Google Ads Attribution Rebuild: Clinic Case Study

A multi-location US medical practice came to me unable to answer one question: which campaign actually brings in bookings and calls? The Google Ads spend was real, but the analytics under it had rotted into a state where no number could be trusted. I audited the stack, tore out the dead weight, and rebuilt the measurement so that bookings keep their source and count once, and phone calls finally carry a source. Here is exactly what was broken and what I did about it.

Quick answer

I did not invent a lift number for this one, and I would be suspicious of anyone who did after a job like this. The honest, verifiable result is that the data now says something true. Bookings retain their source and are no longer counted twice, and calls that used to arrive anonymous now come with a campaign attached.

The problem: which campaign brings real bookings and calls?

The owner was spending on Google Ads across several locations and getting reports that contradicted each other. The Ads interface claimed conversions. GA4 disagreed. The call-tracking tool showed a third number. When three systems tell you three different stories, you stop making decisions and start guessing, which is an expensive way to run paid search.

The root cause was not one bug. It was years of accumulated cruft in GA4 and Google Tag Manager, plus two measurement gaps that quietly erased the source of the two things that actually make this business money: online bookings and phone calls.

What was broken

The audit of GTM, GA4, and the call-tracking setup turned up a familiar pile of problems. This is what accumulates when a container gets edited by several hands over a few years and nobody ever removes anything.

• Double-firing tags. The same conversion event fired from more than one tag, so key actions were counted twice. Any rate calculated on top of that was wrong before it started.
• Leftover legacy Universal Analytics. UA tags and a hardcoded property ID were still sitting in the container long after UA stopped collecting, adding noise and confusion.
• The booking conversion lost its source. The confirmation step ran in a context that started a fresh GA4 session with no referrer, so bookings filed as Direct or Unassigned. The campaign that earned the booking got no credit.
• Calls had no source. A large share of phone calls arrived in GA4 with no source at all, because the call-tracking tool was never wired to push a call event with its attribution into GTM.
• Google Business Profile and Maps collapsed into Direct. That traffic arrives with a weak referrer, and GA4's default channel grouping dumped it into Direct, hiding a real and separate acquisition channel.

None of this is exotic. It is the default end state of a tag setup that nobody prunes, and it is exactly the mess that good Google Ads conversion tracking is supposed to prevent.

What I rebuilt

I worked in one direction: make every conversion true, then make it legible. The rebuild went in this order.

1. Removed the dead and duplicate tags. I pulled the leftover Universal Analytics tags, the hardcoded IDs, and the duplicate conversion tags. One action, one tag, one event. This alone stopped the double counting.
2. Rebuilt the booking conversion to keep its source. I captured the original source (gclid and referrer) at first touch and carried it through to the confirmation step, so the booking event fires with the campaign that earned it instead of starting a blank session.
3. Scoped the booking event to fire once per session. A session-scoped condition and a booking identifier used only for de-duplication mean a refresh or a back-button never inflates the count.
4. Wired call source from the call-tracking tool. I connected the call-tracking tool into GTM so a completed call pushes an event with its source into GA4. Calls stopped arriving anonymous.
5. Built a custom channel grouping. A rule for Google Business Profile and Maps traffic pulls it out of Direct into its own channel, so local discovery stops masquerading as people typing the URL from memory.
6. Delivered a Looker Studio dashboard. The owner now sees bookings and calls broken out by source and campaign in one place, instead of reconciling three tools by hand.

The booking event that now fires carries only what measurement needs and nothing a medical practice must protect:

{
  "event": "booking_confirmed",
  "booking_id": "b_8f3a2c",        // used only to de-duplicate, never a patient identifier
  "value": 180,                     // a service price band, not a bill and not PHI
  "currency": "USD",
  "source_preserved": true,         // gclid / referrer captured at first touch
  "session_scoped": true            // fires once per session
}

That last point matters more here than in most industries. GA4 is not HIPAA compliant and never will be, so the entire rebuild was designed to keep protected health information out of analytics. Only a source, a timestamp, and a non-identifying value ever leave the system, any phone identifier that has to move between tools is hashed, and no patient data is persisted anywhere in the tracking layer. You can measure marketing honestly without ever putting a patient into a tool that was not built to hold one. If you later want to push real booking value back into Ads, the same discipline carries into an offline conversion tracking setup where an anonymous click id plus a value is the only thing that travels.

The result

The deliverable was not a bigger number. It was a trustworthy one. After the rebuild:

• Online bookings retain their traffic source and count once per session, so the campaign that earned a booking gets the credit and nothing gets double-counted.
• Tracked phone calls carry a source and campaign instead of vanishing into an anonymous, sourceless majority.
• Google Business Profile and Maps traffic shows as its own channel rather than inflating Direct.
• The owner has a Looker Studio dashboard that answers the original question directly: which campaigns produce bookings and calls, by source.

From there, the practice can actually manage spend. You cannot optimize toward a conversion the system keeps mislabeling, and you certainly cannot optimize toward a call it never saw. Fix the measurement first and every budget decision after it gets easier.

If your paid search reports argue with each other and you have quietly stopped trusting any of them, that is the symptom I fix. Send me your GTM container and a screenshot of your GA4 conversions, and I will tell you within a call whether it is a cleanup or a rebuild, and what the conversion tracking setup would involve before either of us commits to anything.