Almost all data analysis books, tutorials, and articles are politically correct to write MECE as the criterion for data analysis. This thing quickly became a "crumbling" in the data field:
- What exactly is MECE
- Why must MECE
- Am I doing MECE
No one answered a lot of question marks. Today we will explain them systematically.
1. What is MECE
MECE is the abbreviation of Mutually Exclusive Collectively Exhaustive, which is simply summarized into 8 words: mutually independent and completely exhaustive. It sounds terrific, and if it can be exhausted completely, then the analysis must be impeccable. The question is: how to do it? Here is the most intuitive example:
Problem scenario
Little sister Zhang Yuxuan from the project team is late again! I was 19 days late last month! Originally, my colleagues were late for two or three days, and HR wouldn't care about it, but 19 days late this month was too exaggerated, so I told the leader.
My sister paper cried so Pear Flower complained with rain:
- Many people in the project team are late, and you don’t care
- Then you don’t care about traffic jams during heavy rain
- The project is coming online, but it is late
- I worked overtime the day before, it would be late
- ……
Question: How to analyze the actual situation?
2. The consequences of not using MECE
Novice data analyst, immediately take a pen in this list:
Reason 1: Traffic jam
Reason 2: Rain
Reason 3: Project goes online
Reason 4: Overtime
Reason 5: I’m lazy
Reason 6: Other
If they do so, they will soon discover that various reasons are intertwined! When it rains, traffic jams will naturally occur. Overtime and project launches often overlap. Laziness and all problems overlap. Then it might rain, work overtime and be lazy...it's hard to tell how much it affects. Even if it is clear, there is another one waiting. So I couldn't help but want Baidu's "How to Build a Multi-factor Independent Inspection Model", or go to the WeChat group to ask, "Is there a HR analysis leader from Tooteng's company, urgent, online, etc.!".
This is the disadvantage of not using the MECE rule to disassemble the problem: data analysis is illogical and becomes a pure cross-tabulation. In fact, many newcomers do this. When they encounter problems, they pull out the dimensions of channels, time, products, users, etc., and cross the problem indicators one by one. After crossing, see which column is low: that's it! Finally, I was questioned: How to distinguish when multiple reasons are intertwined? I can't answer.
3. How does MECE operate?
The first step of MECE: Determine the goal.
Note that in reality, multiple factors are entangled in one problem, so how to classify the problem first depends on the goal of the decision: if you want to kill them all, you can let them go. For example, the problem at hand, the first thing to do is to make a clear distinction: whether you want a sister paper or a sister paper.
Diaoren: Strictly demand, as long as you participate in a little personal factor, it is your problem!
Helping people: lenient requirements, as long as they can be explained by external factors, they will not blame the individual
By clarifying the goal, you can grasp the scale when multiple factors are mixed, so as to avoid deviation of thinking and go straight to the core problem.
The second step of MECE: sort out the problem step by step.
Note that they are mutually independent and completely exhausted, which is the final result of MECE operation. It does not require one step to the end, all the reasons can be exhausted in one click. When decomposing each layer of causes, using dichotomy is the most convenient way to achieve mutual independence and complete exhaustion of requirements. Therefore, there can be many logical levels for analyzing the problem, but the indicators for each layer are as few as possible, and the division is clearer. .
For example, if it is set: help people, see if there is really too much work, this big goal. Then when decomposing the problem, the first level can be: overtime/no overtime. This is a two-category, it must be independent + exhaustive. Then mark all the days with overtime work records on the previous day as: overtime work (as shown below).
At the second level, we can distinguish whether everyone is working overtime collectively or only one person is working overtime. This is another two-category, which is still independent + exhaustive at this level (as shown below).
The third layer, working overtime by oneself, may also be caused by too much work, so this layer can be broken down again (as shown below).
Note: The ultimate goal of analysis is to guide business improvement, so the analysis logic should point to where the business can land. For example, the dismantling of this logic does not consider the weather at all. Because the amount of work can be arranged by the leader; the law does not blame the public, but the leader can accept it. When discussing within the scope of business initiative, try not to involve factors beyond the control of the business, so that it can directly lead to a useful business conclusion. Instead of talking about the weather, the final conclusion is: "Please learn the magic of how to call the wind and the rain"-this will definitely be criticized.
Similarly, when decomposing the second branch of logic, now that the big goal has been set: to help people. It can be justified by reasons such as "it's raining and traffic jams in the city". Note that there is another trick here: choose the cut-in dimension, choose the quantifiable dimension.
For example, it's raining, "it's raining, traffic jams in the city" sounds like a good reason, but:
- How to quantify rain?
- Light rain, heavy rain, heavy rain?
- Light rain is also in traffic?
- Must be late in a traffic jam?
These are difficult to quantify clearly. So you can change to a simpler quantification method. "It's raining and traffic jams in the city" point to the result: "Everyone will be late." Then just look at "Are you all late" (as shown below).
The third step of MECE: Substitute data to quantify.
To do data analysis, not to talk about the data is to be rogue. After the classification logic is determined, the data must be filled in according to the logic, and finally the data speaks.
Talking about toxicity out of measurement is to be rogue, so after substituting data, we must first look at the proportion of each type of problem. The problem ratio itself can explain the problem to a large extent. This is also the biggest advantage of using the MECE method to disassemble the problem: to avoid being distorted by an example, everyone looks at the number and talks (as shown below).
The fourth step of MECE: derive business conclusions.
Finally, the business conclusion can be derived. The business conclusion contains two aspects:
First: Judgment from the overall structure. Is Xiaomeizhi laziness subjective or objective reasons?
Second: a small conclusion for each detailed problem point. When it comes to counseling, I should simply let it go.
After inferring, you can directly set observation indicators and continue to observe the trend of the problem. Observations include:
- Quantity change: Did the number of days late have decreased?
- Structural changes: Is the number of days reduced due to objective reasons?
- Refinement of the problem point change: Is the number of overtime days due to work distribution reduced after the workload is reduced?
In this way, by analyzing logic + business actions + data tracking, we can clearly see whether the problem is solved under the guidance of data, so as to achieve a good data-driven effect.
4. Obstacles that prevent MECE from working
For data analysts, the biggest obstacle comes from: do not understand business, do not communicate with business, do not promote business. Think of data analysis as a homework. Anyway, there are several ready-made dimensions in the database. I made all the comparisons. If the one is low, I will say which has the problem.
For the business side, the biggest obstacle comes from: being independent and not looking at data. I caught the headlines of a few cases, how to use data to quantify, how to use data to assess, nothing is discussed. Either just stay still and rise to the attitude level, let alone quantitative assessment.
The advantage of data analysis is that it can fight against judgment errors caused by individual cases and emotions in business development. Therefore, carefully sorting out the business logic, clarifying the goals, and deriving step by step until the landing monitoring is a good way to see the day. Of course, after reading it, some students will say: Can you give me an example of operation? If you are interested, follow the public account [Ground Qi School], and we will share in the next article.
Author: Chen grounded gas, micro-channel public number: down to earth school. A data analyst with ten years of experience has launched a series of data analysis courses and has more than 20,000 students.