Resources
In-House vs Agency: Behavioral Health Marketing Economics
The decision comes down to two questions: what the option actually costs once every line is counted, and which functions you need to control directly. In-house tends to suit organizations with stable volume and existing admissions reporting; an agency tends to fit a documented capacity gap in buying or production. The right comparison unit is fully loaded cost per verified admission, not salary versus retainer.
By Mike Hulick, Founder, Marketing Powered

When each model fits
Picture the meeting where this decision gets made. The CEO has a recruiter packet for a marketing director. The COO holds the last agency invoice, with fees and media on separate lines if the contract was clear. The admissions director has a call log that does not match the campaign report. Salary and retainer are the only numbers on the table, and neither one is the right comparison.
Consider in-house when most of the following are true for your organization:
- Admissions volume and service lines are stable enough that a full-time marketer has continuous work, not burst launches followed by idle months.
- Intake already records call outcomes, verification of benefits where applicable, and admissions in a system someone maintains.
- You can recruit and retain a media buyer or marketing lead who works inside current sensitive-category advertising rules, with a backup plan if that person leaves.
- Leadership wants marketing, clinical review, and admissions using the same definitions.
- You can fund ramp time, expressed as money and calendar months, before the function produces useful reporting.
Consider an agency when its proposed staffing and start date address a documented capacity gap: paid media depth you cannot hire, landing pages and tracking your current team cannot build, or coverage across locations one specialist cannot hold. A proposal that names who will work on the account and when work starts is evidence about capacity. A proposal that only describes the category is not.
Dividing functions is a legitimate option for several lines in this table. Treat a split as a hypothesis to test in the cost model, with a named owner for each function. This page provides a decision framework, not a comparative study; it does not establish that either model performs better. The same build-or-buy choice appears in hospital service lines, outpatient groups, and other regulated organizations. Intake steps, consent rules, and platform limits differ by service, so reuse the method and replace the inputs.

Why salary versus retainer misleads
A marketing director salary set next to a monthly agency fee is a partial invoice, not an acquisition cost. Fully loaded cost for a defined period includes paid media spend plus every operating line required to run that spend. If you leave media out, label the result "operating cost per admission, excluding media." Do not call it fully loaded acquisition cost.
For an in-house build, the lines a finance lead should expect are:
- Fully loaded labor: salary, benefits, payroll taxes, recruiting, counted once
- Incremental management time from the supervising executive, if not already inside labor
- Tool stack: analytics, call tracking, CRM, reporting, creative software
- Creative and landing page production the hire will not cover
- Allocated ramp cost: recruiting, training, delayed output converted to money across the analysis period
- Turnover contingency entered as a dollar amount, because departure restarts the ramp
- Paid media spend for the same period
For an agency relationship, expect a different mix:
- Agency fees separated from media
- Pass-through costs, including any fee calculated from media spend, entered once and not also inside the media line
- Paid media spend
- Internal time for briefing, approval, review, if incremental
- Onboarding cost before campaigns and tracking are running
- Switching cost if you leave: rebuilding tracking, pages, or exports you do not own
No salary band, retainer range, or waste percentage belongs in this model as a default. Those figures vary by market, scope, and contract. Enter your own numbers. If the media budget is identical in both versions, it raises both totals by the same amount and does not create the difference between them. It still belongs in the total, because cost per admission is unreadable without it.
Assign each function an owner
Choose an owner for each function before you sign anything. Do not choose one model for the entire department and discover at renewal that nobody owns the account or the definition of a qualified call.
- Strategy and budget allocation. Possible owner: In-house, or shared with an external advisor. Control to keep: Final decision on spend, service lines, target payer mix.
- Paid media buying. Possible owner: In-house or outsourced. Control to keep: Ad accounts in the organization's name, admin access retained.
- Creative and landing pages. Possible owner: Either or shared. Control to keep: Brand, clinical, and compliance review before launch.
- Tracking and analytics. Possible owner: Either, with an internal owner. Control to keep: Tag setup, analytics property, data exports under your control.
