instrumentation-planlisted

Turns a feature or flow into an instrumentation plan - the questions to answer, the success and guardrail metrics, an event taxonomy with a consistent naming convention, event properties, identity and consent rules, funnel and retention definitions, and a QA checklist - so product analytics can evaluate the design after launch. Use when specifying analytics for a new feature, reviewing an existing tracking plan, defining activation or retention metrics, or preparing an A/B test. Triggers on instrumentation, tracking plan, event taxonomy, analytics events, event naming, product analytics, success metrics, guardrail metrics, funnel analysis, activation metric, retention metric, north star metric, A/B test metrics, experiment design, measurement plan, data minimization for analytics.
stritar/skills · ★ 0 · Testing & QA · score 63

Install: claude install-skill stritar/skills

# Instrumentation plan Produces the plan that lets a team learn whether a design worked. Metrics first, events second: an event that answers no question is cost without information. The plan is a deliverable engineers implement and analysts query; it is not the analysis itself. Post-launch evaluation of results belongs to `silver-measure`; experiment analysis to the team's analytics tooling. ## Inputs - The feature or flow: screens, states, entry and exit points. - The decision the data must inform (ship, iterate, roll back, expand). - Existing conventions: current event names, analytics vendor, identity model, consent framework. Read them before inventing new ones. ## Workflow ### 1. State the questions and the metrics Write two to five questions in the form "did users X after Y". Derive: - **Success metrics** — the outcome the design intends (task completion rate, time to first value, adoption of the new path). - **Guardrail metrics** — what must not get worse (error rate, support contacts, churn of the old path, latency, accessibility-related drop-off). - **Activation and retention definitions** where relevant: the action that marks a user as activated, and the window and action that count as retained. Make each definition computable from the events below. Every metric gets an owner, a baseline (or "none, first measurement"), and a target or a stated hypothesis. Read [references/metric-patterns.md](references/metric-patterns.md) for the standard funnel, a