Paid acquisition into clinical and channel buyers
An audience architecture for reaching clinical titles on LinkedIn
A device company launching paid social into clinicians usually starts the way a software company would. Pick the titles, add seniority, exclude the obvious noise, split by persona. Then the audience estimate comes back too small to save, and the team widens it until it no longer describes anyone in particular.
The problem is arithmetic. Clinical titles are small, fixed populations, and the whole country’s supply of some of them is smaller than a single ad set is supposed to be.
How many clinicians are there to target in the first place?
Fewer than most plans assume. Federal occupation data for 2025 counts 267,330 physical therapists, 30,500 athletic trainers and 9,390 orthotists and prosthetists in US jobs. Every LinkedIn audience built on those titles is a subset of those numbers, so the ceiling is set before a single facet is chosen.
The Bureau of Labor Statistics’ Occupational Employment and Wage Statistics program publishes these counts. They describe the population a profile could possibly match, before any platform filter applies.
LinkedIn’s own guidance says the minimum audience for an ad set is 300 member accounts, and suggests a minimum of 50,000 to drive results. For Sponsored Content it suggests 300,000. Put the two sources side by side and the shape of the problem is obvious. Every athletic trainer employed in the US amounts to 61% of the suggested general floor. Every physical therapist amounts to 89% of the Sponsored Content floor. Of the 13 clinical titles in the file below, 6 employ fewer than 50,000 people nationally, before any question of how many of them maintain a profile or match a title string.
A software playbook assumes the audience is large and the job is to narrow it. In clinical targeting the audience is already narrow, and the work is pooling titles without losing the ability to tell them apart afterwards.
What does the audience architecture look like?
Nine cells, grouped by the role a title plays in the sale: the clinical user who touches the device, the clinical prescriber who recommends it, the economic buyer who funds it, and the channel rep who carries it. Each cell is one ad set, and its name encodes the grouping so reporting can be read without opening the targeting panel.
| Cell name | Role | Title group (BLS anchor) | Facet logic | Exclusions |
|---|---|---|---|---|
US_CU_REHAB_TTL | Clinical user | PT, PTA, OT, OTA (29-1123, 31-2021, 29-1122, 31-2011) | Job Title OR list, current or past | Own employees; competitor company list |
US_CU_SPORT_TTL | Clinical user | Athletic trainer, exercise physiologist (29-9091, 29-1128) | Job Title OR list, current or past | Own employees only |
US_CU_SPORT_SKL | Clinical user | Same people, found by skill | Member Skills AND a clinical job function | Members already in US_CU_SPORT_TTL |
US_CU_RESP_TTL | Clinical user | Respiratory therapist (29-1126) | Job Title OR list | Own employees; competitor company list |
US_CU_OANDP_TTL | Clinical user | Orthotist, prosthetist (29-2091) | Job Title OR list, current or past | Own employees only (never an industry exclusion) |
US_CP_ORTHO_TTL | Clinical prescriber | Orthopedic surgeon, sports medicine and PM&R physicians, orthopedic PAs (29-1242, 29-1229, 29-1071) | Job Title OR list; no seniority | Own employees; competitor company list |
US_EB_CLINIC_TTL | Economic buyer | Clinic director, practice owner, rehab manager (11-9111) | Job Title AND seniority | Clinical-user cells |
US_EB_SUPPLY_FN | Economic buyer | Purchasing, supply chain, value analysis (13-1020 in hospitals) | Job Function AND seniority AND employer list | Clinical-user cells |
US_CH_DIST_LST | Channel rep | Sales rep, territory manager (41-4011) | Uploaded distributor company list AND sales titles | Own employees |
naming: {geo}_{role}_{group}_{method}[_{variant}]
geo US | country code | state code for a regional test
role CU clinical user touches or fits the device
CP clinical prescriber recommends or orders it
EB economic buyer funds or contracts for it
CH channel rep carries it for a distributor
group REHAB | SPORT | RESP | OANDP | ORTHO | CLINIC | SUPPLY | DIST
method TTL job title SKL member skills
FN job function LST uploaded company list
variant optional, for a creative or offer test: _A, _B
rules
one ad set = one cell name; never mix roles in a cell
clinical cells carry no seniority facet
buyer cells always carry a seniority facet
a SKL cell always excludes its TTL sibling, so no member is counted twiceHow does each cell work?