- Call handling and intake. Possible owner: Accountable internal owner, with execution evaluated separately. Control to keep: Definitions, access permissions and outcome reporting stay internal.
- Reporting. Possible owner: Shared. Control to keep: A report that joins spend to the deepest stage you can measure.
Keep internal accountability for intake definitions, access permissions, and outcome reporting regardless of where execution sits. If a contractor handles calls, you still need a written definition of qualified, a rule for who can open recordings, and a record of who reports the admission. Accountability is not the same decision as where the person answering the phone sits.
Platform rules are part of the buying function, not a footnote for legal to discover after an account is limited. Review current platform policies for your specific service category before assigning ad buying responsibilities, because location and category restrictions change. Whoever holds the buying seat needs working knowledge of the rule that applies to the service being advertised. Certification with a third party does not, by itself, settle every platform or legal requirement; consult counsel on what applies to your program structure.
Free Audit
Want a straight read on where your budget is leaking?
Measure spend through admission
Stand in the place where the numbers have to meet. An admissions coordinator finishes a call, applies a written definition of qualified, and records whether verification of benefits was completed and whether the person was admitted. Finance has the media invoice for that week. Marketing or the agency has a platform export of spend and clicks. The model comparison is only as good as the join among those three records. If they live in three systems and nobody reconciles them, you are comparing stories.
Build the chain in stages and assign an owner to each:
- Spend and clicks. Platform exports stored where you can retrieve them after a vendor change. Owner: marketing or the media vendor, with finance holding the invoice.
- Inquiries. Calls, forms, chats tied to a campaign source. Owner: whoever runs tracking, with an internal owner of the property.
- Qualified inquiries. One written definition applied the same way in every report. Owner: admissions, even if marketing drafts the first version.
- Verification of benefits completed. Use this stage only where verification is part of the path. Owner: admissions or patient access.
- Admission. Deduplicated, from a defined cohort and period. Owner: admissions, reconciled to census or equivalent record.
Do not assume marketing stops at inquiries and admissions starts later. If two teams use different definitions of qualified, one side will optimize toward a definition the other does not use. Write the definition before you compare models, and record who owns it. Judge both models at the same deepest stage you can measure with data you control. Cost per lead can diagnose a break between click and inquiry; it should not decide the staffing model.
Where HIPAA applies to records in this chain, assess any tracking or analytics setup against your organization's privacy requirements before enabling it. Consult legal counsel on what constitutes permissible use for your specific program structure. For programs subject to Part 2 confidentiality rules covering substance use disorder records, confirm applicability with counsel before sharing data across systems. Document attribution gaps and exclusions in the report instead of filling them with an assumed rate. Joining spend to admissions requires authorized data, and that join will have holes.

Build the cost model twice
Run the model once for each staffing option over the same period. The first pass holds inquiry volume and conversion rates equal so you are comparing cost structure, not a hoped-for change in performance. A second pass may change volume or conversion; label it a scenario, not a measured result.
Fully loaded acquisition cost equals paid media spend plus fully loaded labor plus incremental management cost not already in labor plus tools plus creative production not already in labor or fees plus agency fees and pass-through costs not already in media plus allocated ramp cost plus turnover contingency as money. Include each expense once. If call tracking or CRM is inside tools, do not add it again.
Scenario-estimated admissions are not observed admissions. Define each rate on the stage before it: qualified-inquiry rate equals qualified inquiries divided by inquiries; verification rate equals verifications completed divided by qualified inquiries where that stage applies; admission rate equals admissions divided by verified or qualified inquiries depending on whether verification is in the path. Do not force a verification stage onto a service line that does not use one.
Measured cost per admission equals fully loaded acquisition cost divided by deduplicated observed admissions from a defined cohort in the same period. That is the figure to use when records exist. If observed admissions equal zero, report cost and count separately; do not divide by a modeled number and call the result measured.
Name an owner for each step:
- Finance or COO fills labor, tools, fees, media, ramp, turnover contingency for each model and checks no line is counted twice.
