Blog
Data engineering, in practice.
How to move data from any source to any destination — extraction, CDC, transformation and loading, told by the team that builds Januss.
All articles
20 articlesMedallion architecture: what it is, when to use it, when it is overkill
The medallion architecture became an industry standard, and it also became overkill on plenty of small projects. What each layer solves, what must not go into each one, and when two are enough.
Airbyte or Januss: the connector is similar, the rest is not
Airbyte and Januss handle the off-catalogue source the same way. The difference shows up in what happens after the data lands — and in who operates the platform that made it land.
The worst time to find out the data is wrong is in the meeting
The pipeline ran, the dashboard refreshed and nobody was alerted — because nothing broke. Whoever spots the error is usually whoever uses the number, and that says less about the team than about where the test runs.
Pentaho or Januss: when it is worth keeping what already runs
Switching integration tools is rarely about features. It is about who runs the server, who maintains the routines and what happens when the person who built it all leaves.
dbt, Python or Januss: what each one solves on the data path
Comparing dbt, Python and Januss as if they were alternatives leads to the wrong choice. They cover different stretches of the data path — and what is left for your team to build changes with each design.
Python in the data stack: which part really has to be Python
Not all Python in a data stack is the same thing. One layer needs the language, another is habit, and the third is maintenance nobody hired anyone for. Tell the three apart before deciding.