistening things silently". People often don't think it's very powerful when there is data to see, but if there is no data to read, some people will be anxious. Therefore, if you want to find the pain points of your business, it is best not to forcefully sell: "I have an artificial intelligence alpha big dog model, which is tested and accurate, so why don't you try it!" Instead, first look at what issues the other department is most concerned about , what data is missing the most.
There are four common cases of missing data:
There is no basic data, and I am eager to see the data
I have data but don’t know how to interpret it, so I’m in a hurry
With data and interpretation, I want to further verify the idea
Have data, have interpretation, want to make further predictions
When receiving the needs of the business side, the real needs must be clear. For example, "user portrait" may be just a word of mouth when the project is launched. Whether the business is not clear about the current situation of the user, or wants to do something based on the portrait, we must understand clearly. It was not clear at the beginning of the project, and the middle process should be gradually clear, otherwise people questioned "what's the use of you!" after a lot of labeling, it would be dumb eating coptis...
03. The main points of the report type project
The number of report-type projects is the largest, but it is the easiest to be overlooked by data analysts. Many newcomers always dislike it because it is not technically complicated. But in fact, report-type projects are the easiest to achieve results. The key is: do what the leaders care about and what the leaders can see. When receiving requirements, distinguish the report users, give priority to visualizing the needs of the leaders, so that the leaders can intuitively feel the data.
Moreover, through report-type projects, the cooperation attitude of business parties can be effectively identified. If the business side has a good attitude, then they can cooperate in depth. Now that there is a business monitoring report, the next step is to analyze abnormal business trends. First record the abnormal points caused by non-business active behavior, and then analyze in depth:
How large a change is abnormal
what causes the abnormality
How to detect anomalies in advance through data
With these accumulations, we can further do automatic abnormality reminder + problem diagnosis, so that the simple data display can be improved to a higher level, and at the same time, it can lay a solid foundation for subsequent in-depth analysis.
04. Essentials of Analytical Projects
In the original impression of many people, data analysis should be to get a bunch of numbers, and then Mami Prajna coaxes them to analyze and tell the business three sentences, so that the business can earn an extra 180,000! As a result, people tend to have high expectations for analytical projects. But in fact, analytical projects are particularly easy to step on mines. Insufficient understanding of the business, lack of monitoring data, and lack of experience in abnormal analysis will make problem analysis superficial. When working on a project, "loud thunder and little rain" is the norm.
Therefore, analytical projects are incubated on the basis of report-based projects, and the success rate is relatively high. If it is found that the business side has insufficient monitoring and unclear understanding of the problem itself, it can return to the report-based project to start. After a certain amount of accumulation, if you want to be effective, the best way is to first agree on the business assumptions, and figure out what the business side has no confidence in and what they have confidence in. It is easier to falsify than to prove true, and it is easier to produce results by directly verifying hypotheses.
If the problem involves too many intractable diseases that are difficult to quantify, there is another solution, which is to convert the problem into a test project. Directly see what solutions the business side has at hand to solve the problem, and then test that method works. This can also output a solution to the problem.
05. Key points of test projects
Test-type projects are relatively easy to succeed. In essence, testing also belongs to the situation of "the business has no data, and I especially want to see some data". But it should be noted that what to measure, you have to think clearly in advance. The most important thing in testing is to have a preliminary understanding of the factors that affect the results, test the key factors you want to test, and control other interference items.
Therefore, the general page design test is easy to succeed, and the consumption result test is easy to mess up. Because there are few test points in the page design, it is easy to produce accurate and stable results. However, there are too many factors that affect the consumption results. Before doing the test, I did not think about it clearly. It is easy to cause the results to fail due to the low comparability of the test schemes, the large differences in the test groups, and the failure to exclude key interference factors.
Therefore, before doing the test, it is necessary to conduct basic analysis. It is necessary to sort out which factors will have an impact, and how big the difference is between several sets of test plans, which can effectively improve the quality of the project.
06. Key points of predictive projects
The key to predictive projects is to confirm the real forecasting needs and avoid blindly gambling "I want to be 100% accurate". Not only can't it be done, but it doesn't make sense. For example, the lost user prediction mentioned at the beginning, if the operation is to invest all resources to recall lost users, then change the goal to forecast: "Which people will naturally return" can save money. If the operator wants to obtain the maximum effect, the goal can be changed to: "Which way of recall is the user expected to respond?" In this way, multiple rounds of pushes can be made to maximize the awakening of users. In short, it is more effective to figure out the operation plan first than to build a model behind closed doors.
07. The sense of ritual is very important
Data analysis projects especially need a sense of ritual! Because the results of data analysis rarely become white money, everyone may have almost forgotten it when the summary is made at the end of the year. So be sure to do a full ritual. For example, when the project started, we held a meeting with our partners, and at noon we had a round table meal together. At the end of the project report, I specially made an appointment with the big boss, and took a group photo after the report. The results of the project should be posted on BI as much as possible, and a "Data Asset Big Screen" will be specially made and placed in the boss's office. Scroll through it every week to see how much new data has been accumulated, and how much benefit has been generated to help the business. Of course, the "Double Eleven Combat Big Screen" must be displayed, and remember to take pictures to record the grand occasion of colleagues cheering in front of the big screen...
There are many specific methods, and you can adopt them according to the style of your own company, but there is only one core idea, that is, unite with colleagues more, hold more meetings, and use more systems. Don’t silently submit a ppt, submit an excel, and submit a csv. . The lining has been worked hard, and the face must be enough!