- The admissions lead supplies qualified, verification, and admission counts from actual records, states period and cohort, and confirms whether verification applies.
- Marketing or the vendor supplies inquiry volume by source and the location of the export.
- CEO or CMO reviews the equal-outcome comparison first, then any labeled scenario, and decides which functions to move, keep, or test.
- The group records assumptions that move the result so the next review can challenge them instead of inheriting them.
If step two or three is blank, that is the finding. You cannot compare models until the chain exists. Fix measurement before you approve a hire or a replacement vendor. This framework does not establish which model performs better; outcomes depend on admissions conversion, payer mix, local competition, and record quality. It is a way to put your own numbers in one place, not a prediction.

Questions to settle before you sign
Before proposals arrive, the COO and CMO should lock scorecard weights with a blank sheet. Each vendor gets the same questions in the same order, in writing. A polished presentation does not get to rewrite the weights after the fact. The scene to prepare for is a conference table with the contract, ad-account access list, and admissions field list, not a capabilities deck.
Ask at least these:
- Account and data ownership. Are ad accounts, analytics properties, call-tracking numbers, recordings, and CRM exports in your organization's name? What is the written exit process and file format?
- Admissions reporting. Which fields can the vendor actually see: qualified inquiry, verification of benefits, admission? What will they state in writing that they cannot see?
- Category-specific advertising rules. For the services you advertise, which current platform restrictions apply, and how does the team demonstrate that knowledge in the account rather than a slide?
- Measurement design. How will tracking, call recording, and any join to outcome data be reviewed under your privacy and business-associate requirements?
- Team and access. Who works on the account, and how often do you speak with those people rather than a salesperson?
- Fee structure. How are fees, pass-through costs, and media spend separated in the contract?
- Evidence. Which references can you contact, and what can they confirm without handing over another organization's confidential results?
- Exit terms. What you keep, who keeps admin access during transition, how long exports remain available.
Document ownership and export rights before signing rather than discovering restrictions at exit. Include call handling in the comparison so the staffing decision does not ignore admissions execution. Use the same assumptions for both models; do not compare a favorable agency case with an unfavorable hire or the reverse. Do not count only the retainer. Do not split functions without a reason tied to a control you need to keep.
Rehab marketing costs are not a single market price. Payer mix, geography, and competition move them, and this page does not publish a benchmark. For budget construction, see the rehab marketing budget guide. Keep this comparison on ownership, measurement, and fully loaded cost.
Questions, Answered.
When you can sustain the workload, retain required skills, maintain admissions reporting, and fund ramp time before useful output. If those conditions are weak, complete the cost model with your own numbers before approving a hire. An agency fits when its staffing plan and start date address a documented capacity gap in buying or production that one internal specialist cannot cover.
Use one chain for both models: spend, inquiry, qualified inquiry, verification of benefits where applicable, admission. Judge both at the deepest stage you can measure with data you control. For a measured figure, divide fully loaded cost by deduplicated observed admissions from a defined cohort. Use rate-based projections only as labeled scenarios, not as evidence of performance.
The same questions in the same format for every vendor: who owns accounts, tracking, recordings, and exports; which admission fields the vendor can actually see; what they cannot measure stated plainly in writing. Put ownership and export rights in the contract before signature so you are not discovering limitations at the end of a relationship.
Hospital service lines, outpatient groups, and other regulated providers face the same build-or-buy choice. Intake definitions, consent rules, and advertising restrictions differ by service, so keep the method and replace the inputs. The operating questions for behavioral health organizations are the reference point here, not a boundary around it. If your attribution chain has blank stages or you cannot join spend to admissions with data you own, the next step is understanding where those gaps sit and what they cost before you commit headcount or vendor budget. Request an acquisition audit
Ready to see where your marketing can improve?
Request a free audit of your paid media, landing pages, attribution, and measurement setup. You'll get a straight assessment of where the opportunities are.
or email us at info@marketingpowered.ai