KPI Tree Builder: design and audit a metric system

KPI Alignment traces Goal → Strategy → Tactic → KPI. Those links explain why a measure is attached to a strategy; they are not formulas. Metric Decomposition starts from a selected outcome and lets you mix mathematical identities with clearly labeled influence hypotheses, diagnostics and guardrails. Arithmetic scenarios run only across complete formula paths.

Use Formula only when the parent is defined by the children, for example GMV = Orders × AOV or Conversion = Purchasers / Eligible visitors. A correlation such as faster onboarding being associated with retention is not an identity. Record it as an Influence hypothesis and add the evidence or rationale you actually have.

No. It is an arithmetic scenario. Given the formulas and values you supplied, it recalculates the selected outcome while holding other declared inputs constant. It does not estimate whether a feature will cause an input to move, whether the change is feasible, or whether that input should be prioritized.

Yes. Import the metrics CSV and the relationships CSV separately. The importer detects the file type from its headers, validates referenced metric IDs and rejects relationships that would introduce a cycle. Imported legacy-style relationships are not silently treated as causal or mathematical facts; you still control their declared semantics.

The audit checks graph integrity, metric-definition gaps, relationship/formula semantics, decision usefulness and measurement readiness. It can flag cycles, unresolved roots, missing definitions or units, weak rate definitions, ratio ambiguity, incompatible additive units, missing data sources and other concrete unknowns. It deliberately does not collapse those findings into an arbitrary quality score.

No. The share link contains a compressed copy of your tree data in the URL. Compression is not encryption. Anyone who receives the link can read the tree, so do not put sensitive internal metrics in a share URL unless that is acceptable. Large trees should be shared using an exported JSON or Markdown artifact instead.

No. The root can be any clearly defined Product or business outcome that you need to decompose. If your real question is which candidate North Star should anchor the system, use the North Star Metric Finder first. This tool owns decomposition and audit, not North Star selection methodology.

A KPI tree describes how metrics relate: which links are mathematical identities and which are hypothesized influences. OKRs describe commitments: an objective, the key results you will be accountable for, and the review cadence. A key result may reference a tree metric, but the tree never turns every node into a commitment.

No. Only formula identities are treated as mathematical facts, and only for the definitions, units, denominators and time windows you declared. Every influence edge stays a labeled hypothesis with its own rationale, evidence note and next test, and arithmetic scenarios never calculate through it.

Stop when the next decomposition no longer changes a decision. A useful rule: decompose until each leaf is either actionable enough to inform an experiment or decision, or diagnostic enough to explain movement. Extra levels without new decision value only make the tree harder to review.

A guardrail is a metric you monitor to avoid winning the wrong way, such as refund rate next to GMV or support contacts next to activation. Guardrails are tagged explicitly and kept out of the arithmetic identity unless they are mathematically part of it.