# Next-day bakery production planning

Synthetic business case

Train and evaluate the portable random forest on the instance’s synthetic sales history, inspect store-SKU forecasts and review production/distribution actions.

## Business question

How much should each store bake tomorrow, and can the forecast evidence and production approval be reviewed together?

## Who it is for

Bakery-chain operations, store planning and central production teams

Perishable products can sell out when production is too low and generate waste when it is too high. Planners need to know which history supports a forecast and how it becomes a production task.

The portable policy trains on instance history using sales windows, calendar and store-manager features, preserving temporal holdout evaluation, predictions and evidence records.

Planners review evaluation results and business rules before approval. Approval, dispatch and closure are recorded separately; synthetic-data metrics do not establish customer production accuracy.

## Business objects

- Stores and SKUs
- Sales history
- Future features
- Model evaluation
- Production and delivery tasks

## Try the workflow

1. Inspect the business records — Select STORE-MALL-018, SKU-CROISSANT-006 and the guide’s start date, then inspect sales history and future features.
2. Analyze and check the evidence — Run training and forecasting on current history, checking training-row counts, temporal holdout MAE/MAPE and source rules; these metrics describe the synthetic example only.
3. Change an input and recalculate — Edit the forecast-feature record specified in the guide, recalculate and compare current forecasts and production proposals rather than using old prefilled results.
4. Review actions and their records — After reviewing forecasts and production rules, approve, dispatch and close the current task as guided, then inspect model snapshots and receipts.

## Adapt to your business

- Prepare valid store-SKU sales history; the current policy requires at least 60 historical dates.
- Confirm store-manager inputs, pack sizes, production buffers and early-delivery proportions.
- Evaluate errors and business outcomes on your own data; allow time for importing the larger package.

## Case package

Version: 0.1.4 · 530,012 rows · 64 resources · 14731216 bytes

Package language: Chinese

SHA256: 9fd83f8f9382a29b5f925dfb2a629d2a25937fc8f3f9e46b2489b018c644c1f3

Import checks the required capabilities and your account permissions.

[ZIP](/case-packages/retail-sales-forecast-demo/0.1.4/case.zip)

[Import guide](/en/cases/import)

Required capabilities: domain-case-workflow v1

This synthetic case is not a customer success story.