Analytics and measurement
A conversion tracking checklist before you spend
Check conversion definitions, event firing, consent, campaign attribution and reporting before paid campaigns launch, with practical tests and clear ownership.
OppRise Editorial · Reviewed · 7 min read
The short answer
Before spending, define a genuine business outcome, verify its Analytics and advertising settings, test successful and failed journeys, validate values and duplicates, check consent and attribution, and reconcile results with business records. Launch only when critical defects are resolved and someone owns ongoing monitoring.
What to take away
- Measure confirmed outcomes separately from clicks and other intent signals.
- Verify each reporting destination rather than trusting matching event names.
- Test failure paths, duplicate triggers, consent states and external providers.
- Document reporting differences before using CPA or ROAS to make decisions.
- Measure team efficiency with testing time, rework and defect coverage.
1. Define the outcome before configuring the event
Start with a measurement contract: the business outcome, its technical trigger, the system that confirms it and the person responsible. A button click is evidence of intent, not proof that an enquiry arrived. A checkout visit is not a paid order.
Hypothetical example: a plumbing business counts an enquiry only after its server accepts the form. Telephone clicks remain diagnostic events because a tap does not establish that a call connected. Qualified enquiries are recorded separately after review. This preserves funnel visibility without presenting every interaction as commercial success.
Choose a main outcome that is meaningful and observable. Optimising towards an earlier action may provide more feedback, but risks attracting people who never progress.
- Write inclusion and exclusion rules, including spam, staff tests and repeat submissions.
- Name the business record that confirms success: accepted enquiry, booking or paid order.
- Keep diagnostic actions separate from outcomes used to evaluate acquisition.
2. Separate GA4 key events from advertising conversions
Google distinguishes GA4 key events from Google Ads conversions. Marking an event as important in Analytics is not the same as verifying its advertising configuration. Review the destination account, conversion source, counting settings and campaign goals rather than assuming a familiar event name means everything is connected.
Hypothetical example: an accounting firm measures consultation requests through both a native advertising tag and an Analytics import. Both may be useful for comparison, but allowing both to represent the same outcome in campaign goals can inflate the apparent result or distort optimisation.
Keep a register showing each outcome, its Analytics event, advertising action and intended role. Preserve historical actions where needed for reporting, but document which actions current campaigns should use.
- Confirm account and property IDs, not just display names.
- Check which actions are primary or secondary and which goals each campaign uses.
- Record counting methods and attribution windows before launch.
3. Test successful, failed and repeated journeys
Test the business journey, not merely whether a tag appears in a preview tool. A valid event needs the right trigger, timing and payload. Check the browser event, the receiving platform and the operational record wherever access allows.
Hypothetical example: a course enquiry form displays a validation error when a required field is empty. If submission-button clicks trigger the conversion, failed attempts look successful. Triggering from confirmed acceptance avoids that error, provided the implementation exposes that state reliably.
Duplicate tracking can come from overlapping plugins, hard-coded tags and container tags. A clean first submission does not prove that reloads, back navigation or repeated callbacks are safe.
- Test valid submission, validation failure, network failure and rapid double-clicking.
- Reload the confirmation page and revisit it directly.
- Repeat on mobile and desktop, recording timestamps, expected counts and observed counts.
4. Validate values, currency and transaction identifiers
For purchases, a correct event count with incorrect revenue is still broken measurement. Document whether reported value includes GST, shipping, discounts and refunds. Different definitions can be legitimate, but comparison requires consistent treatment.
Hypothetical example: an order contains A$120 of goods after discount plus A$10 shipping. Sending 130 while finance reports 120 creates a predictable difference, not necessarily a tracking fault. Sending cents as dollars creates a much larger implementation error.
Use stable transaction identifiers to support duplicate detection and reconciliation, while avoiding personal information in identifiers. For leads, label assigned values as estimates. A nominal enquiry value is not realised revenue and should not be presented as such.
- Check numeric value, currency code and identifier against the order record.
- Test a discounted order and the intended refund-reporting process.
- Document whether deposits represent deposit revenue or full booking value.
5. Verify consent behaviour and third-party boundaries
Test accepted, rejected and changed consent states against your organisation's approved privacy requirements. Do not bypass consent to improve event counts. Confirm that event payloads and URLs do not expose names, email addresses, phone numbers or sensitive free-text answers.
Hypothetical example: a clinic sends visitors to an external booking provider. Its own website can observe the booking-link click, but confirmed appointments require supported provider instrumentation or a permitted reconciliation process. Label the click honestly if confirmation is unavailable.
