Night's Watch turns raw multi-currency bank balances into a single ranked view of which portfolio companies are at risk, why, and what to do — with every number traceable back to the raw file behind it.
coverage = expected transferable cash at next payment (T1) ÷ cash to repay at T1At-risk companies ranked worst-first: understand why, pull loan follow-up, get a P1–P4 action, log it.
Open the desk OntologySource → Bronze → Standardisation → Silver → Golden. Click any input to trace its lineage and grain.
Explore the model End to endA new bank file → a High warning → a credit-team action → the updated database, step by step.
Play it through Now → 1m → 2mWhat each data horizon unlocks and how the See → Decide → Act workflow sharpens as more data lands.
See the roadmap Trust & lineageEvery metric with its formula, as-at date, raw source file(s) and the evidence behind it — the reference the desk hovers link into.
Open the dictionary WalkthroughA Teams alert fires, the credit specialist opens the desk, drills into the loan, decides, acts — end to end with subtitles and voice.
Play the demo Future · previewThe later-phase vision: an AI agent sweeps the portfolio overnight and presents a morning briefing — the analyst reviews and approves, not finds and analyses.
See the mockupSynthetic data only. One company mirrors the numeric distribution of a real bank export as the worked example. Ontology & schema: docs/ontology.md · schema.sql.