dbt has taken the world by storm, empowering analysts and data engineers alike to leverage versionability, testability, reusability, reproducibility, and a declarative approach to data transformation. And dbt has plans to cover more of the data translation layer with their recent announcement of the metrics layer at dbt Coalesce. (For more on the metrics layer, follow Benn Stancil’s fantastic substack.) If you’ve heard of the idea of headless BI, this will scratch that itch.
Whether you use dbt or a headless BI platform, this trend leads to downstream data applications that can be rapidly deployed and are much more nimble, lightweight, and actionable. That means less business logic embedded in data marts, cubes, or directly in the BI or analytics tool, and more time focused on using data to make better decisions.
See Sisu for an example of a company driving towards smart insights. See ThoughtSpot for a compelling take on fewer dashboards and more embedded answers. See Eppo for data experimentation. And see Lightdash for an example of simpler, dbt-friendly, code-defined dashboards.
4. ‘Mo code AND no code
We mentioned that dbt really demonstrated the value of an as-code approach to data transformation. Tools are diverging in two directions — more code AND no code. It’s really not a battle of india whatsapp number data which will win, but rather of which you prefer for your use case and your technical level.
No code tools provide a low barrier to entry and fast time to value. As-code tools provide versioning, testability, a declarative approach, and more. And the best no-code tools will sit on top of as-code frameworks to get the best of both worlds together.
For more on the role code plays and the right data user experience, check out:
Cindi Howson (ThoughtSpot) on the importance of low-code/no-code for analytics
Tejas Manowar (Hightouch) on low-code/no-code UX for reverse ETL integration
Erik Bernhardsson on data tools: the good, the bad, and the ugly
Nick Schrock (Elementl / Dagster) on modern data stack and orchestration
5. Data mesh rolls up its sleeves
A top trends list wouldn’t be complete without mention of data mesh. It was quite possibly the hottest topic of 2021 in data management, and perhaps also the most polarizing.
Data as a product? YES!
Self-service infrastructure? HECK YES!
Domain-driven ownership? YES I THINK? LET ME LOOK THAT UP.
Data product architecture quantum? HUH?
While there will continue to be detractors of data mesh with fair reasons to be pessimistic, with scrutiny comes pragmatism. We predict data mesh is going to quickly pass through the hype phase with a practical approach in hand. In 2022, we’ll see case studies on how companies are incrementally implementing the parts of data mesh that are fit for purpose, figuring out what a good data product really looks like, and codifying federated computational governance in an agile, bottom-up way.
What do you think?
Do you agree that the above will be among the data management trends for 2022? Join our honest, no BS conversations about enterprise data management with data leaders and practitioners by subscribing to our podcast Catalog and Cocktails.