The chain, in order
Impression. Click. Call or form. Answered call. Qualified call. Assessment scheduled. Assessment completed. Admission. Every marketing report should be able to walk that chain and show a number at each step.
Most reports stop at step three. That is the boundary where the agency's data ends and the facility's data begins, and it is exactly where the useful information starts. An agency reporting through step three is reporting what it can see, not what you need.
The single most useful number is cost per admission by channel. It compares paid search against organic against referral against everything else on the same axis, and it is the only number that tells you where the next dollar goes.
The metrics that mislead
Impressions and sessions go up when you buy broader keywords, which is the easiest thing in the world to do and usually makes admissions worse. Keyword count goes up when you rank in position 47 for more terms. Bounce rate goes down when your tracking fires a second event.
Lead volume is the dangerous one, because it looks like the right metric. A campaign producing 200 form fills from people with no insurance and no intent to admit is worse than one producing 20 calls from families ready to place someone this week, and the first report looks better.
Ranking reports have a narrow use. Rankings for the specific terms that produce admissions in your market are worth watching. An average position across 400 keywords is a number that cannot be acted on.
Call quality is the real measurement problem
In behavioral health the conversion event is a phone call, which means the measurement lives in the call. Volume alone tells you nothing, because a large share of inbound calls to any treatment center are wrong numbers, vendors, current clients, or people outside your payer mix entirely.
The fix is call tracking with recording, plus a scoring pass where someone tags each call: qualified, unqualified, existing client, spam. Once calls are scored, cost per qualified call becomes real and every channel comparison changes.
Two operational numbers come free with that setup, and both cost programs more admissions than any marketing decision does. Answer rate, the share of inbound calls a human picked up. Speed to lead, the minutes between a form submission and the first outbound attempt.
Attribution in a long, private decision
Families research for weeks and use several devices, often without telling anyone. The last click before the call is frequently a branded search, which makes SEO and paid brand terms look brilliant and everything upstream look useless.
Do not over-engineer this. Track the first touch and the last touch, ask a source question during intake, and accept that some share is unattributable. A directionally correct number you act on beats a precise model nobody trusts.
The HIPAA constraint is real here too. HHS has published guidance on the use of online tracking technologies by covered entities, parts of which have been challenged in court. Whatever the final shape, know what your tags collect on condition-specific pages and who receives it. Server-side tracking and call tracking without third-party pixels on clinical pages are the usual safe paths.
What a report should look like
One page. Calls by channel, qualified calls by channel, cost per qualified call, admissions by channel where the CRM supports it, and cost per admission. Then a short list of what changed and what happens next month.
If a monthly report takes twenty slides to say the same thing, the length is doing work the numbers cannot. Ask for the one-page version and see what survives.
Getting to that report is mostly plumbing: call tracking numbers on every channel, a CRM field that carries the source, and a Salesforce or equivalent build that does not lose the source between the first call and the admission. That work is unglamorous and it is what makes every other number honest.