Reproductive-health data
What may safely travel back to an advertising platform
Closed-loop measurement works when the data returned to an ad platform is limited to a click identifier, a neutral stage name, and a value. Everything that identifies the person or describes their care stays inside clinic-controlled systems. This note maps the boundary.
Clinics hear two contradictory messages. Advertising vendors say better measurement requires sending more data back to the platform. Privacy counsel says health-related tracking is under active regulatory scrutiny, citing HHS guidance on tracking technologies and FTC enforcement around consumer health data. Both messages contain truth, and the way through is precision about what actually needs to travel.
The signal the platform needs
Bidding systems optimize against conversions attached to clicks. To close the loop, the platform needs to learn that a specific past click eventually produced an outcome the advertiser values. That requires exactly three things: the click identifier it issued in the first place, an event name, and optionally a value and timestamp. It does not require knowing who the person is or what care they sought.
| Stays inside clinic systems | May travel in a well-designed flow |
|---|---|
| Patient identity and contact details | The click identifier the platform issued |
| Diagnosis and treatment details | A neutral stage name such as consult_scheduled |
| Clinical notes and outcomes | A date or timestamp where allowed |
| Patient-portal activity | An assigned value for the stage |
Neutral naming is doing real work
Event names are data. A conversion called ivf_consult_booked encodes treatment information into a system the clinic does not control; a conversion called consult_scheduled does not. The same discipline applies to URLs, which should never carry treatment types, and to form-field values, which should never ride along with analytics events. These are design choices made once, at setup, and they cost nothing in measurement quality.
What should not happen
- Treatment types encoded in event names or URLs.
- Patient lists uploaded to advertising platforms for matching or audiences.
- Remarketing audiences built from visitors to fertility-treatment pages, which personalized-advertising policy restricts regardless of technical capability.
- Tracking scripts firing inside authenticated patient portals.
- Free-text form content forwarded to analytics or advertising tags.
The review that settles it
A useful exercise for any clinic: produce a one-page data-flow map listing each event that leaves the website or CRM, the fields it carries, and the system that receives it. If the map cannot be produced, that is itself the finding. If it can, counsel reviews a short factual document instead of an abstraction, and the marketing team knows exactly which changes would cross a line. Hyphen builds this map as a standard part of diagnostic work.