Does the model understand how your industry works?

Does the model understand how your industry works?

An analytical model can fit historical data and still misunderstand the business. Industry validity depends on whether it represents the economic relationships, constraints and behavioural responses that govern outcomes�not merely whether it predicts the recent sample.

Make the causal structure explicit. Map value flows, capacity, substitution, contracts, regulation, seasonality and upstream or downstream dependencies. OECD input�output tables illustrate why structure matters: inter-industry sale and purchase relationships expose how a shock propagates beyond the sector where it begins. A firm model needs the equivalent logic at its own decision scale.

Challenge proxies and boundaries with domain experts. A variable may correlate with demand while omitting rationing, channel conflict or a regulated ceiling; historical elasticity may fail after a structural change. Compare alternative theories, document assumptions and identify operating conditions in which the model should be restricted or overridden.

Validation must combine conceptual soundness, outcomes and ongoing monitoring. Revised US interagency model-risk guidance issued in April 2026 reinforces a risk-based approach tailored to model use and complexity. Backtesting alone is insufficient when the future regime, product or customer mix differs from the development data.

Use scenario, sensitivity and stress tests to reveal unstable interactions, then monitor whether relationships remain credible in production. Record expert overrides and realised outcomes. A model understands the industry to the extent that its structure survives informed challenge and continues to support decisions when market conditions move beyond the period it learned from.