Clinical cells use title lists with OR logic and no seniority layer. LinkedIn combines several values inside one facet with OR, so a long list of title variants grows the audience. Adding a second facet narrows it. For populations this small, every added facet costs more reach than it buys in relevance.
The title list is the load-bearing part. LinkedIn says its Job Title attribute is primarily member-entered, with inference applied when a member types a free-text title. Clinicians write their titles in dozens of ways: credential suffixes, setting prefixes, “staff” and “senior” variants. A cell with three canonical titles and none of the variants is a cell that silently misses most of its population.
The current-or-past option matters more here than in most categories. LinkedIn lets an advertiser target past holders of a title. A physical therapist who moved into clinic management is still a physical therapist for the purposes of evaluating a rehab device, and a current-only cell drops them at exactly the moment they gain budget authority.
Seniority is deliberately absent from clinical cells. LinkedIn maps job titles to seniority, and to job function, using its own taxonomy. A staff therapist and a lead therapist judge a device the same way, but a seniority layer can split them. Seniority belongs in the buyer cells, where rank changes who signs.
The skill cell exists for the smallest populations. LinkedIn suggests skills targeting where title targeting is less effective, and a sports medicine audience is the clearest case: athletic trainers work across schools, clinics and hospitals, often under titles that do not say “athletic trainer” at all. Excluding the title cell from the skill cell keeps the two from counting the same member twice.
Where do the clinicians actually work?
In very different places, title by title. Respiratory therapists sit overwhelmingly in hospitals; athletic trainers sit mostly in educational services; physical therapists split across outpatient offices, hospitals and home health. Any company-industry facet or employer list has to follow that split, or it removes the people it was meant to find.
In the settings file, 80.2% of respiratory therapists work in hospitals. Only 37.9% of physical therapists work in offices of physical, occupational and speech therapists, with 26.9% in hospitals and 11.6% in home health. Athletic trainers have 39.3% in educational services. A single hospital-industry facet would capture most respiratory therapists and miss most athletic trainers.
The most useful line in the file is the one that argues against a reflex. Orthotists and prosthetists have 25.2% of their national employment inside medical equipment and supplies manufacturing. The reflex in device marketing is to exclude manufacturing employers to keep competitors out of the audience. For this title, that exclusion removes a quarter of the population, most of whom are fitting patients in manufacturer-owned practices.
Which exclusions help, and which ones backfire?
Exclude people, not categories. An exclusion on your own employees, and on an uploaded list of named competitors, removes members who will never buy. An exclusion on an industry, a seniority band or a broad function removes clinicians who look like noise from outside the category and are, in fact, the audience.
LinkedIn’s help pages define an exclusion plainly: it reduces the audience to members who meet the earlier selections and do not meet the excluded criteria. A member who matches an excluded industry is gone, however well they fit the title list.
Company Industry, by LinkedIn’s definition, is the employer’s stated primary industry. The employer chooses it. That is the mechanism behind the orthotist problem: the practice says it is a manufacturer, so every clinician on its payroll inherits the label.
Selections are also finite. LinkedIn permits at most 200 selections across demographic targeting, inclusions and exclusions combined. A rehab title list with credential variants can run long, and a team that spends that budget on exclusions will run out of room for titles first. Competitor exclusions belong in an uploaded company list, which LinkedIn accepts at up to 300,000 companies, rather than as individual selections.
How do you keep clinical user, economic buyer and channel rep apart?
