What are the Main Characteristics of a Data Product?
I have delved into the background of how a data product has been introduced in a data mesh infrastructure and the unique benefits and value it can offer to an enterprise — decentralised data ownership with dedicated domain teams, leading improved data quality, agility and scalability. Read the previous medium article — 'Why does a Data Mesh matter?'.
The centralised platform and data domain teams should promote the key attributes of a data product to increase efficacy of producing the intended result to business domains, as well as to successfully manage it.
There are 6 key attributes that a business needs to consider when new data product is introduced or their existing data products are reviewed.
1. Discoverable 🔍
Data products should be easy to navigate and explore by keeping centralised data catalog or registry that publishes their ownership, metadata, operations agreement and other policies. The catalog or registry serves as the "one-stop-shop" entry point for those who are keen on exploring and consuming datasets from a data product, while enabling the structural engagement to start with a data product.
2. Addressable 🧑🏽💻
Each data product is located at a permanent and unique address, which guarantees continuous use irrespective of its evolution over time and in accordance with access policies.
3. Trustworthy 🤝
Data product owners must take ownership of managing data quality and adhere to an approved service level objectives through a data contract which explains how data quality dimensions can be managed.
The APRA is a Prudential Regulation Authority in Australia, and has published the best practice to manage data risk through the guide 'CPG 235' (similar to BCBS 239). Through this guide, the dimensions typically considered in the assessment of data quality include:
- Accuracy - the degree to which data is error free and aligns with what it represents
- Completeness - the extent to which data is not missing and is of sufficient breadth and depth for the intended purpose
- Consistency - the degree to which related data is in alignment with respect to dimensions such as definition, value, range, type and format, as applicable
- Timeliness - the degree to which data is up-to-date
- Availability - accessibility and usability of data when required
- Fitness for use - the degree to which data is relevant, appropriate for the intended purpose and meets business specifications
4. Understandable 🗣️
It is important to incorporate well-defined schemas with well-described semantics and syntax. This enables data consumers to independently access and utilise the data without requiring additional support.
️5. Interoperable (Standardised) ✍
The metadata (e.g. headers) and data types of enterprise-wide information can be standardised. The business entities such as application and employee details can be centrally managed by the platform team as master data. This helps data consumers easily combine one domain of a data product with others.
6. Secure 🔐
Each data product should be inherently secure with access control that only allows the authenticated data consumers to access datasets. The measure to protect the personal identifiable information (PII) must be in place. The access control is centrally managed with the platform team, and they configure approval process to provision access in consideration with ownership and stewardship of a given data product.
Now here is the question — Does your team have a data product meeting all these 6 key attributes? If it only fulfils several, what are the attributes that your data product misses out and needs to reinforce with?
Please share your stories via comments and follow my medium channel. Thanks for reading! The next article will cover the detailed values, that a data product can bring in to an enterprise, with examples.
Here are other articles you may be interested in: