At the beginning of many reflections on data use, optimization or valorization, there is often a mapping approach. Indeed, how can we manage something whose volume or complexity we can't evaluate?
But when it comes to mapping data, we are often faced with their complexity and volume. We hit a wall: how can we present, in a way everyone can understand, something that is constantly evolving and that is seen from so many different points of view? How can we promote the understanding of this data?
Dictionary, glossary, catalog... what is all this?
It can be tempting to lump these terms together. We see it this way:
- The dictionary: a collection of words and expressions that define them with essential and non-essential attributes. For the record, the German word for dictionary, Wörterbuch, simply means "book (Buch) of words (Wort)". Wikipedia goes further by introducing the notion of reference, that is, a corpus aiming to teach and inform.
- The glossary: a collection of "glosses", that is, rare or foreign words associated with a definition. The aim is to describe or comment on terms semantically. Unlike lexicons, which describe all the words of a language, the glossary aims to describe words centered on a domain.
So we have two objects that, although they share a common objective (explaining words), are not built in the same way:
- The dictionary aims to be exhaustive, referential and precise.
- The glossary aims at popularization.
What does it mean for data mapping?
If we adapt these terms to data management, we can build the following definitions.
What is a data dictionary?
"inventory of the intangible asset of data, which allows users to discover and explore all available data sets, improve their understanding of that data, facilitate collaboration with other users to enrich the quality of those assets, and create more value from that data." (decideo)
This definition is close to the tradition of data management, so it leans toward IT. The idea is to:
- Carry out a data inventory that is as complete and detailed as possible: for each piece of data, a brief description, technical attributes (size, format, type...), and storage (which dataset or database does it depend on?)
- Be able to manage this data: what is its quality? Does it need to be protected? Is it sensitive? Is it subject to particular constraints?
- Allow exploration of this dictionary and collaboration around it: having a repository that gives control over data knowledge is good, but letting everyone take ownership of it is better.
The notion of value is deliberately set aside, as we would rather put it in the glossary.
In practice, a data dictionary is often a list of the data present in the organization, with:
- Its classification or hierarchy. Take an analogy with a dictionary entry, as Wikipedia does: "the dog is an animal of the class of mammals, order of carnivores and family of canids". By analogy, a classification for data management says that the data depends on a more or less complex chain of management: table, container, database...
- A definition: what the data means and how to describe it.
- Attributes: the information needed to understand its management context.
What is a glossary?
Defined above as a tool for popularizing data, the glossary is a synthesis of the organization's various terms, defined semantically, that is, in a way that's understandable from a functional point of view. It isn't intended to translate all the data contained in the dictionary. Its main purpose is to render and describe the main terms used in the organization.
In practical terms, the glossary aims to:
- Define the terms used in the organization in a business way: what does this indicator mean, and how is it calculated?
- Offer a certain modeling of the data in order to place it in a usage context everyone can understand. Data often depends on a context, and several points of view can be used to describe it: by data silos or verticals? by organization?
- Provide management attributes: operating domains, management constraints, intrinsic quality of the information.
How do the dictionary and the glossary work together?
When we talk about defining data, we have two complementary points of view:
- The dictionary defines the company's data from a technical point of view and in an exhaustive way.
- The glossary offers a functional "popularization" of the data.
Each deals with part of the need. What is the point of knowing the exact definition of an employee if I don't know the different elements managed in my information system? Take this "employee" object:
- In my glossary, I learn that the definition of an employee is "a physical person who has an employment contract with the organization with an end date in the future". Several attributes can describe this object: management responsibility, whether the data is personal, and so on. It can also be linked to other types of information: address, banking information, etc.
- In my dictionary, I learn that my employee is in fact a notion potentially scattered across several tables or even several systems.
Linking the elements of the dictionary with those of the glossary adds greater value to the data mapping process.
Beyond the small semantic discussions above, data mapping is modernizing. It follows the underlying trend in data management: opening up to business communities for better ownership. The challenge now is to know how to model the glossary so that it is understandable and valuable to everyone.