HiLo.FM
Forecast Modification — Functional Specification

Version 0.1 (draft)2026-08-03Status: for review

Specification for a standalone, high-performance forecast modification tool (also known as demand assumptions, or risks & opportunities). The system accepts modification entries at any level of aggregation (HI space) and disaggregates them — at light speed — to the base planning grain (LO space: Product × Location × Week). The document is technology-agnostic: it specifies behaviour, data, interfaces and performance obligations, not implementation.

—Chapters

• 01Overview & ScopePremise, non-goals, demonstrator character, reset
• 02Concepts & GlossaryHI/LO, entry, grain, basis, split table — and the shared example dataset
• 03Data ModelMaster data trifecta, planning calendar, forecast, split tables, HI/LO stores
• 04HI EntriesScope, value semantics, overlap policy, state machine
• 05Disaggregation RulesForecast-proportional basis, NPI phase-in/out, sourcing split tables
• 06Disaggregation FlowTime → Product → Location pipeline, rounding, batch, performance
• 07User InterfaceScreens, interaction model, responsiveness obligations
• 08Webservices (API)Resources, operations, error model — full parity with UI
• 09MCP InterfaceTool surface for AI agents, mapped onto the API
• 10Demonstrator & Seed DataSeed profiles, reset procedure, demo scenarios, authentication

—Decision log & open points

Design decisions taken so far, and points explicitly deferred, are tracked on a separate page so the chapters stay stable while discussion continues.

→ Decisions & Open Questions

—Reading conventions

• Requirements are numbered per chapter: OV-01, DM-03, DR-07, … and are individually referenceable.
• MUST / MUST NOT / SHOULD / MAY carry their customary (RFC-2119-style) meaning.
• Amber-bordered requirements are provisional — agreed direction, parameters to be fine-tuned.
• All chapters reuse one shared example dataset so worked examples are cross-checkable.