Write an outcome contract

For each conversion, record the business definition, event name, authoritative system, eligibility rule, deduplication key, value and currency behavior, owner, and test procedure. “Lead” might mean any form submission, a valid contact request, or a sales-qualified opportunity; those are different outcomes and should not share one unlabeled metric.

Separate micro-conversions that diagnose progress from final outcomes. A pricing-page view, form start, and completed enquiry can belong to one funnel, but adding all three into a single conversion total exaggerates success and prevents a useful drop-off analysis.

Choose the denominator before the result

A session conversion rate asks what share of eligible visits produced the outcome. A user rate asks what share of estimated people did so during a window. A landing-page rate may include only sessions that began on that page. State the denominator in the report title and keep it consistent across comparisons.

Exclude only with a documented rule. Removing bots, employees, test traffic, unsupported markets, or already-converted accounts may be reasonable, but an exclusion chosen after seeing the result creates bias. Preserve the raw count and publish the eligibility logic.

Implement and deduplicate

Use browser events for interface milestones and authoritative server events for accepted transactions or created records. Assign one stable event ID when both paths report the same outcome. Validate required parameters, reject impossible values, and protect the collection endpoint from retries or replay.

Test success, validation failure, refresh, back navigation, network interruption, duplicate clicks, payment rejection, consent choices, and cross-domain handoff. Reconcile analytics conversions with the system of record on a schedule and annotate implementation changes.

Report the uncertainty and limits

Show conversion count and eligible count beside the percentage. For a rate, include a confidence interval or at least warn when the sample is small. A change from one conversion to two is a 100% increase but rarely grounds a strategic conclusion. Use experiments when the question is causal.

Attribution adds another uncertainty layer. Campaign credit depends on observed touchpoints, identity, lookback, and model. Keep the business outcome total separate from channel-attributed totals, and explain why attributed rows may be revised or fail to sum cleanly.

Direct answers

Questions this guide answers

What is conversion tracking?

It is the recording and validation of meaningful outcomes, connected to an eligible visitor or session population.

Should conversion rate use users or sessions?

Use the denominator that matches the decision, state it explicitly, and do not switch between users and sessions in the same comparison.

How are duplicate conversions prevented?

Assign a stable transaction or event ID and deduplicate retries, browser/server copies, refreshes, and repeated callbacks.

Evidence

Primary-source ledger

  1. Mark events as key eventsGoogle Analytics Help · accessed 30 July 2026
  2. Manual events and goalsClicky · accessed 30 July 2026
  3. Confidence intervals for proportionsNIST · accessed 30 July 2026

Scope note: A tracked conversion is an observed implementation event. It does not establish incremental impact unless the measurement design supports a causal conclusion.

Found an outdated fact or a material omission?Read the corrections and recheck policy.