Reprinted Source: https://mp.weixin.qq.com/s/XTnjVI1m1QPzHyeMbGEjIw
The data do students have often heard saying: "Building Sales / Operations / product data monitoring system." This thing apart to see each word know, heard together face mask ignorant circle, often ask:
-
What is a data monitoring system?
-
Data System in this stuff and what's the difference?
-
I did these two reports, in the end count system?
-
Why I did not feel anyone is "monitoring" the? ? ? !
These questions, today the system to answer. The reason why students have doubts, because a majority of companies do not really pay attention to the data, the students worked hard to make the report a few people did not see. Secondly, a large number of students will only respond when a huge problem has occurred +, the lack of ability to judge beforehand. Based on the data lead to even mention the views, the service went back to the sentence: "I knew it" (see below)
1. What is the data monitoring system
Monitoring, by definition, supervision and control. In theory, can not supervise and control data, such as the traditional workshop director, production team leader, invigilator, are the front-line on-site supervision and control. But Enterprise Group headquarters thousands of miles away, and there is no way to send clairvoyant ears of site supervision. Therefore, there is a data monitoring: to monitor and control data through the index. When the business becomes complex, a single data can not meet the monitoring requirements, and therefore there is a data monitoring system.
But it does not mean Data System. Data System in a simple supervisory role and can not control. Data do students know best, to issue reports every day, ten men to have a person look pretty good. Even data line of love to ignore the answer, without mentioning the supervision and control. So simple data index system is just a tool, how to combine tools and management processes, data is the most difficult part of the monitoring system.
2, how to build a data monitoring system
As the name data monitoring system, monitoring + control, so build data monitoring system contains two key tasks:
-
Establish a data index system, the business conduct supervision.
-
The application data management processes to achieve control.
Before sharing has been how to create a data index system, we can see " what data analysis systems are? How to build? "Today the focus of talk about how to control the control method.
Many students do a technical background data, the mention of "control" the most intuitive can think of is to criticize their own to get up, he got married reminders born baby's maid. Then emotion: I never did, "Sales / Operations / Products / risk control, how do I control ah ......" In fact enterprise management completely without such a trivial tangle. Just as we will not need to build cars, but also to control the car (driving a car), the key to solve the following four questions:
▌ first clear who should control
Point action must be clear. Many students will say: "GMV down, and engage in high" That just did not say the same. GMV appears to have a relationship with each department, the company-wide thing. But everyone is responsible, everyone is equal no responsibility. Too macro target point is unknown, naturally can not play the role of control. You can call specific points, such as:
-
GMV drop → → high-end users consume fewer users think of a way operators
-
GMV drop the big decline of certain goods → → think of a way Merchandise Operations
-
GMV drop too low → → external traffic channels serving think of a way
In short the overall objectives, specific to a department, a team, it is best to hang directly Ah KPI / OKR index, clear this point, to be effective. To achieve this goal, when the index system construction data, do the dimension of classification can not be free to do, but to and division of labor relating to the classification dimension (such as branch offices, merchandise, customer hierarchy etc.) stand out, easy to implement Responsible.
▌ second, control Shashi
If you really wait until the fall GMV die before to shout: engage in high, too late. In fact, many factors affect business trends can be predicted in advance, such as:
-
Positive factors: a big promotion, new products, the traditional peak season
-
Negative factors: the system BUG, out of stock, the traditional off-season
-
Uncertain factors like: system revision, season
Thus, there is a control time status and to determination of. This four standard terms are:
-
Over the past XX + negative attention to the problem →
-
XX + past experience found positive →
-
XX + future risk to guard against negative →
-
Future Forward → Tips + XX opportunity
Therefore, when building data monitoring system, we have to summarize past experience, to understand the future of internal plans, and even collect some competition, so the situation early to judge ready to rush in advance multi shouting "vigilance", "tips" rather than after all have seen, just yell "engage in high"
▌ Third, how much control efforts
This time you can call "engage in high" up! Is a drop, yes, was a Chen ridicule "engage in high" itself is not the problem, is a standard recommended way all the way. When given control of opinions, talk about the direction, repeat efforts, and finally the details. The severity of different issues, corresponding to the intensity is different.
There are also standard, then surgery:
-
Pay attention (transaction occurs)
-
To Action (engage in high and pursue low)
-
To immediate action (!!!)
If you do put in place to assess the problem, after the immediate action is further described. For example: "If the mass flow rate can not be improved within 3 days, this month will KPI non-compliance." Do this kind of speculation and prediction, and therefore so many students are tangled, predicted right in the end. There is absolutely no need.
In management, mobility is more important than accurate forecasts . To predict a problem, if the business action to force the issue early on is finished processing, where also what precise prediction. If the business does not act, he has been entangled: "I see no problem now." "If it natural bounce of it," and finally will miss the boat, it would definitely flutter Street, also need to predict?. So often do estimation is the most simple logic, such as the performance of 30 million a month, it has 1 million a day. 10 million traffic into 200,000 to buy, so you have to buy 50,000 2,500,000 flow, similar. Control is done in order to promote business operations, not to find reasons to excuse, remember to remember.
▌ fourth, complete control of what effect
Do the most important control is the effect. The effect is hierarchical:
-
Primary: control traffic is received
-
Intermediate: Business Press control action
-
Advanced: actions have a significant effect on the index
Build a data monitoring system, the most part is the result of no less review. And the effect of which is to look at the level, look at the specific effect size. Often it is impossible not to index the last change, but the business side are they heard suggestions last basket off by disclosing to panic just came to ask "why." The ring appears to be simple, actually difficult to do. A cocky attitude to many businesses, too lazy to communicate. And secondly, a lot of students do data thin-skinned big stuffing, too thin skinned afraid to communicate, focusing instead engage in a variety of algorithmic models, only you know. This makes it difficult to achieve smooth operation of the monitoring system.
Finally, it is emphasized that: data monitoring system is used to find the problem, not to solve the problem. Surveillance monitoring, supervision and control, he was not actually teaching and discussion. Have you ever seen Which slave overseer and a textbook in the field to supervise work, are waved his whip hand "pop!" Sound. Some students do extremely complex to monitor, cram a bunch of indicators want direct reading problem analysis, this would greatly slow down the monitor operating efficiency. In the direction is not clear, the more we look at the data, the more chaotic mind.
3, why do not you data monitoring system
Common problems, such as:
-
Lack of criteria for judging Data System. Only know that shouting "up," "down," and I do not know reported "good", "bad"
-
Data indicators index really is. And management processes out of touch, there is no clear responsibility, there is no corresponding people to follow
-
Lack of data other than traffic grooming. I did not know what, do not know doing, do not know doing
-
The lack of follow-up questions confirmed feedback. I do not know other people have not heard, I do not know people did not, did not know people did not become dry, light daze staring at the numbers.
-
Greedy big multi-plug a lot of indicators, colorful look tall on. The results run a few tired like to see the number of weighs.
Essentially data monitoring system is the thermometer. Body temperature is important, but reflect the body temperature fever is more important . Want to understand the symptoms, is much more natural to study medicine, was found after people rushed to hospital fever cough, fever patients have to understand who contacted, been there, have to do further tests rather than grabbing hold thermometer. These are the real fell to the ground so that the data monitoring place. It has recently been measured thermometer to measure to which the taste, you goods, you fine chemicals.
