How to accurately identify master data?

What is master data ？

MBOK The definition in is like this, aunt ：
Master data is related to business entities （ Such as employees 、 Customer 、 product 、 Financial structure 、 Assets and Location, etc ） The data of , These entities provide contextual information for business transactions and analysis . The entity is the object of the objective world （ people 、 organization 、 Places or things ）. Entities are called entities 、 The instance is represented by data / The way of recording indicates .

Found no ？ Master data is related to entities ~~~ So the master data is actually closely related to the model .
In fact, understanding master data is very simple , Simply speaking , In the core business , Non numerical key data . This understanding is not accurate , But it's easy to understand .
But this buddy's problem , Obviously, this article can't solve , Because he must have encountered ambiguities when identifying master data , It's impossible to distinguish .

Master data identification

Want to confirm whether the two contents are master data , We have to start with the definition of master data , Start with the characteristics of master data .
Of course , There are also some side door techniques that can assist in recognition .
Shixiufeng, teacher Shi in his book 《 A book on Data Governance 》 There are two methods in it ：
1、 Master data feature recognition method
2、 Business impact and sharing degree analysis matrix method
Master data feature recognition method as the name suggests , Just compare the characteristics of the master data ：

Let's see if these data have the above characteristics , If there are both , So . If oneortwo are missing , Think about .
Generally, it can be classified as 6 A question ：
1、 Whether it reflects the core value of the business ？ This is very, very important ！（ The customer information must be , But the province where the delivery address is located is not the core value data ）
2、 Whether it is an independent entity ？（ Commodities are independent and indivisible entities , But temporary goods are not ）
3、 Whether it is relatively stable ？（ The reason why we add relative , Some master data will change , Such as customer information ）
4、 Whether it will be shared in other systems ？（ If only a single system is used , Even core values , Generally, it will not be listed as main data ）
5、 Is it unique ？（ If this data is not forced to be unique , The whole situation may repeat , Then you can kick it out ）
6、 Whether it works for a long time ？（ For short-term use , Generally, it is not used as master data . But this long-term and short-term business related , For example, the timeliness of Internet orders is different from that of shipyards , The high probability of the former will disappear after half a month , The latter usually lasts for several years ）
As for the business sharing matrix method , In fact, it depends on the importance and sharing of this data ：

According to the degree of importance and sharing , The priority level naturally comes out . As for those that are neither important nor shared , Naturally excluded .

Differentiate challenges

Although there are already methods , But sometimes I encounter unfamiliar businesses , It will still be covered . Generally, reference data and master data are easily confused .
Because the reference data has many characteristics very similar to the master data , For example, it is also effective for a long time 、 Share across systems 、 It's also important. （ The value is not necessarily high ）、 Very stable 、 Global uniqueness, etc .
Some familiar ones are ok , But once linked to the business , If you don't know the business , There is almost no way to separate from the master data .
DMBOK The difference between the two management priorities is put forward in ：
For reference data and master data , The focus of management is different ：
1） Reference data management （Reference Data Management,RDM）. You need to control the defined field value and its definition . The goal of reference data management is to ensure that organizations have access to a set of accurate and up-to-date values for each concept .
2） Master data management （Master Data Management,MDM）. The value and identifier of master data need to be controlled , In order to be able to cross system 、 Consistently use the most accurate of the core business entities 、 The most timely data . The objectives of master data management include ensuring the accuracy and availability of current values , At the same time, reduce the related risks caused by those ambiguous identifiers （ those Identified as entities with multiple instances and those involving multiple entities ）.

As for how to distinguish the two , In addition to distinguishing the value of data , There is also a more ingenious way ： Look at the number of fields in the table and the amount of data .
Generally speaking , The data set of reference data is usually smaller than the transaction data set or master data set , Low complexity , Have fewer columns and rows .
So if you see one 3 Data table for column , It is impossible to distinguish whether this is master data or reference data , Then it must be right to blindly guess a wave of reference data ~~~
Of course , More important distinction , Or to see its value , Is the degree and value of participating in the core business . And whether the data is completed “ Identify and manage relationships between data from different systems and processes ”.
Reference data will be reflected in the core process , For example, the province where the delivery address is located , But the degree is relatively low , The value is not particularly great , It doesn't even matter if it's lost （ It can be deduced from other data ）.
and , This data also eliminates the need to identify and manage the relationship between the data of the continuous system and process . therefore , It is no problem to list it as reference data .

Summary

Master data is easily confused with reference data .
There are many ways to distinguish , There are two conventional methods ：
1、 Master data feature recognition method
2、 Business impact and sharing degree analysis matrix method
There are many unconventional methods ：
1、 Management focus differentiation
2、 Field 、 Data volume judgment method
3、 Data association method of different systems
4、 Empirical judgment