Author: Chen grounded gas, micro-channel public number: down to earth school. A data analyst with ten years of experience and CRM experience in multiple industries.
Many students suffer from lack of data analysis ideas. They are either in a daze with data or do not know how to use data to demonstrate business problems. I searched articles on the Internet, and all I saw were: structural thinking, contrasting thinking, and underlying thinking. Today, we will explain in general, how to form ideas, so that everyone can understand.
1. Basic analysis ideas, so build
The weather has been torrential rain recently. One day you were walking on the street and you were soaked by the sudden torrential rain. He ran home, feeling cold, trembling, and sneezing. What would you think? ——The common sense of life tells you: You may have a cold! At this time you may choose to ignore it and just carry it over. Maybe take some cold medicine, because you assume you have a cold.
After a few days, there were no symptoms of chills, tremors, or sneezing. If you feel that your cold is over, you won't take medicine anymore. But if you find that you are still not well, you may even feel your head is hot. Damn, I hurried to find an individual thermometer to measure it, and when I saw 38 degrees, I felt nervous, so I went to see the doctor.
This is an example of a complete data analysis idea. Many students may wonder. What! This is the data analysis? What about the underlying logic, fission thinking, and CNN model? Remove all kinds of fantasy terms, remove the huge and complex statistical and mathematical methods. The essence of analysis is as simple as this:
- Observe the phenomenon (cold, shaking, sneezing)
- Combination principle (symptoms of a cold)
- Make an inference (I have a cold)
- Take action (take medicine / carry it)
- Verify the hypothesis (take medicine for 3 days/carry it for 3 days)
- Further analysis (disappearance of symptoms/intensification of symptoms)
- Take further action (ignore it/see doctor)
Analysis ideas have little to do with specific tools. In the era without data, there are qualitative analysis methods; in the era with data, there are survey analysis based on questionnaire data, business analysis based on transaction data, and user APP/website. "Big" data analysis of behavior. The richer the data source, the more accurate the data, the more available analysis methods, and the more accurate the results, but the basic idea is the same.
Second, further improve ideas, do so
Since the analysis idea is such a simple thing, why it always feels hard to find it? Note that the above example seems simple, but there are some basic premises behind it:
- Observed in place: We observed cold and shaking
- Understand the principle: we know that colds have these symptoms
- Action can be taken: we can decide whether to take medicine or not
- The results of the action can be observed: we can observe the effect of taking medicine
These premises often do not exist in real work.
- Data analysts don’t understand what sales/operations/products/after-sales are doing
- The data analyst only knows the number of runs and has no knowledge of sales/operations/product principles
- The business action plan does not know the data. I don’t know if there are any rectification problems, when and what they are.
Without knowing the above, it is naturally impossible to connect the analysis logic and business results, and experience cannot be accumulated.
This is like a person who has not been exposed to the rain or has no medical knowledge, sitting in an air-conditioned room and typing on a keyboard every day. You can ask him what will happen if he gets uncomfortable in the rain. You can't ask why. To think about this dilemma, it is inevitable to understand medical common sense. It's just like a doctor's diagnosis. It doesn't mean that the doctor has to check all the diseases, but if he understands the medical theory, he can diagnose the condition in an orderly manner.
Of course, the best state is to establish a process of setting data indicators→data monitoring process→data early warning problems→analyzing problems→exploring countermeasures→testing→validating assumptions→summarizing experience→circulating monitoring. The so-called "big factory experience" that many students envy is actually just that the process runs smoothly.
3. More complicated ideas, think like this
However, if you only understand medical theory, you can't become a real doctor. Because when I see a doctor in reality, I can clearly say: "I started to have a fever on the 27th, and it has continued to this day. During this period, my body temperature has not fallen below 38 degrees. On the 27th, I was caught in the rain. There are no other symptoms and no sore throat." ——There are very few such rational and clear patients.
Real patients always just hum upon entering the door: eh, eh, eh, doctor, I feel sick all over, please help me!
The doctor asked: Why is it uncomfortable?
Patient: The whole body is uncomfortable...