Cross-domain measurement, embedded forms and payment redirects depend on provider capabilities and configuration. Browser-side and server-side implementations both need privacy review; moving collection to a server does not remove consent obligations.
- List every domain and embedded provider in the conversion journey.
- Check what survives redirects and what the provider actually exposes.
- Document expected measurement gaps instead of treating missing observations as zero demand.
6. Check campaign labels through the whole journey
Google Analytics uses manually supplied UTM parameters to populate campaign dimensions. Agree naming rules before distributing links: source, medium and campaign should describe the acquisition consistently. Use content labels when distinguishing placements or creative variants is useful.
Hypothetical example: a newsletter uses utm_source=customer_newsletter, utm_medium=email and utm_campaign=spring_service. Test the actual shortened or redirected URL, not just the original spreadsheet entry. Confirm the intended campaign information reaches the landing page and appears in the appropriate report.
Do not add acquisition UTMs to internal navigation. That confuses the meaning of campaign labels. For advertising connections using automatic tagging, check compatibility before adding manual parameters rather than assuming more tags improve attribution.
- Use a shared naming register with agreed casing and separators.
- Exclude personal information from campaign parameters.
- Test final destination URLs, redirects and relevant session-level reporting.
7. Reconcile discrepancies without demanding identical totals
Compare operational records, Analytics and advertising reports using named date ranges, time zones and definitions. Allow for processing delays, attribution rules, consent limitations and cancellations. Agreement is useful evidence, but exact equality is not always a realistic acceptance criterion.
Hypothetical example: a booking system records 24 reservations while Analytics shows 19 booking events. Investigate missing providers, consent states, delayed reporting and trigger failures before calling the difference normal. Conversely, an advertising platform may credit bookings under its own attribution rules without measuring additional unique reservations.
Set an investigation rule based on observed behaviour and business risk, not an invented universal tolerance. Avoid adding conversions across platforms as though their attributed audiences never overlap.
- Reconcile test records individually where safe identifiers are available.
- Maintain a discrepancy log with explanation, owner and next check.
- Keep attributed conversions distinct from incremental business outcomes.
8. Set a launch gate and an early monitoring plan
Classify defects by decision risk. A missing main conversion, leaked personal information or duplicated purchase value should block the affected launch. A missing optional creative label may be acceptable temporarily if documented with an owner and deadline.
Hypothetical example: an accommodation campaign has working enquiry tracking but an unverified external booking completion. If bookings are its core objective, pause launch until that measurement gap is resolved or a deliberately limited alternative is approved. Do not silently substitute enquiry clicks for reservations.
Record a rollback action and monitoring responsibility. Recheck after form releases, checkout changes, consent updates or tag-container publications. Low traffic may mean waiting for evidence, not concluding that a quiet dashboard proves failure.
- Save test evidence, configuration versions and approval date.
- Assign someone to inspect early traffic, conversion payloads and operational outcomes.
- Agree when to pause spend and who can authorise resumption.
9. Measure checklist efficiency without overstating the tools
Track active testing minutes, rework minutes, unresolved defects and time to sign-off. Hypothetically, reducing repeated account-ID checks could shorten preparation while leaving test coverage intact. Include maintenance and training time, and compare similar launches. A faster later launch does not prove the checklist caused the improvement.
Its product scope includes owned work in RiseProjects, source-labelled reporting in RiseInsights and tracking context in RiseCampaigns, with campaigns created paused following approval. These are not substitutes for implementation testing.
OppRise's Google Tag Manager connection is read-only: it can review container names and IDs, but never edits or publishes tags. Reporting depends on connected-source access and data quality; channel actions also depend on permissions, integration status and provider approvals. Review the linked product pages and solution overview before planning your workflow.
- Measure efficiency alongside defect escape rates, not speed alone.
- Keep tag implementation with an authorised technical owner.
- Treat period comparisons as observations, not causal proof or guaranteed savings.
Frequently asked questions
Is a thank-you page enough to track a conversion?
Only if it reliably follows confirmed success and cannot create false conversions through direct visits or reloads. Test those conditions rather than assuming the page is trustworthy.
Should every GA4 event become a key event?
No. Reserve key events for actions important to the business. Retain diagnostic events for funnel analysis, and separately check which advertising conversions campaigns use.
What if a booking provider cannot report completed bookings?
Track observable actions honestly, document the gap and investigate supported exports or integrations. Do not rename booking-link clicks as confirmed reservations.
Does passing this checklist prove advertising is profitable?
No. It improves measurement readiness. Profitability also depends on costs, margins, lead quality and fulfilment; attributed conversions alone do not establish incremental return.
Put this into practice with