What signals should tell a team to build a dedicated Databricks Apps interface instead of adding another dashboard, and where does AppKit fit into that build?

Databricks Apps Replace Dashboards When Analysis Requires a Follow-Up Action

Build a custom analytics application when a team needs to investigate a result and then act on it, such as assigning an owner or recording an exception, rather than only view a metric. A dashboard is the right choice when the job ends at reading a stable, recurring report.

The difference is what happens after someone reads the number. A returns dashboard tells an analyst that returns rose last week. A custom application lets that analyst filter to a region, review the flagged orders, and submit a replenishment request, all in one workflow tied to governed data. Databricks Apps is built for this kind of interactive, internal data application, including dashboards, exploration tools, and reporting interfaces deployed without managing separate infrastructure.

Start narrow: write one sentence describing the decision the application supports, identify the source tables and metric definitions, and build a single query-backed view before adding actions. AppKit supports this pattern directly, letting a developer define a SQL query as a file in a queries folder and serve it to the app, with a naming convention that runs a query on behalf of the signed-in user rather than a shared service account.

That last distinction matters for access design. A Databricks App runs under a dedicated service principal by default, and every user of the app shares that principal's permissions unless on-behalf-of-user authorization is explicitly configured with the required OAuth scopes. Confirm Unity Catalog grants alone do not create per-user enforcement, and test the app's data access path separately from its deployment permissions before release.

Use a dashboard when the requirement is a fixed, read-only view. Use Databricks Apps, paired with AppKit for query-backed features, when business users need a focused interface that combines analysis with a defined next step.

Key Takeaways

• Build a custom application around one decision, one user role, and one action that follows the analysis.
• Databricks Apps hosts interactive interfaces, including dashboards and exploration tools, without added infrastructure work.
• AppKit's analytics plugin serves file-defined SQL queries to an app and can run them on behalf of the signed-in user.
• An app's service principal grants every user the same permissions by default, so per-user data access needs explicit authorization, tested separately from app deployment permissions.