(╯‵□′)╯︵┻━┻
What to do at this time? Doctors can only communicate from 0:
- Ask the family: "Where is he sick?"
- Check medical records: "What is the medical history before?"
- Do testing: "Take individual temperature/test blood/take a film to see?"
- Asking behavior: "What did you eat? Where did you go?"
- Do a test: "Is it more painful if I press you?"
Approach the truth step by step.
This is not all medical knowledge, but more communication skills and reasoning logic. The professional doctors Mr. Chen met not only talked about the condition, but also patiently communicated with the patient about the medical insurance policy, family income, and parent-child relationship. These factors have nothing to do with the condition, but are directly related to the money, time, and energy the patient spends on seeing a doctor. In the end, it will affect the effect of seeing a doctor.
Data analysis work is very similar to this situation. Because the business department itself is the person who runs wild in the rain, when they get wet, and what they feel when they get wet, they feel it earlier than data analysts. Just like many people who have a cold choose to carry it by themselves, many business departments also like to choose: carry it by themselves. So the question that comes to the data analyst in the end is often "Oh, it hurts, I can't explain it clearly"-because it was clear that I handled it myself.
This means that data analysts must be like doctors, not only knowing how to add, subtract, multiply and divide, but they also have to be able to see and hear in order to get to the point of the problem. Good problem analysis is three-quarters of success. The problem can be clarified, and the follow-up advancement can save worry and effort.
Four, deeper issues
Of course, he can only shout: Pain is not the most annoying patient, at least he has the willingness to treat the disease and can communicate. The following categories are more annoying and dangerous:
- Empiricism: The doctor please give me cold medicine directly! I have 18 years of experience in colds, your diagnosis does not match my experience
- Refuse to face it: I was fine when I came, how come you are infected with the virus!
- Refusal to invest: You must not spend a penny on me, the diagnosis can be correct without blood test or filming! Otherwise you are a liar!
- Falsification of data: I secretly took a fever-reducing medicine, and then when I took the temperature of 36.7 degrees, I was confused...
It is these people who refuse to cooperate when seeing a doctor, and it is these people who kill the doctor when they die of illness - these doctors are the most annoying and the most hated.
Data analysts will also encounter such problems:
- Empiricism: The old man has been in the industry for 20 years, everything that does not fit my feelings is wrong!
- Refuse to face: Who said I did not do well! I do not do worse performance/the whole industry is not good!
- Refusal to invest: the project goes online quickly, so save the point/data governance is too troublesome, business must be fast! Stop doing it!
- Data tampering: artificially brushing traffic, brushing reading/activity rules to leave room for wool/change target, reference group
Of course, these people will blame the data analysts and say: "It's all because you can't make accurate, comprehensive, objective, in-depth, and clear analysis. You won't use artificial intelligence Alpha Big Dog!" Many students encountered this If you make things difficult, you will also ask: "How to analyze it".
Note: This is no longer a question of analysis. The problem now is that these people are messing around and trying to shake the pot. Use rational and normal logic to communicate with them. What is needed at this time is the means to deal with the enemy. Specifically, how to advance the project tactics. So when you really meet these people, please don't suspect that there is a problem with your own thinking, but look at how to seek advantages and avoid disadvantages, and strive for a good result, at least not in vain.
V. Summary
The basic process of establishing analysis ideas is:
1. Observe phenomena (market reputation, business feedback, indicator changes)
2. Combine principles (business logic + analysis logic)
3. Make inferences (establish assumptions)
4. Take actions (take business actions based on assumptions)
5. Verify assumptions (Inspection results, accumulation of experience)
6. Further analysis (continuous monitoring)
To establish an analysis idea, the assistance needed is:
1. Have a good data collection, data warehouse construction, and data governance basis
2. Have a good business communication mechanism
3. Have a basic understanding of business logic
4. Can observe the results of business actions
5. Distinguish non-technical issues and distinguish evil people/ Good people
Above, if you are interested, we will find a specific new business online scenario in the next article, and see how to establish an analytical thinking from 0 to 1. Stay tuned.