Many students said: There are many articles from 0 to 1, but in the face of complex problems, how to build data analysis ideas? First, the term "complex" has different meanings in different levels of data analysts. For Xiaobai, when the leader conveys orders, the words "model" are complex problems. When they hear the word "model", newcomers start to flip through "Watermelon Book", "Statistical Learning", and "Machine Learning". "Model" blood battle for 300 rounds. And experienced students know that these are not the real complications in the enterprise. Let's look at a specific example:
Scenario: In the e-commerce industry (mainly products such as paper books, video CDs, etc.), customer service leaders have very strong opinions on logistics leaders, believing that logistics issues affect customer satisfaction. However, the logistics leader said: All problems such as delayed delivery and damaged packaging during the delivery process have been dealt with. How could there be logistics problems. Now there is a demand for analysis, which requires the establishment of a comprehensive and detailed customer satisfaction evaluation index system.
01 What is a real complex problem
Question 1: After receiving this demand, which keyword will you Baidu?
- Evaluation index
- Customer Satisfaction Index
- Customer satisfaction indicators
- Logistics customer satisfaction index
Many newcomers think of this problem: simple. Isn't it just to build an evaluation index? You can see 8 articles of this kind online a day. I am also familiar with the term "customer satisfaction", and it is not a mysterious term like private domain traffic, precise portrait. So I started with the four keywords on Baidu and started working on one that seemed feasible. But the key question is: The question at hand is that you don’t know how to evaluate customer satisfaction?
Not! The immediate problem is that the customer service and logistics departments are engaged! This is the big problem. The so-called "customer satisfaction" is just a reason for arguing on both sides. If the standard of "customer satisfaction" cannot be agreed by both sides, no matter how it is defined in the book, as long as you throw it out, one of them will be sprayed to death. This is the number one problem. So this question should not be chosen at all. The first step is to understand where the specific dissatisfaction is.
Another newcomer said: Since the customer service is dissatisfied with the logistics, the customer service records the user's incoming call, there is a "customer complaint" item, just take out this indicator directly (as shown below)
This involves the second problem: customer satisfaction is an indicator with rich meaning but very difficult to collect data.
- What is satisfactory in the end?
- Is it true that customers are not satisfied with complaints?
- Is it true that customers will not complain if they are satisfied?
Not necessarily! Especially when it comes to logistics issues:
- Maybe the customer pretended to lose his temper just to make the customer service process faster.
- It is also possible that the customer is silent, but returns in the end! Return! Return!
- It’s more likely that customers call for consultation/suggestion, but lose their temper: why not
Relying on only one field: Complaints, cannot truly reflect the situation. For example, the "customer dissatisfaction" given by the customer service leader is the following scenario (as shown below)
This involves the third problem: how to truly identify customer complaints/non-complaints in various complex data. According to the statement of the customer leader, all customer calls must be transferred to text records + keyword filtering to identify the situation. But obviously this is too time-consuming and laborious, and a simple solution has to be found.
However, this involves the fourth problem: the work process of customer service has to be adjusted. If the work process is not adjusted, there will still be a large number of true and false complaints mixed in other calls, and the follow-up is still impossible to follow, the customer service will still complain endlessly, and the logistics still does not know where it is wrong. However, the matter of adjusting the process involves the business department's ability, willingness, and thinking. At this time, if someone pops up and says, "Aren't you doing data with artificial intelligence big data? We can't do it as usual. Duang can analyze it clearly. It must be because you are not capable." It's not that you want to blow his head.
Convex
- Conflict of interest
- The meaning of the indicator is unclear
- The original data content is messy
- Related procedures need to be changed
These are the real difficult problems in the eyes of old birds. However, this is also the real business scenario of the enterprise. The data is perfect and the meaning is clear. The thing lying quietly in the excel sheet waiting to be modeled, only exists in the online articles. The reality is that various interests are entangled, data is mixed, and the process is unclear. What should I do? ? ?
02 How to build analytical thinking
Summarize the problem this time. On the surface, it looks like: customer service feedbacks many logistics problems and low customer satisfaction. Looking deeper, customer satisfaction and logistics are not uniform in terms of customer satisfaction, which makes it impossible to solve the problem. Looking deeper, many of the customer's problems are not caused by logistics, but they are blamed on logistics. The customer service does not make a distinction, but hits the door with a brainstorm.
Under such circumstances, there are three solutions:
First: neutral judge. If you get a higher-level authorization, or if the two departments can talk calmly, hoping that the data department will stand in the middle as the judge, you can use this idea. At this time, you can focus on the customer dissatisfaction feedback from the customer service, sort out the problems level by level, and clear out which ones are true and which are false complaints:
Second: pretending to be Xiaobai. This idea can be used if the two departments are unable to fight each other and are determined to fight. The data department is like a little white rabbit with no injuries to humans and animals, saying: "You see that we don’t understand the logistics business, nor the customer service business. If there is a need to distinguish which calls are unsatisfactory, the business can give specific distinction rules. We extract data according to the rules".
Yes, let the two quarrel with each other, and decide exactly what is dissatisfied, where and according to what standards, the data department will act as a counting machine. And only give data, not judgment. This is very embarrassing, but can protect yourself in the departmental melee.
Third: solve the problem. Note that customers always want more benefits, so complaints about true and false customers are unavoidable. However, it costs money to improve the distribution capacity of logistics. Just like all the bosses say they want to improve customer satisfaction, but you ask him to spend 10 billion to increase satisfaction-in all likelihood, he disagrees.
Therefore, from the perspective of problem solving, the first step is not to establish customer satisfaction indicators, but to first define logistics service principles, such as how long the longest delivery time is, and how much is the delivery damage rate, and so on. With this standard, customer service can be promoted in the second step. When responding to customer complaints, first distinguish whether there is any violation of the service principle. If there is, the logistics execution is not in place, and transfer to logistics processing; if not, you have to rely on customer service efforts, or appease customers, or explain principles to customers. In this way, everyone can maximize the solution to the problem within a limited cost.
If you use problem-solving ideas, the analysis needed is not one to establish a customer satisfaction index system, but three coordinated analyses
- According to the logistics principle, the current customer situation analysis is not in place
- Based on the logistics principle, the analysis of customers' true dissatisfaction and false dissatisfaction
- Based on the existing customer service appeasement methods, analysis of the final handling of true/false customer dissatisfaction
The complexity of analysis is greatly increased. In fact, problem-solving-oriented analysis logic is very complicated and depends on the business processing capabilities of data analysts.
03 Summary
You will find:
Generally, the data analysis ideas in online articles are neutral and judge-style. The authors like to think of themselves as the biggest boss and give pointers to the country, which is really damn good.
In general real work, they all pretend to be noobs. "Please think through the business yourself" "I'm just a data-runner, I don't understand anything"-in the end, people often scolded "useless" and "what did you analyze".
Generally, when bosses solve problems, they use problem-solving thinking. What can be thrown to data analysts is three independent access tables. The students who ran count still don’t know what they are doing
In fact, the three methods are all right to look at individually. The difficulty is not to do a certain method, but to judge the situation, combine the real problem points, the current system data, and the determination to deal with the problem, and choose a practical method. This requires data analysts, if they really want to participate in solving problems, they have to observe the situation and build ideas from problem communication, meetings, and chats, instead of just grabbing a keyword and Baidu as the beginning.
However, some classmates will say: Teacher, my leader firmly mentioned the word "model", it feels so difficult to do. I worked so hard to make a model, but the leader said, "It's useless!" "I'm not talking about this!" What should I do? If you are interested, we will share in the next article: what is the true meaning of the model in the leader's mouth, so stay tuned.
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.