Short answer: stop trying to rebuild perfect user-level tracking. Collect fewer signals that you own, ask buyers directly how they found you, and use holdout tests to answer the questions attribution can no longer answer.
Accept the new baseline
Between consent banners, tracking prevention in browsers and ad blockers, a meaningful share of sessions is simply not measured. This is not a configuration error to fix. It is the environment.
The practical consequence: your analytics numbers are a consistent sample, not a census. They are still useful for comparing periods and campaigns, as long as you do not compare across a change in consent behaviour without noting it.
Own the signals that matter
First-party events collected on your own domain survive most of these restrictions.
- Fire one key event on a successful enquiry, not a dozen weak ones.
- Capture campaign source on landing and keep it for the session, then send it with the enquiry.
- Store the enquiry with its source in your CRM, where it is no longer subject to browser policy.
The event mechanics are covered in GA4 events for lead generation.
Ask the buyer
The cheapest attribution tool is an optional field on the form: how did you hear about us? Self-reported data is messy and biased toward recent memory, and it still routinely beats modelled attribution for offline and word-of-mouth channels that analytics never sees at all.
Compare it with your analytics report. Where they disagree systematically, the gap usually names the channel your tracking cannot see.
Run holdouts instead of arguing about credit
When attribution cannot settle a question, an experiment can. Turn a channel off in one region or for one segment for a defined period, and watch total enquiries rather than channel-attributed ones.
It is blunt, it costs something, and it answers the question that matters: what happens to the business when we stop.
Respect consent, and say what it costs
If analytics only loads after acceptance, your reports under-count by the share of people who decline. Record the date you introduced the banner and treat it as a break in the series. In Thailand, the obligations around personal data are set out by the Personal Data Protection Committee.
A measurement setup that quietly ignores consent is not more accurate. It is a liability dressed as a dashboard.
Report in a way that survives scrutiny
- One definition of a qualified enquiry, agreed with sales.
- One key event, unchanged for at least two quarters.
- Self-reported source shown next to analytics source, not merged into it.
- A written note of every change that breaks comparability.
Where to go next
Decisions about spend come next: splitting paid and organic budget. If your pages are not being found or quoted in the first place, start with AI search optimization.
Want the measurement plan rebuilt around first-party data? Book a fit call.