Connecting Social Media Measurement to Admissions

Why platform dashboards stop where the admissions team begins

Meta’s Ads Manager and LinkedIn Campaign Manager will report reach, impressions, engagement rate, URL clicks, and follower growth without complaint ¹. What they will not report, and increasingly cannot report, is whether a person who saw a paid post later called the admissions line, completed a verification of benefits, showed up for an assessment, or was admitted. That gap is not a data-integration oversight. It is the direct consequence of how substance use disorder information is regulated and enforced.

HHS OCR has stated that pixels, cookies, and similar tracking technologies may not be used in ways that produce impermissible disclosures of protected health information, and a privacy-policy notice alone does not authorize disclosures to a tracking vendor ². The 42 CFR Part 2 final rule, effective April 16, 2024, with compliance required by February 16, 2026, keeps heightened protections around SUD records even as it improves coordination with HIPAA ³. In April 2024, the FTC banned Monument from disclosing health information for advertising after alleging that custom events and identifiers sent to Meta and Google exposed details about therapy and medication management ⁴.

The practical result: platform dashboards end at the click. The admissions record begins on the other side of a compliance boundary, and the measurement system has to be built to span it.

The two-layer measurement model

The compliance perimeter that reshapes the stack

A KPI hierarchy mapped to systems of record

Every social metric belongs to exactly one system. When a CMO cannot name that system for a given number, the number is either duplicated across dashboards or exposed in a place it should not live. The matrix below is the operational anchor: each row is a metric, and each metric has one owner, one storage location, and one privacy constraint that governs whether it can be shared with a third-party ad platform.

| Metric | Data owner | Storage system | Third-party sharing constraint | Reference ||—|—|—|—|—|| Reach, impressions | Paid social manager | Ad platform (Meta, TikTok, LinkedIn) | Platform-native, no PHI | ¹|| Engagement rate | Paid social manager | Ad platform | Platform-native, no PHI | ¹|| Follower growth | Organic social lead | Ad platform, native analytics | Platform-native, no PHI | ¹|| URL clicks | Paid social manager | Ad platform, GA4 | Coarse click signal only | ¹|| Time on page, bounce rate, pages per visit | Analytics owner | GA4 or first-party analytics | No PHI attached to session | ¹⁰|| Qualified site event (form start, chat open) | Analytics owner | Server-side endpoint, first-party warehouse | De-identified visitor ID only; no service-line labels sent to ad platforms | ²|| Tracked admissions call | Call center director | Call-tracking platform under BAA | Recording and transcript are PHI; only coarse call-occurred signal may cross the boundary | ²|| VOB submitted | Admissions operations | CRM under BAA | PHI; never transmitted to ad platforms as an event | ²|| Assessment scheduled or completed | Clinical intake | CRM or EHR under BAA | PHI; subject to 42 CFR Part 2 consent rules | ³|| Admission (treatment episode) | Clinical operations | EHR | PHI; subject to 42 CFR Part 2 consent rules | ³|

Two rules govern the whole matrix. First, a metric moves down the table only through the server-side bridge, which strips clinical meaning before any signal reaches an ad platform. Second, ownership does not shift with the metric: the call center director still owns call quality even when a coarse conversion signal is echoed back to Meta for delivery optimization. Audit the stack row by row. Any metric that appears in two owners’ reports, or in an ad platform with a clinically meaningful event name, is a finding.

Interpreting engagement without confusing it with demand

A rising engagement rate on a family-facing Reel is not evidence that admissions will lift next month. It is evidence that the creative held attention among the audience the algorithm chose to deliver it to. The distinction matters because the largest peer-reviewed test of the question found that online social-support seeking was not significantly associated with substance-use outcomes among the adults studied ¹³. Interactions on recovery content describe an audience already inside a support conversation, not a pipeline of people about to call an admissions line.

The 2025 cross-sectional study of adults with SUD adds useful scope: 40% of participants with past-year treatment attendance used social media for recovery support, and women and younger adults were more likely to do so than adults 50 and older ¹². That describes a real audience for alumni content, aftercare, and community programming. It does not describe prospects sizing up a level-of-care decision, and confusing the two inflates the perceived value of top-performing organic posts.

A more useful frame comes from the systematic review of engagement measurement in mental-health interventions, which recommends assessing behavioral, cognitive, and affective dimensions rather than a single activity number ¹¹. Applied to social, that means treating comment sentiment, saved-post rates, and dwell time on landing pages as separate diagnostic signals from qualified calls. Engagement tells the creative team what to make next. Admissions volume tells the CFO what the channel produced. Neither number should be asked to answer the other’s question.

Audience and creative allocation by primary substance

Creative planning that treats “SUD prospects” as one audience wastes budget on the wrong hooks. SAMHSA’s TEDS 2024 recorded 1,583,823 admission episodes, with alcohol as the primary substance for 36.5% and methamphetamine for 14.2% ¹⁵. The remaining half is spread across opioids, cannabis, cocaine, and combinations, each with distinct family dynamics, referral pathways, and search language. A paid social plan that allocates spend proportionally to that distribution—or intentionally over-indexes on service lines the facility actually treats—will beat a plan that runs one generic “get help today” creative across every audience segment.

