You need to put a word in front of the term “Semantic Layer” and call it a “BI Semantic Layer”

#AgileData

I am going to be pedantic on your semantics again.

The 90’s have returned (or is it the 2000’s?) and Semantic layers are cool again.

It’s a hot topic on LinkedIn and Data Slack channels.

The need to pass context or understanding to the LLM to improve its Text to SQL response is driving the resurrection.

I was working with a side hustle customer and we were discussing the need for a BI Semantic layer. This layer would be between the final Consume data layer and the BI tool, in this case PowerBI.

We discussed if PowerBi was the only tool that would access the Consume layer or if there would be a BI Toolkit approach. This would drive the decision if we could just use the semantic technology in PowerBI or if we needed a separate technology capability that would power any tool.

We also discussed the desire to push left as much as possible, so what metric calculations could be pushed to left to the Design layer, what could be put in the Consumer layer, what should be in a separate BI Semantic layer (and did we even need one), and what would be in the BI Tool.

But here is the thing.

Isn’t your data catalog a Semantic layer?

Isn’t your knowledge graph a Semantic layer?

Isn’t your Ontologies and Taxonomies a Semantic Layer?

In #AgileData we hold all the business logic for the transformation rules in one place, we augment them with Context and we expose them so they can be reused. Isn’t that a Semantic layer?

We convert and hold natural language version of the rules as well as the SQL code, isn’t that a Semantic layer?

We are working on iterations of creating and maintaining Business Glossaries and Data Dictionaries, aren’t those Semantic layers?

So as I have mentioned before, we should talk about “Active Marketing Customer” vs “Active Risk Customer”, should we not then talk about “BI Semantic layer” rather than a generic Semantic term?