Define the decisions the data should support
A training website wants to know which channels bring people who complete a course inquiry. Start with that decision, not a list of every button that can be clicked. Define the business outcome, its owner, and the action you would take if performance changes. Tracking a behavior without any plausible decision often creates maintenance work rather than insight.
Sketch the journey: course view, syllabus download, inquiry started, and inquiry accepted. Separate diagnostic interactions from outcomes. A download may show interest but should not automatically carry the same importance as a successfully submitted inquiry. Write the definitions in plain language so marketing, development, and admissions agree on what each measurement means.
Build an event dictionary
For each event, record its name, exact trigger, parameters, data types, owner, and validation method. Prefer a recommended GA4 event when its meaning matches your behavior. Google lists generate_lead for a submitted form or information request and documents prescribed parameters for recommended events. [1](https://support.google.com/analytics/answer/9267735)
Trigger a lead event on confirmed success, not merely on a submit-button click. Failed validation and server errors should not count as completed inquiries. Use a testable success signal from the application where possible. Define how single-page navigation, refreshes, and multiple forms behave so two developers do not implement incompatible interpretations of the same event name.
Keep private information out of collection
Do not send names, email addresses, phone numbers, or free-text inquiry messages as analytics parameters. Review URLs and page titles too: a confirmation URL containing an email address can leak data even when your custom event looks clean. Google's policy guidance explicitly warns against sending personally identifiable information to Analytics. [2](https://support.google.com/analytics/answer/6366371?hl=en)
Use controlled categories such as course_type rather than arbitrary user-entered text. Review consent and privacy requirements for your audience and implementation. GA4 data redaction can help remove likely email addresses and selected query parameters, but Google describes it as an additional control rather than a transfer of responsibility away from the collector. [3](https://support.google.com/analytics/answer/13544947?hl=en)
Test collection before interpreting reports
Use a development or controlled testing setup and inspect events in debugging tools. Submit a valid inquiry, submit an invalid inquiry, refresh the success page, navigate backward, and double-click the button. Verify event names, parameter values, and the number of events emitted. A polished dashboard cannot repair duplicate or missing collection.
Check whether both a direct tag and a tag-manager configuration send the same event. Inspect enhanced measurement settings where they overlap your custom implementation. Keep screenshots or redacted payload examples with the event dictionary. If a required custom parameter needs registration for your intended reports, configure that deliberately and document the reporting scope instead of assuming every collected field appears everywhere automatically.
Choose outcomes and attribution definitions carefully
Mark genuine business outcomes as key events according to your measurement plan. Record the counting and attribution settings used by the reports you share. A session-based channel view and an advertising attribution report can answer different questions, so conflicting-looking totals do not always indicate a broken implementation.
Connect lead quality through an appropriate, privacy-reviewed process rather than assuming every form submission has equal value. Keep CRM accepted leads distinct from analytics submissions and explain reconciliation gaps caused by consent choices, blockers, timing, or definitions. Do not force the systems to agree by quietly changing numbers. Differences are often valuable clues about the limits of the measurement.
Maintain a measurement release checklist
Before a website release, rerun the event tests for changed forms and navigation. Maintain an owner for every important event, and remove abandoned events from the reporting plan. Keep a dated change log for tag changes, consent configuration, filters, and outcome definitions so analysts can interpret trend breaks.
Your first report can be modest: course views, confirmed inquiries, inquiry rate under a stated denominator, and CRM-qualified outcomes by a clearly defined acquisition view. Add a caveat about observed rather than complete behavior. Reliable GA4 work is the discipline of maintaining definitions and validating collection, not the ability to fill a screen with charts.
Sources & further reading
Last editorial review: 27 October 2026. Examples are educational, not live employer data or guaranteed outcomes.