COMPEL Specialization — AITE-VDT: AI Value & Analytics Expert Template 3 of 5

This template builds a three-level KPI tree for an AI feature following Article 12’s methodology and rooted in Kaplan and Norton’s Balanced Scorecard (Article 13). It is designed for Miro/Lucidchart (visual tree) plus Excel or Sheets (wiring sheet) but translates cleanly to any modeling tool.

KPI Tree: [Feature Name]

Feature: [Short name] Owner: [Role, name] Version: [1.0] Date: [Date] Paired Balanced Scorecard perspective(s): [Financial / Customer / Internal-process / Learning-and-growth; a feature can map to multiple]

1. Level 1 — Business outcome

The top of the tree is a single business outcome stated in the same terms leadership uses. One sentence.

Outcome: [The business outcome this feature contributes to]

Quantified target: [Specific numeric target and timeframe]

Source of target: [Business case, strategic plan, OKR, etc.]

2. Level 2 — Drivers

Level 2 breaks the outcome into drivers — the intermediate variables that cause movement in the outcome. Typical trees have 3–5 drivers. More than 5 signals over-decomposition; fewer than 3 signals under-decomposition.

Driver 1 — [Name]

Description: [What this driver represents] Causality to outcome: [How does movement in this driver produce movement in the outcome? One or two sentences.] Baseline value: [Current value] Target value: [Target value] Owner: [Role accountable for this driver]

Driver 2 — [Name]

[…]

Driver 3 — [Name]

[…]

(Add additional drivers up to 5 if needed.)

3. Level 3 — Metrics

Each driver has 1–3 leaf metrics that measure it. Leaf metrics are what actually get computed from data. A driver without a leaf metric is a driver that cannot be measured; fix the driver or drop it.

Driver 1 metrics

Driver 2 metrics

Driver 3 metrics

[…]

4. Causal-validation check

For each driver, answer:

Driver 1

Q: Is the driver-to-outcome causality empirically supported? [Evidence: prior analyses, external benchmarks, academic references, pilot data]

Q: What would weaken the driver-to-outcome link? [Confounders, correlation-without-causation risks, etc.]

Q: What is the expected response time between driver movement and outcome movement? [Days, weeks, quarters]

Driver 2

[…]

Driver 3

[…]

5. Leading-lagging balance

Leading indicators predict outcomes; lagging indicators confirm. A good KPI tree balances both.

Leading-indicator count: [Number across all drivers] Lagging-indicator count: [Number across all drivers]

Balance check: Each driver should have at least one leading indicator. Drivers with only lagging indicators are drivers you cannot act on in real time.

6. Balanced Scorecard alignment

Gap analysis: Identify Balanced Scorecard perspectives with no driver or metric coverage. Drivers feeding every perspective is ideal; gaps are acceptable if deliberate but should be disclosed.

7. Wiring sheet (spreadsheet companion)

The wiring sheet is an Excel/Sheets table with one row per leaf metric. Columns:

The wiring sheet is the bridge between the conceptual tree and the operational dashboards. Every metric in the tree has a row; every row is traceable to a real data source.

8. Visualization

Recommended visualization

ConcentricRingsDiagram (from DiagramPrimitives): outcome at the center, drivers in the first ring, metrics in the outer ring. Color by leading/lagging type or by Balanced Scorecard perspective.

Alternative visualization

Matrix (tree): outcome at left, drivers vertically, metrics to the right of each driver. Easier to print; less visually striking but clearer for executive review.

9. Version-control

KPI trees evolve as features mature. Version history preserves the decision trail for auditors and for new team members.

Appendix A — Common failure modes and fixes

Appendix B — Linkage to other artifacts

• Measurement plan (Template 1): [Reference]
• VRR (Template 4): [Reference]
• Portfolio scorecard (Template 5): [Reference]
• Operational dashboards: [URLs]