Data is everywhere around us. It is like we are surrounded by data. Even with so much available data, we are unable to utilize it to the fullest. This may be due to poor understanding of the ginormous data available or poor documentation or not being able to find the data that is important to us.
It was then concluded that it was important to integrate data across enterprises, store and analyze it. It was understood that historical data was of great importance for the present understanding of data. Summarization of data still holds immense value to organizations.
Data warehouses and Data marts were introduced as a result. They are used to take data from the data creation systems and after applying logic to it, extract the data with the required tools and store the data.
The data is stored in structures which can support our process of analysis. They are concepts that broadly define the creation of sets or tables for reporting and analysis. So necessarily we are collating data, moving it and storing it in a form that is meant to give reports on it as an output.
This way we can look at data across multiple systems and also reduce our burden of analysis of data generated by data generation systems. They do not focus on daily operations and transactions but focus on modelling and analysis of data.
For precise and explicit business intelligence, companies count on data warehouses. Data marts too serve a similar purpose so what is it that makes them different from each other?
Data warehouse vs Data mart
The table below provides difference between Data warehouse and Data mart.
| Data warehouse | Data mart |
| Centralised system of storage (covers data on various subject) | Centralised system of storage (covers data on various subject) Decentralised system of storage (focused on particular user groups) |
| Dozens or hundreds of data sources | Typically just a few data sources |
| Stores data on various subjects | Stores data related to a particular topic |
| Detailed form of data | Summarized form of data |
| The data is slightly denormalized. | The data is highly denormalized. |
| Snowflake, Fact constellation schema is used | Star and snowflake schema is used |
| Built on top of databases and other data-generating systems | Built generally on data warehouses |
| Top-down model | Bottom-up model |
| Objective is useful for data visualization, analysis, business intelligence | Objective is store data useful for a particular part of subjects |
| Flexible, data-oriented and long life. | Restrictive, project-oriented and short life. |
| Harder to build due to large amounts of data with a high risk of failure | Simpler to build due to lesser amounts of data with less risk of failure. |
| The Data Warehouse is vast in size. | a data mart is smaller in size than a data warehouse. |
| Business-wide analysis | Department-specific analysis |
What is data warehouse?
A data warehouse was described as a subject-oriented, integrated, time-variant and non-volatile collection of knowledge in support of management’s deciding process by W.H. Inmon. It is maintained separately for the organisation’s operational database. It’s meant to provide a platform for information processing and historical analysis of data. It provides a simple concise view around a particular subject by excluding data that is not useful for the process.
Data warehousing was the process of creating and using data warehouses. They are constructed by integrating multiple, heterogeneous data sources. Data here doesn’t need an operational update. Only two operations of loading and accessing data are required, data is loaded in static format and doesn’t need any modification.
Data warehouses don’t contain the most recent information. It serves the purpose of correlation between data from different source systems. People often confuse themselves between a data warehouse and a database.
A data warehouse is a top layer on databases, which procures (Extracts, transforms and loads) data from them and stores them for analysis. This data is then processed for its insights.
A data warehouse is used for online analytical processing (OLAP) which serves knowledge workers in the role of decision making and data analysis. It is used to extract important insights and streamline business processes and is an important element in business intelligence. It enables business users to access query-relevant data faster and also improves data consistency and quality.
Data warehousing has applications in various fields of Finance, Telecommunications, Transport and many more.
What is data mart?
The data mart is an independent, logical subset of Data warehouse. Data focuses on specific users and is focused on a particular functional area of people. It is a small repository of data. Data mart may have a similar structure as Data warehouse. However, it takes way less time to set up a data mart i.e. a few months. This could be due to its smaller size and data extraction from a lesser number of sources.
A data mart is preferred to be used at a dedicated business unit. Thus data mart is a data warehouse with a limited scope and whose data can be analyzed by summarization. Therefore data for the entire organisation is not needed. Several data marts can be used by a single business. Designing a data mart is a lengthy and costly process. Even so, every sector of business must have a data mart of its own.
Author
Shriya Upasani
MIT World Peace University
