artifactData, experimentation & growth

Instrumentation Plan

An instrumentation plan specifies the data a product must emit to answer priority questions, including event logic, properties, ownership, and validation.

Download Markdown (.md)

01What it means

An instrumentation plan, often called a tracking plan, is an implementation specification for analytics data. It states which events and properties to collect, exactly when they fire, where values originate, how users or accounts are identified, and how quality, privacy, testing, and release will be managed.[1][2][3]

Decision need to governed analytics data Specify meaning and trigger before implementation emits an event
  1. 01Decision + questionWhy the data is needed
  2. 02Event contractName, meaning, trigger, actor, object
  3. 03Properties + identityAllowed values and source systems
  4. 04Governance + testPrivacy, consent, payload, edge cases
  5. 05Release + monitorQuality, documentation, version lifecycle

A tracking plan includes identity, values, ownership, privacy, testing, monitoring, versioning, and deprecation—not names alone.

02Why it matters

Analytics questions cannot be answered reliably when collection logic lives only in code or memory. A shared plan aligns product, design, engineering, data, privacy, and quality assurance before implementation. It also creates acceptance criteria for verifying that observed events represent real product behavior.

03When to use it

Create or update an instrumentation plan for a new product, feature, journey, experiment, or analytics migration. Begin with decisions and questions rather than tracking every interface action. Revisit the plan when flows, identity models, consent requirements, or metric definitions change.

04How it works

  1. List the decisions, questions, journeys, and measures the data must support.
  2. Specify each event’s name, meaning, trigger, actor, object, properties, allowed values, and identity requirements.
  3. Record platform coverage, source system, owner, privacy classification, consent behavior, and retention constraints.
  4. Implement in a test environment and verify payloads, duplicates, ordering, missing values, and edge cases against acceptance criteria.
  5. Release with monitoring, documentation, versioning, and a process for correcting or deprecating data.

05A practical example

Example

For a document-sharing flow, the plan defines “Share Link Created” only after the server confirms creation. Properties include document type, permission level, and entry point, but exclude document names. QA tests web and mobile, offline retries, duplicate submissions, and anonymous-to-signed-in identity transitions.

06Where it breaks

A plan adds coordination cost and can become stale if treated as a one-time spreadsheet. Passing payload checks does not prove the selected events answer useful questions. Excessive collection increases privacy, security, storage, and governance burdens.

Common mistakes

  • Instrumenting every click without a question or decision it supports.
  • Describing an event name but not its exact firing condition and source.
  • Testing only the happy path or sending development traffic into production data.

07Sources

  1. What Events Do You Need?Amplitude Docs · Official planning guidanceWhat Events Do You Need?
  2. Getting Started with AmplitudeAmplitude Docs · Official instrumentation documentationGetting Started with Amplitude
  3. About EventsGoogle Analytics Help · Official analytics documentationAbout Events
View saved topics

Start typing to search titles, definitions, categories, and frameworks.