Give each role its own ad set and never let one cell hold two roles. The clinical user judges the device, the economic buyer judges the spend, and the channel rep judges whether it is worth carrying. They need different creative, different offers and, above all, separately reportable results.
The economic buyer is the one role where seniority earns its place. Federal data counts 17,180 medical and health services managers inside therapy offices and 11,450 buyers and purchasing agents inside hospitals. Those are the people who sign, and a seniority layer is the cleanest way to separate them from the clinicians who report to them.
The channel cell is the hardest to size honestly. The file counts 86,680 technical and scientific sales representatives at professional and commercial equipment wholesalers, a category that includes medical equipment distributors and a great deal besides. Title alone cannot isolate the reps who carry orthopedic or respiratory lines, which is why that cell is built on an uploaded distributor list. The same logic governs how channel demand is recorded once it arrives: the rep who books the order and the source of the demand are different parties, and a lead-source schema built for distributor channels keeps the two apart.
Where a category is crowding (the pattern in where new device competitors are entering), the clinical-user cells feel it first, because every entrant targets the same small population.
Where does it break?
Four places. National employment is only a ceiling, so the true audience is smaller by an unknown margin. Sports medicine physicians have no occupation code of their own. A cell that runs too long fatigues a population that cannot be refreshed. And retargeting needs care on condition pages.
The ceiling overstates the audience. Nothing in the federal data says how many of these clinicians hold a LinkedIn profile, keep it current, or list a title the platform can map. The architecture is built to survive a small match; it cannot reveal what the match is.
Physicians hide in one bucket. Sports medicine and physical medicine and rehabilitation physicians are counted inside Physicians, All Other, which holds 342,720 jobs, 24 for every orthopedic surgeon in the file. The prescriber cell has to rely on title strings, and the bucket’s size says nothing about the specialty’s.
Small populations fatigue. An audience of a few thousand orthotists sees the same creative quickly. Rotating variants within a cell, rather than widening the cell, keeps the audience definition stable while the message changes.
Retargeting has its own limits. LinkedIn says its Insight Tag should not be installed on pages that collect or contain sensitive data. A device site with patient-facing condition pages needs a deliberate decision about which pages carry the tag before any retargeting cell is built on them.
How was this built?
From two queries against the Bureau of Labor Statistics OEWS May 2025 estimates: national employment for 19 title and industry pairs, and employment for six clinical titles across seven care settings. LinkedIn facts come only from LinkedIn’s help pages, fetched and snapshotted on the same day. No LinkedIn member count is asserted anywhere.
The data was retrieved on 2026-09-30. The BLS public API refused the request under its daily quota, so the recipe streamed the same series from the published OEWS flat file instead; the sources file records which path served the numbers. OEWS counts wage and salary jobs. Self-employed people are not in the estimates, which undercounts owner-operators such as practice-owning therapists and chiropractors. Employment estimates are rounded to the nearest ten.
Standard errors vary. Orthotists and prosthetists carry a relative standard error of 9.7%, orthopedic surgeons 10.9%, and sales representatives in medical equipment manufacturing 13.6%. Treat those rows as directional. The role column is Carmine’s classification of each title’s part in a device sale; it is not a BLS field.
What the data cannot tell you is anything about behaviour. A count of respiratory therapists says nothing about how many influence a purchase, read LinkedIn, or respond to an ad. It sets the size of the room. Who in it is listening is a question for the campaign’s own results, and for the evidence base covered in who generates device clinical evidence.
What is the smallest first step?
Build the title list for one clinical cell, properly. Pull every variant the population actually uses, set it to current or past, add only an own-employee exclusion, and name it with the convention. One well-built cell teaches more about the population than nine rushed ones.
Then check the forecast against the ceiling. An estimate near the national employment for that title means the cell is picking up people it should not. A tiny fraction means the title list is missing variants.
The rest follows in order: the matching skill cell for the smallest populations, then the buyer cells with seniority, then the channel cell on an uploaded list. The paid acquisition hub collects the related pieces.