Interventional psychiatry has one of the longest and most private consideration cycles in healthcare. Patients research for months, discuss it with a therapist, raise it with a psychiatrist, tell nobody else, and eventually call. Standard attribution cannot see any of that.
Why simple attribution fails here
- Long cycles. Months between first awareness and first call, well beyond most tracking windows.
- Many touches. Search, a directory, your site, an AI answer, a conversation with a clinician, your site again.
- Offline decisive moments. The therapist conversation that actually decides it is invisible.
- Cross-device and cross-context research. Phone late at night, laptop later, no shared identity.
- Deliberate privacy. Patients researching psychiatric treatment often use private browsing, and you should expect and respect that.
- Untrackable channels. Payer directories and referrals pass no data.
The consequence is that last-click attribution systematically over-credits branded search and under-credits everything that created the demand.
The privacy constraint comes first
Before any attribution design, accept a hard boundary. In a psychiatric practice you should not be building individual-level tracking of who researched what.
- Do not attempt to identify individual website visitors.
- Do not send treatment-page URLs or condition-specific parameters to advertising platforms, since that can expose sensitive inferences about identifiable people.
- Be extremely cautious with tracking pixels on treatment and booking pages. Regulators and enforcement bodies have taken a keen interest in tracking technologies on health websites.
- Prefer aggregate measurement over individual journeys.
This rules out several techniques common in other industries. That is appropriate, and it is not a serious loss, because the approach below works without them.
The approach that works
Combine one direct question with aggregate trends:
- Ask at intake, always. How did you hear about us, recorded into a structured field, per chapter eight. This is your primary attribution instrument. It captures the invisible channels nothing else can see: referrals, payer directories, and word of mouth.
- Ask a second question for high-value treatments. What made you decide to call now. The answers reveal the actual decisive moment, which is frequently a clinician conversation rather than any marketing you paid for.
- Track channel trends in aggregate. Profile actions, search impressions and clicks, AI presence rate, directory coverage.
- Compare trends against enquiry volume over months, not weeks.
- Maintain a change log so you can align what you did with what moved.
Reading it honestly
Accept correlation rather than pretending to causation:
- If directory accuracy improved and insured enquiries rose two months later, that is credible evidence, stated as such.
- If you fixed everything at once, you cannot separate the causes. Change one significant thing at a time where you can.
- If enquiries rose the week after a change, that is probably noise. These cycles are long.
- Expect lag. Work done this month may show up two or three months out.
What to actually judge
Rather than trying to attribute each patient, judge the whole system on a few figures:
- New patient enquiries per month, trended.
- Enquiries by source, from the intake question.
- Enquiry to evaluation rate.
- Evaluation to treatment rate.
- Cost per new patient for genuinely paid channels only.
The two conversion rates are the ones practices neglect. A practice with plenty of enquiries and a poor evaluation rate does not have a visibility problem, and no amount of listings work will fix it.
Be sceptical of certainty
Anyone presenting a precise attribution model for this specialty is overstating what the data supports. The honest position is that you can see channel trends, you can ask patients directly, and you can correlate carefully against a dated change log. That is enough to make good decisions, and it is more than most practices have.
Building a report an owner will read is next.