Resources

Category: Uncategorized

The stockout you can predict is not the one that hurts

Regular demand is easy to forecast and rarely causes stockouts. The damage comes from the events your model has three examples of.

Demand sensing is not forecasting with a shorter horizon

One extrapolates history. The other reads signals that history has not absorbed yet. Confusing them produces a fast forecast that is wrong in the same way as the slow one.

Assortment decisions are where retail AI earns its keep

Replenishment optimises what you already sell. Assortment decides what you sell. One of these has an order of magnitude more upside.

The first decision you automate should be reversible

Not the most valuable one. Not the most frequent. The one where a mistake costs a correction rather than a customer.

Why on-premise is a decision-architecture question, not an IT preference

Where the model runs determines what data it can see. What it can see determines what it can decide.

Predictive maintenance fails at the work order, not the model

The model predicts the bearing will fail in eleven days. Then nothing happens for nine of them.

The audit trail is the product

Organisations do not refuse to automate decisions because the model is inaccurate. They refuse because nobody can explain the decision afterwards.

Agentic AI in production planning: what actually gets delegated

Not the schedule. The hundred small adjustments to the schedule that a planner currently makes by hand, each one defensible and none of them written down.

Retail replenishment is a decision problem wearing a forecasting costume

Every replenishment system forecasts demand and then applies a rule. The forecast gets the investment. The rule decides the outcome.

Why your forecast is accurate and your decisions are still late

Forecast accuracy is measured against what happened. Decision quality is measured against what you could have done. The two are not the same number.

Request a Demo