Key takeaways
- Measure qualified outcomes.
- Keep sample size visible.
- Distinguish clicks from signups.
Define a meaningful event
Choose a completed signup, qualified enquiry or booked walkthrough. A button click is not necessarily a business outcome. Check the event fires once and document exactly what it represents.
Keep original sources inspectable
Group identifiable assistant referral domains in analytics while retaining the source record. Test the rule on real visits. Some journeys lose referrer information; direct traffic alone is not evidence of AI discovery.
Use honest denominators
In an illustrative sample, five qualified enquiries from twenty observed referral visits is 25%. The sample is small and may not represent future traffic. Compare similar periods and pages before calling a channel better than organic search.
Fulfil the recommendation
If an answer presents your product as reporting software, the destination should explain reports, limits and next steps. Keep essential information visible. Do not require a form just to discover whether the product fits.
Check cross-domain attribution
A marketing-site click to a separate product domain is not a completed signup. Configure cross-domain measurement only where you control both properties and have appropriate consent. Otherwise report outbound actions and product conversions separately.
References and further reading
Primary documentation and product references. Practical checklists and worked examples above are editorial guidance, not claims about private ranking algorithms.