Website analytics should answer a defined question about customer behaviour or business outcomes. Reliable measurement needs clear event meanings, consistent campaign tags and evidence that the implementation records what the report claims.
Define each event before adding a tag
Write down the action, triggering condition, expected parameters and business meaning. Distinguish an interaction from a completed outcome. A form submission interaction may happen before an enquiry has been accepted and stored successfully.
Google Analytics separates automatically collected, enhanced measurement, recommended and custom events. Important events can be marked as key events. Choose that status because of the action's meaning rather than marking every interaction as success.
Recommended names such as generate_lead and qualify_lead require appropriate configuration. Sending qualify_lead should reflect an actual qualification decision under agreed criteria. The name does not make that decision or define the criteria.
Use consistent names and campaign tags
Check the current naming rules and reserved lists of the platform. GA4 event names are case-sensitive, begin with a letter and use letters, numbers or underscores without spaces. Keep a shared event dictionary, so similar actions do not acquire several accidental names.
For manually tagged campaign links, agree source, medium and campaign values, and any useful content distinctions. UTM values are case-sensitive. Consistent lowercase conventions can reduce accidental fragmentation, provided they fit the reporting system's definitions.
Test the finished links, redirects and landing pages. Do not place personal customer details into campaign parameters. A naming convention is useful only when the team applies it to the published links.
Keep personal information out of ordinary collection
Inspect URLs, form fields, page titles and event parameters for inappropriate personal information. Google Analytics policies prohibit sending personally identifiable information through ordinary collection, including email addresses. Do not send private enquiry messages into an analytics report.
Represent approved states and categories where appropriate. Review any connection to CRM records against the platform's policies and the business's information-handling arrangements. A pseudonymous identifier still needs appropriate design and controls.
Validate the actual website behaviour
Test valid submissions, invalid answers, server failures and repeat actions. Compare the event sequence with the application and CRM outcome. Check whether one real action generates duplicates through several tags or retries.
Use supported diagnostics within the intended privacy configuration. Events can be absent from debugging views because of consent or privacy controls, so interpret the result alongside the test's settings. Do not weaken the live privacy arrangements merely to make a diagnostic event appear.
Enhanced measurement does not collect every internal interaction. Outbound click measurement concerns links leaving the current domain, with cross-domain configuration affecting treatment. Define and test internal navigation measurements explicitly when they are needed.
Select the right scope and metric
First-user, session and event acquisition dimensions describe different scopes. Choose the scope appropriate to the question. A customer's own answer about discovery is separate evidence, not another name for the source field most convenient to export.
The GA4 Lead acquisition report defines New leads as unique users with a generate_lead event, rather than every occurrence of that event. Qualified leads similarly relies on qualify_lead. Check the actual metric against the CRM's definition before expecting matching totals.
CRM reports can distinguish influenced contacts, attributed revenue and original or latest traffic sources using their own definitions. Record the relevant product configuration and calculation instead of assuming that identical labels mean identical measures.
Make comparisons deliberate
Record reporting dates, timezone, filters and comparison method. Matching days of the week may be useful where trading patterns vary. A convenient percentage does not establish that the change being investigated caused the difference.
Plan experiments with a hypothesis, suitable allocation, outcome definition and enough observations for meaningful analysis. Check instrumentation before beginning. New and returning user metrics describe what the measurement system recognises, not whether someone has bought from the business.
Measurement review checklist
- Document event meanings and important outcomes.
- Apply consistent names and campaign values.
- Inspect collection for inappropriate personal details.
- Test consent, errors, duplicates and real delivery.
- Check scope, metric definitions and report settings.
- Record the context of comparisons and experiments.
The paid search measurement guide covers bidding and conversion settings. Our CRM handover guide explains operational enquiry records.
