Metric Store is an important part of the Modern Data Stack (MDS) today. Basically it enables to centralize your metrics and KPIs for a unified access and semantic.
As Metric Store is a typical MDS topic, it is considered as part of MDS 2.0.
This is part of my “The Modern Data Stack” blog series, focussing on the Metric Store part.
In simple terms a Metric Store can be seen as a layer between one or many data warehouses and BI tools. The idea behind is to de-couple the metrics from front- or backend to replace these components easier (supporting fail-fast mindset).
There are some names and terms around Metric Store:
Metric Layer - let us assume it is the same as Metric Store
Headless BI - let us assume it is the same as Metric Store
Semantic Layer - in general SL is a broad concept delivering meaning for metrics and characteristics of underlying data sources1. It can typically include a Metric Store 2
If you check different solutions you can see evolution as name and concept changes of the offerings in the market so that today these terms are nearly similar to use.
Metrics are defined in three different solutions3:
BI-Tool (Tableau, Power BI, QlikSense, …)
Data Warehouse (Snowflake, SAP BW, Redshift, …
Operational System (ERP, CRM, …)
Metric Stores should ideally deliver modeling capabilities, access management, caching and APIs for access through BI tools4.
Why is a Metric Store important for a company? A nice overview is if you think about metric and the logic behind5:
Same Metric / Same Logic
Same Metric / Different Logic
Different Metric / Same Logic
Different Metric / Similar Logic
From the timeline Metric Store came up since mainly 2021 as seen here with typical vendors. Some vendors may come from a semantic layer context with a longer history like Cube and dbt - so I just set them where they started with their Metric Store offering:
Popular vendors for Metric Store are Cube (2016), Transform (2019), Weld (2021), MetriQL (2021), Propel (2021) and dbt (2022) with it’s semantic layer