Engineering, production, governance, cost, communication · insight-1289
Keep a model inventory that links each model to its data, owner and consumers
- Evidence type
- practitioner heuristic
- Difficulty
- medium
- Roles
- ML engineer, Data scientist, Applied AI
- Industries
- banking, insurance, cross-industry
Summary: situation, action, reason, boundary
Situation
A source table is being deprecated and nobody can list which models, spreadsheets and vendor scores read it (hypothetical).
Action
Maintain one inventory recording for each model its purpose, owner, risk tier, input datasets and features, downstream consumers, version and review date. Include vendor and spreadsheet models, make registration a deploy gate, and review the list each quarter.
Reason
Supervisory guidance treats a comprehensive inventory, with enough information to understand risk for each model and in aggregate, as common practice.
Boundary
Organisations with more than a handful of models, and any regulated model estate.
Symptom or risk
Impact analysis becomes a search by email, dependants break silently, and incident, licence or privacy questions cannot be answered across the portfolio.
Exceptions
A few models fit in a wiki table; dedicated tooling pays off with dozens.
Trade-offs
An inventory rots unless registration is enforced at deployment.
Pitfalls
Listing models but not the datasets they read, so impact analysis still fails.
Lessons and concepts
Claims and their sources
SR 26-2 states it is common industry practice for banks to maintain a comprehensive set of information on models under development or in use, enough to understand model risk individually and in aggregate.