Loading SMB cash flow risk cockpit…
Fetching receivables, forecast, and customer risk signals.
Fetching receivables, forecast, and customer risk signals.
This project keeps the runtime scorer honest: the live product uses a rule-based baseline while benchmark artifacts and project-native readiness reports show what exists beyond the demo path.
rule-based · target is_late_15
0 rows · 0 positive labels
Reference artifacts for modeling behavior, not runtime replacement.
Rule-based baseline for SMB cash-flow risk MVP. Weights are heuristic, not learned from historical data.
Keep the current runtime scorer rule-based. External benchmarks are useful as benchmark evidence and feature-discovery inputs, but they do not yet justify replacing runtime scoring.
Keep the live application on the interpretable rules model until native data volume clears the minimum training thresholds.
Project-native learned scoring remains deferred until enough native invoice history exists.
Reach at least 200 native rows and 25 positive late-payment examples.
These runs help compare modeling behavior and surface feature ideas. They do not validate direct runtime transfer into this app.