Overall Service level indicators are traditionally classified on 3 dimensions - availability, performance and correctness. While they mean different things when used in context of Data product (and their meaning changes based on type of data product) , all 3 hold true. However, there is one more characteristic that deems important to serve as an indicator - specifically for data. That's security.

Details

The same holds true for a Data product:

1. Availability
   1. Technically the uptime the user can expect.
   2. But also how accessible it is - support for different access patterns.
   3. Ability to mutate/fix and delete (forget like [Machine Learning](Machine Unlearning in 2024 - Ken Ziyu Liu - Stanford Computer Science))
2. Performance
   1. How fast can we get the desired insight out of the data product, or fix an observed anomaly.
3. Correctness
   1. What time window does the product capture, how old, how recent.
   2. How much of the data is present vs missing
4. Security
   1. What are the level of access controls in place

Examples

• Completeness (like freshness)
• Query-ability - SQL like, programmatic, drag drop, excel.
• Mutability (like forgetfulness )
• Bias-freeness

Related

• Data Product - Observability

Tags

#data-engineering