Several promotion routines of data governance

One 、 Top level design method

seeing the name of a thing one thinks of its function , The top-level design method is to make a plan for the top-level design of data governance first , Then follow the plan .

Peng you, who has done consulting, knows , The top design 、 Strategic consultation will be based on strategic objectives KPI, Then set up corresponding supporting projects , And sort according to priority , Finally, an execution path is formed .

What to do this year , What to do next year , What to do first , What do you do after , The planning is clear .

Then just follow the chart . The general logic is like the following figure ：

The benefits are obvious , First there's noodles , And then wired , Finally, there are various dot like items , A little bit of implementation , Naturally, there is no effect .

But such a plan is very, very extravagant , Because this scheme works slowly , The requirements for the organization are very, very high . There are few organizations that can stand it , It usually has to work quickly .

Basically, only a few government units and a few enterprises have used this method to achieve the success of data governance .

Two 、 Technology driven approach

Some sensitive friends have noticed it , This is called “ Technology driven approach ”, Not technology led or something .

In fact, this method is the data governance method adopted by most enterprises . Why , It's very simple , Because most data governance projects are initiated and implemented in the information department .

Since it's the Technology Department , Of course, the technology department promoted . Speak true , I've seen too many similar things , There are few effective .

《 Huawei's way of data 》 I said yes “ Business leading ”, That's true , But little has been done . The reason is simple , Ass determines head . The main responsibility of the person in charge of business is to engage in business , It's impossible to take the initiative to do data governance .

There's nothing to say about technology driven routines , It's about data , Solve from the technical level . Routine is the logic of information system construction , Make an item , Do research , Various outline designs 、 Detailed design , All kinds of development 、 Integrate 、 test 、 Deploy , Then accept .

Effect , General! . Because most of them are problem oriented , frequent “ patch up ” Style construction . In the end, it is often all kinds of explosions , Report explosion , Index explosion , Data explosion .

Then start the index system 、 Data quality system , Patch by patch , In the end, no one dared to move .

in the final analysis , Because the problem of data is systematic , Technical reasons are just one of them . The reason for this phenomenon is that the business participation is not enough .

In Enterprises , Who makes money , Who has the right to speak . Business is naturally a profit center , And technology is generally a cost center . Let pure technology drive data governance , It's like asking your son to urge your father to quit smoking .

3、 ... and 、 Apply traction method

If technology promotion is a child cart , So the application of traction is easy for strong cattle to pull cars . There are applications in front of traction , The latter things seem very natural .

It is not unreasonable for many enterprises to build data systems and like to get a large screen first . Because no “ use ” What is valuable is worthless .

Although the users of large screen are relatively single , The practical value is relatively low , But after all, there are still usage scenarios , What is stronger than the pure technology development and construction without use scenario is not a bit .

Based on data application , Reverse requires high-quality data supply of each link , Promote the construction of data governance system , It's also a good choice .

But doing data management in this way , Will always fall into one-sided 、 A partial victory . There are applications , Data quality can be managed , No one cares about the data quality without application .

Four 、 Standard antecedent method

I have only met a few real cases of standard current law , Because it needs to be set in advance and the data standards must be observed when building the business system

5、 ... and 、 Regulatory driven approach

Strong supervision is usually the policy issued by the superior unit , Subordinate units execute . And I can't do it well , There will be punishment .

Bank 、 Industries with strong supervision such as insurance follow the policy . Do not do data management well , Not according to EAST、1104 Submit data as required , The ticket will come soon .

Don't try to fool , If you have the ability to make a full set of false data , The fake one is the same as the real one , The cross check relationship between statements is correct , That can't find flaws in all dimensions . Yes, of course , In fact, this strong supervision mode can also be implemented within the enterprise , But it takes “ Privilege ”. This premise is usually difficult to achieve .

There is a clever way , Namely iso . For example, the state is pushing DCMM iso . There is a special advantage of implementing standards , Is to put “ Standard implementation rating ” List in the organization's annual goals , In this way, a huge enterprise can be formed within the enterprise “ potential energy ”, Form a strong regulatory trend . When we put “DCMM iso ” This big stick waved ,  Naturally, it is much stronger than a certain department or several departments to promote data governance .

We do... For an enterprise DCMM When implementing the standard , The technical department has established a system for the early discovery and issuance of data 、 technological process . But like most companies , After sending it, it will become an empty paper . The business feels that security control is too cumbersome , Not at all .

It's not the same now , The technical department borrowed “ iso ” The reason of , The business is required to implement the previously issued systems and processes . Although the business is unwilling , But implementing standards is an enterprise level goal , You have to do , It's just Half push and half push 了 .

At the end of the day , Regulatory driven approach , Is taking advantage of the situation , Take advantage of the trend required by the superior policy , Take advantage of the trend of national standards . Use the general trend to promote the departments that could not be pushed , Dredge the process with high resistance .

6、 ... and 、 Quality control law

The quality control method is actually no way , It can also be regarded as the early prototype of data management . Because speaking of , The theoretical system of data management goes back , Actually, I came to From the quality management system .

ISO9000（ Quality management standard system ）、TQM（ Total quality management system ）、CMMI（ Capability maturity integration model , Not just software ！）, All belong to the general management system .ISO9000 Later developed ISO8000（ Data quality management standard system ）,TQM Extend out TDQM（ Comprehensive data quality management system ）. and CMMI The association is also 2014 Introduced in the DMM（ Enterprise data management capability maturity model ）. This is the quality management system in the data field .

China refers to CMMI Wait for a crowd of data management systems , stay 2018 The data management maturity evaluation model was officially released in （DCMM） National standard , This is a later story .

As in other industries , Quality cannot be bypassed . Whether it's business , Or technical , I believe you, Peng you are no less Scratch your head for data quality problems . There's something wrong with the quality , The data won't work , It can even affect wrong decisions .

therefore , Forced by various data quality problems , The internal and external departments of the enterprise take it seriously , Gradually solve the problem of data quality .

 

Data quality control is obvious , It's problem oriented . But you can't treat your head and feet with headache , There has to be a methodology .

Generally speaking, there must be a specific demand , Including data quality control objectives 、 Evaluation criteria 、 Decision rules, etc .

Then proceed from the phased objectives and needs , from Take precautions 、 In process monitoring 、 Ex post verification Three aspects of quality control , Solve all kinds of data problems .

When solving , Generally, a system for improving data quality will be established special , From a technical 、 technological process 、 The system 、 Mechanism and other aspects improvement , On a regular basis assessment , Establish data quality problems and Solutions The knowledge base , It is convenient to quickly locate and solve similar problems in the future .

In the process , Take the problem of data quality as tow , Comprehensive use of metadata 、 Master data 、 Data standards 、 Systems and norms methods ,“ build ” So as to use , Naturally, there will be no situation that it can't be used .

7、 ... and 、 Interest driven approach

With Benefit sharing For the root , With “ achievement ” Oriented , Establish a system that meets Some core personnel Benefits The goal of , Then just push it .

There are many specific operation methods , For example, successful case law 、 Cooperative winning method 、 Award method 、 Calligraphy 、 Conference law, etc , There is also the life-saving law for Internet enterprises “ Open source method ”.

8、 ... and 、 Project construction law

This is easy to understand , Just get a data governance project , Build slowly .

In fact, data governance has been carried out until now , It has also formed a set of very perfect processes , Relevant product capabilities have also been very comprehensive .

The project I was involved in , It basically covers the whole process of data , What data consultation 、 Data collection 、 Share exchange 、 Several positions 、 Data standards 、 Metadata 、 Master data 、 Data quality 、 Data visualization 、 Data analysis, etc .

At present, the effect is better , It is a combination of consultation and implementation .

Do a consultation , Check the current data status , Fully grasp the future strategy and current situation of the enterprise , Then according to the data management system , Make a gap analysis , Work out the specific tasks to be performed , According to the time schedule , Dismantle and plan the project .

  Then in the implementation of the project , First penetrate a scene , And then slowly expand the results from both the in-depth and horizontal levels , Build metadata 、 Master data 、 Index system 、 Data quality management system, etc , Continuously consolidate data infrastructure , Provide high-quality data supply for front-end data applications .

Nine 、 summary

In fact, why are there so many data quality problems ？ It's simple , There is no standard . Without standards, there is no right or wrong , Naturally, it will be in a mess ！ The standard has , You can determine what's right , What is wrong . Later execution 、 Monitoring and control have a basis , Data quality is guaranteed .

Actually do things , It is nothing more than to use all kinds of forces . Or use your own strength , Or use someone else's power .

As long as there is room to show , Things can still be done . I'm afraid I won't give you space , Strength does not support , That's a dead end .

Reference material ：

1. WeChat official account （ Big data architect ）-《 Data governance work 8 A propulsion routine ( On )》

2. WeChat official account （ Big data architect ）-《 Data governance work 8 A propulsion routine ( Next )》