Analytical Explainability
From KPI to Story: A Repeatable Explanation Template
A consistent template makes AI explanations easier to generate and trust.
Rendering diagram...
KPI → change → drivers → segments → caveats.
TL;DR
- • Structure turns numbers into narratives.
- • Templates reduce ambiguity.
The problem (layman)
- • Explanations vary in quality and format.
- • AI answers lack consistency.
Why it matters
- • Templates help users interpret answers quickly.
- • They enable evaluation and comparison.
Symptoms
- • AI explanations omit drivers or time context.
- • Different KPIs have different explanation styles.
Root causes
- • No standard narrative format.
- • Measures lack supporting metadata.
What good looks like
- • KPI → change → drivers → segments → caveats.
- • Consistent ordering and language.
How to fix (steps)
- • Define a standard explanation template.
- • Update AI prompts or outputs to follow it.
- • Add required metadata fields.
Pitfalls
- • Templates that are too rigid for complex cases.
- • Ignoring user feedback on clarity.
Checklist
- • Template defined and documented.
- • AI outputs follow the template.
- • Template iterated based on feedback.