DEMO / 06 · Demo ready
Demand and inventory lab
Demand forecasting baselines with rolling-origin backtests and a synthetic replenishment simulator for service level and holding cost trade-offs.
What decision is supported
How much stock to reorder and when, given a demand forecast with known error, so that a chosen service level is met without carrying more inventory than necessary.
Who uses it
A planner at a fictional distributor choosing a reorder policy, and an analyst checking whether a forecast method is better than the naive baseline on a proper backtest.
What data enters
Authored synthetic weekly demand histories for a set of fictional products, with seasonality and promotions written into the generator and documented.
What is computed
Rolling-origin backtests of several baseline forecasters with error metrics per horizon, then a replenishment simulation that turns forecasts into orders under a policy and records stockouts, fill rate and holding cost across a grid of policy settings.
What action is suggested
A policy setting per product with its simulated service level and holding cost, and a plain statement of which forecaster beat the naive baseline on the backtest and which did not.
What evidence supports it
Backtest errors and simulated fill rates are computed on the synthetic fixture and labelled demo. They describe the generator, not a real supply chain.
No metric is claimed for this project. Anything computed by the repository on its synthetic fixture carries the label demo.
What fails or is uncertain
- Lead times are fixed in the fixture; variability in supply is not modelled.
- Costs are demo units chosen to show the trade-off shape, not calibrated to any business.
- Forecasters are baselines by design; the point of the lab is the evaluation discipline, not the model.