What we announced with Databricks at the 2023 Data + AI Summit
At the 2023 Data + AI Summit, Sigma and Databricks showed off a deeper integration between the two platforms. Some of what Databricks announced that week has since been renamed or restructured; the parts specifically about the Sigma integration have held up.
What Databricks announced
Databricks used the keynote to introduce Lakehouse Federation, letting a single Unity Catalog query reach across databases like MySQL, PostgreSQL, Redshift, Snowflake, and BigQuery without moving the data first. That feature has since reached general availability and is still how cross-platform queries work in Databricks today. Databricks Marketplace, a catalog for sharing datasets and notebooks across organizations, launched the same week and is still active.
Databricks also introduced something it called LakehouseIQ, a knowledge engine meant to answer questions about a company's data by reading signals from Unity Catalog and its pipelines. Databricks has renamed and restructured its AI product lineup more than once since 2023, so this is worth reading as a snapshot of a 2023 idea rather than a current product name.
What Sigma announced
Sigma announced Input Tables at the same event: a way for someone working in a workbook to write data directly back into Databricks, instead of only reading from it. That's still how Input Tables work today. A business analyst can log a forecast override, a correction, or a manual estimate straight into a table that Databricks and every other connected tool can then query.
Three ways teams used it
The write-back mattered differently depending on who was using it.
Business leaders
A decision maker could adjust an assumption inside a model and see the result immediately, instead of filing a request and waiting for a data team to rerun something.
Data teams
Whatever a business user entered became part of the same Lakehouse everything else already queried, so nobody had to reconcile a spreadsheet against the warehouse after the fact.
Data science and ML teams
A model could be corrected with a real, current input instead of waiting for a full retrain. This recorded talk from the summit covers that case in more detail.
See it on your own data
Request a demo if you're writing data into Databricks today through a spreadsheet or a script, and want to see what that looks like from inside a workbook instead.
