← Every exhibit

Table Table 3 From the research bench

The Metric Hierarchy in Controlled Experiments

Metric ClassOperational PurposePrimary ExamplesDecision Rule in Experimentation
1. Overall Evaluation Criterion (OEC)The unified strategic objective function that defines organizational successComposite score balancing 30-day net customer retention and annualized margin contributionPrimary criterion for declaring experimental victory and rolling out code to production
2. Direct Driver MetricsLocal transactional metrics measuring immediate behavioral responseCall-to-action click rate, checkout form start rate, onboarding completion rateDiagnostic indicators to understand why the treatment impacted the primary OEC
3. Guardrail / Health MetricsInvariant system and business health indicators that must not be degradedServer latency (p99), checkout error rate, customer support ticket volume, refund rateNon-negotiable veto power: if any guardrail metric breaches thresholds, the test is aborted
4. Long-Term North Star MetricsMacro-economic business metrics tracking durable shareholder valueCustomer Lifetime Value (LTV), Net Revenue Retention (NRR), brand equity indicesEvaluated retrospectively across long-term holdout cohorts to verify sustained impact

Swipe or scroll horizontally if the table is wider than your screen.

Cite Embed

Reference & Evidence

Source: Table from this essay. Sources and interpretation are given in the article.