> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/sdgm/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/sdgm/_mcp/server.

# How do I diagnose a batch prediction data pipeline?

> Check source freshness, graph compatibility, and saved prediction outputs

Check the pipeline in execution order:

1. Confirm that source refreshes completed and inspect row counts and timestamps.
2. Check that the graph snapshot contains the intended data, keys, relationships, and model features.
3. Inspect the population selected for scoring and its prediction times.
4. Check process errors, saved output counts, and scores joined back to entity keys.
5. Compare population and score distributions with previous batches.

Configure recurring checks and alerts in your scheduler or monitoring system. See [Making Predictions](/kumo-engine/user-code-guide/making-predictions) for population selection, scoring, and saved outputs.