"Circles" encountered in data analysis

Unlike a decade ago, what makes a data analyst confused today may not be the little data, but the data; today is not the importance of playing good data, but do not know the harm of the wrong data, that is, the so-called sweet troubles. A data analyst, if you can understand the current data on the core issues, and can clear the solution, you can improve.

This time I want to talk to you about a circle (Feedback loop) that is composed of the four parts of the subject, thought, line, and knowledge, pushing each other forward.

It sounds a bit iffy, but it might be interesting to combine the examples.

First, by

"Suffer" is the feeling of the world around you

At present, as an electric business, to feel the way the company's business is increasingly dependent on data, but today, very few dealers are sure that they have more complete data to grasp the company's situation, there are mainly two reasons:

The first is "blocking", when many electric dealers start to collect data, they find that the data are distributed in different places. To give a simple example, some platforms do not know the number of complaints to accumulate to what extent, because the complaint channels have telephone, mail, micro-bo, there is no uniform caliber, not to collect, or even if there is data, only in the hands of specific operators, can not reach the hands of managers in time, If a company's data only a few people can see, can not be uploaded in time to release, imagine the data driving force is how small. Believe that plugging in the situation in the middle of the electricity business is very common, if the team allows, of course, to tidy up, otherwise it is like a fight with eyes closed.

The second is "scattered", the lack of http://www.aliyun.com/zixun/aggregation/8600.html "> Data operation experience, only know to data, but do not know what kind of data, or empty data can not be." Just a few days ago, there was an electric dealer and I said the operation of their company data, he is a traditional enterprise, collected a lot of "scattered" data, do not know what to do. I give advice is in a pile of scattered data, from their own areas of expertise, such as traditional brand electric power supplier to the supply chain data is very familiar with, can from the most familiar supply chain data to interact with other data related to see, such as a product sold 10000 orders, But there is no thought of the association to the number of people to see but did not buy the data collected, transaction data and foreground browsing data can be linked to find more new problems.

Second, to

"Think" is the recognition and discrimination of things

Data kill people, this is the view of some of the electric business. The most direct reason is that their data is correct, but the objective of the data by the subjective too much influence, and lead to the actual data failure, but to the power to point to the wrong direction. such as not to remove the noise in the data, such as the source is not right, or subjectively want to find not objective data, such as the product manager, is developing a higher charge of new products, the Product manager will continue to convince themselves, desperately in the data to find some you think the value of users, and finally more and more deviated from the objective reality. When the analyst encounters a "subjective" problem, the wrong judgment is inevitable.

Understand this, you can understand that the same data in different people's eyes why will show a very different image, so look at the data, you need to think from multiple roles, and at this time, you will encounter the problem of "away":

We say that the data is not uniform standards, the conversion rate, each part of the numerator and the denominator are not the same, the market department said a channel conversion rate, the website operation said a page conversion rate, exchange up people do not know each other said conversion rate is what, nature is far away. What's more, data and managers have a different definition of data caliber, and managers get data that is far from what he intended, and he uses this data to make decisions and see how it works. This shows that if a company's data standards are vague, can be expected to see how the data operation is difficult to achieve, may be confused by the data.

Third, the line

"Line", that is, the deep thinking and analysis of things

The previous example is better understood, this is good sailing, nautical chart information is not allowed or direction unknown, naturally it is difficult to reach the destination. But in data analysis, also often appear the problem is "astringent": Clear direction, nautical charts accurate, but in the specific navigation process is not working, due to the business understanding is not deep enough, resulting in the analysis of confusion and chaos, the way the data use, ultimately managers can not rely on data analysis for decision-making.

An example that is often encountered in the actual operation is that the repeat purchase rate is reduced. As mentioned above, the 1th question for the data analyst's brain is: what is the definition of the repeat purchase rate, what the denominator is, what the molecule is, and what is the definition of the time dimension to see the repeat purchase rate. But after reading this, there may still be an error, because the complexity of the business is not taken into account. For example, the day a customer at the same time under the two list, is the repeat purchase, or a list? Have new customers grown a lot lately? Is there any change in the percentage of new channels? Haven't you had a promotional campaign for a long time lately? Is the page layout adjusted?

Therefore, it is the actual action, need to put multiple impact indicators into the coordinate system, and the correlation of data is built, will not blindly to repeat the purchase rate decline and worry. To the "line" at this stage, must understand business, or naught.

Iv. knowledge

"Knowledge", that is, the fundamental induction of things, the precipitation of the understanding of things

The last link of being, thinking, doing and recognizing is knowledge, here began to relate to the data analysis of a major problem: "Leave", the analysis of the results and methods can not be summed up in a timely manner and precipitation, the electrical business can only with analysts, and constantly chasing, explore a new perspective of analysis, but may not be useful to the actual operational decisions. As far as today's trend is concerned, analysts ' resources are becoming more and more valuable, and the electrical business needs to think about the appropriate mechanisms and tools to accumulate the results of successful analysis and experience, and apply it to the day-to-day management of the company and turn knowledge into "money" in a timely manner.

It is best to build a system by accumulating and settling knowledge from the data. Perhaps many electric operators think that the establishment of the system is very troublesome, but the actual operation is not cumbersome, but to build ideas difficult to think clearly. More than 10 people three months, you can put the company operating core data in the system, the data analysis of the concept of the system inside, so that all companies can understand and use. There are three key points: one is to do data security, so that people in different positions see different data; second, it is clear that the data standards of different departments, such as the conversion rate of the market is used to refer to the number of customers in the order divided by the total number of customers come in, the conversion rate of the financial department refers to the number of successful customers divided by the total number of customers come in Third, the data can be linked to different departments (if the electricity business has a team can do so), so that data can be used to spread outside the data department.

Here is my personal little experience, from the subject, think, line, know four parts of the circle to help me clarify the data analysis of the four links, and data analysis often need to run data repeatedly, each time reuse, get more harvest. If you want to speak deeply, you have to say it separately. This process, I am also in groping, welcome to share their own methods and thinking.

SOURCE Address: http://blog.sina.com.cn/s ... 0100xo6l.html