The allocation logic connects back to the two-layer model. Primary substance is a clinical field that lives in the CRM and EHR, and it cannot be echoed back to Meta or Google as a conversion parameter without recreating the exact disclosure pattern the FTC challenged in Monument ⁴. What can move is upstream: creative concepts, landing pages, and audience targeting built around alcohol-specific messaging (professional discretion, high-functioning presentation, family confrontation) or methamphetamine-specific messaging (crisis intervention, cognitive recovery, longer stabilization). Performance is then read at the campaign or ad-set level, where the creative variant is known but the individual respondent is not.

The reporting question is whether campaign-level admission mix, joined in the warehouse on visitor or lead ID, matches the intended allocation. When the alcohol campaign produces mostly opioid admits, that is a targeting or landing-page problem to solve in the next flight—not a signal to send to the ad platform.

Reporting social ROI to a CFO when platform conversions are intentionally suppressed

The CFO’s question is straightforward: what did paid social produce this quarter? The answer cannot come from Ads Manager, because the events that carry commercial meaning—VOB submitted, assessment completed, admitted—are deliberately kept off the platform side of the bridge to avoid the disclosure pattern the FTC challenged in the Monument case ⁴. Reporting has to be reconstructed in the first-party warehouse, and it has to be labeled honestly about what it is.

The workable format is a three-column monthly view:

1. Column one is spend by channel and campaign, pulled from the ad platforms.
2. Column two is qualified inquiries, tracked calls, VOBs, assessments, and admissions attributed to social through the visitor or lead ID join, pulled from the CRM and call tracker.
3. Column three is the derived unit economics—cost per qualified call, cost per VOB, cost per admission—calculated in the warehouse, not the ad platform.

Platform-reported conversions appear as a diagnostic column for delivery optimization, explicitly flagged as a coarse proxy rather than a financial number.

Two methodological notes belong on the report itself. First, attribution is last-touch by tracked session unless a multi-touch model has been validated against holdouts; the CFO should see the model named. Second, the suppression is intentional. HHS OCR guidance on tracking technologies and 42 CFR Part 2 require that clinically meaningful events stay inside systems governed by BAAs and consent ^{2, 3}, which means platform-reported ROAS will always understate true performance. The gap between platform-reported conversions and warehouse-reported admissions is a compliance feature, not a measurement failure, and quantifying that gap each month is part of the report.

For board-level rollups, pair the cost-per-admission line with a second indicator drawn from CDC’s evaluation frameworks: qualified calls to the admissions line, treated as a primary campaign outcome the same way CDC treats calls to a crisis line in behavioral-health campaigns ⁹. That indicator is defensible, auditable, and does not require exporting anything sensitive to a third party. It also gives finance a leading signal—qualified call volume moves weeks before admissions close—so social spend decisions do not have to wait for the full episode to complete.

If you manage multiple facilities: rollup rules for a bridged model

For CMOs running two or more facilities under one marketing team, the two-layer model does not scale by copying it. It scales by deciding which layer consolidates and which layer stays local. Platform-side reach, engagement rate, and follower growth roll up cleanly at the brand level because they carry no PHI ¹. First-party admissions data does not, because 42 CFR Part 2 consent applies at the disclosing program, and a shared warehouse that pools SUD records across facilities can trigger redisclosure obligations the individual programs never agreed to ³.

Three rollup rules keep the bridged model intact across a portfolio:

1. Assign each facility its own visitor and lead ID namespace, then map to a portfolio-level ID only inside the warehouse under an executed data-sharing agreement that matches each program’s Part 2 consent language ³.
2. Keep call-tracking numbers and CRM instances facility-scoped, with role-based access limiting cross-facility visibility to aggregated counts, not record-level detail.
3. Report portfolio ROI as a sum of facility-level cost-per-admission calculations rather than a pooled ratio, so a shared creative asset can be evaluated against each facility’s admission mix without exposing which lead admitted where.

The board sees one number. The compliance file shows how it was assembled.

Market context and what it does not tell you

SAMHSA’s NSDUH estimated that 18.2% of people 12 or older—52.6 million Americans—needed substance-use treatment in the past year, and only 19.3% of that group received it ¹⁶. That gap describes a population, not a pipeline. It confirms the category has durable demand, and it says nothing about how many of those 52.6 million live in a given catchment area, will encounter a specific facility’s paid social, or are within days of picking up the phone.

Two misuses of the figure are common in board decks. The first treats the 19.3% receipt rate as an implicit conversion benchmark, as if a facility should convert one in five reachable prospects. NSDUH measures self-reported need and treatment receipt across all pathways—self-referral, court, employer, family, primary care—not paid social response. The second misuse pairs the national gap with a systematic-review finding that 17 of 18 digital-intervention studies produced positive results on at least one substance-use outcome ¹⁴, then implies social ads produce recovery. That review evaluated clinical interventions, not advertising. Marketing generates qualified demand; treatment outcomes are measured separately in the EHR.