How do I improve model performance?

Improve model performance by checking the task, data, graph, and evaluation setup before increasing the AutoML search or tuning model settings.

1. Confirm that the predictive query defines the intended entity, target, and time window.
2. Inspect column types, missing values, label quality, and possible leakage.
3. Verify that graph relationships use the correct keys and have the expected match rates.
4. Compare training, validation, and holdout results for overfitting or distribution shifts.
5. Use explainability to investigate weak cohorts and unexpectedly influential columns.
6. Add tables or columns incrementally and measure whether each change improves holdout performance.
7. Tune the run mode or model plan only after the data workflow is correct.

See Debug Model Performance for the complete diagnostic workflow.