Business Intelligence

Setup

On first use, read setup.md for integration behavior and memory initialization.

When to Use

Use this skill when the user needs to build or improve business intelligence systems: KPI definitions, metric architecture, dashboard planning, executive reporting, and decision review loops.

This skill is optimized for operators, founders, product leaders, finance leaders, and analysts who need clear answers to "what changed, why it changed, and what to do next".

Architecture

Working memory lives in ~/business-intelligence/. See memory-template.md for base structure and status behavior.

~/business-intelligence/
├── memory.md                # HOT: goals, KPI ownership, active decisions
├── metric-tree/             # WARM: objective -> driver -> metric maps
├── kpi-contracts/           # WARM: metric definitions and formula versions
├── dashboard-specs/         # WARM: visualization and drill-down specifications
├── insight-briefs/          # WARM: weekly and monthly decision briefs
├── operating-cadence/       # WARM: review rituals and escalation rules
└── archive/                 # COLD: retired KPIs and past planning cycles

Quick Reference

Load only the file needed for the current task to keep context focused.

Core Rules

1. Start from Decision Questions, Not Charts

Every BI request must begin with one decision question and one owner.

If there is no decision owner, the output is reporting noise and should be reframed before building metrics.

2. Build a Metric Tree Before Dashboard Design

Map each business objective to drivers, then drivers to measurable KPIs.

Do not build dashboards first. Dashboards without a metric tree create disconnected charts and contradictory narratives.

3. Enforce KPI Contracts

Each KPI needs a written contract: definition, formula, grain, source, refresh cadence, owner, and valid interpretation window.

Never compare KPI values across periods if formula version or source logic changed without annotation.

4. Separate Leading and Lagging Indicators

For every lagging KPI, define at least one leading indicator that signals future movement.

If the system only tracks lagging outcomes, intervention happens too late.

5. Brief Insights in Decision Format

Every insight output must include:

• What changed
• Why it changed
• Confidence level
• Recommended action
• Action owner and due date

A BI summary without an action owner is incomplete.