How we cut cost per acquired patient 29% for a multi-state telehealth provider.
7-month engagement (ongoing)·Google Ads + Microsoft Ads + landing page experimentation
A multi-state telehealth provider in a specialty vertical was spending $74K/month on Google Ads, hitting an attractive-looking $162 cost per form submission, and watching the clinical team complain that the patient pipeline was full of one-visit dropouts and non-eligible inquiries. The previous agency was optimizing toward form fills, which is the same as optimizing toward whoever’s most likely to fill out a form regardless of whether they were a good clinical fit or had any intent to complete care. We rebuilt the conversion event hierarchy around patient lifetime value, scrubbed PHI from the data layer to keep everything HIPAA-clean, ran a structured landing page CRO program, and let Smart Bidding learn the shape of a high-LTV patient. Seven months later, cost per acquired patient (the metric the clinical team and CFO actually care about) dropped 29%, monthly new patient volume rose 58%, and projected 12-month patient revenue from paid search more than doubled on roughly flat spend.
−29%
Cost per acquired patient
from $763 to $363
+58%
Monthly new patient volume
97 → 226 / mo
2.2x
Projected 12-month revenue
$2.4M → $5.3M (paid search)
Our previous agency bid for form fills and showed us beautiful CPA numbers. The patients converting were the wrong patients. SMB Circus rebuilt our bidding around patient lifetime value and the same spend started bringing in the patients our clinical team actually wanted.
What we walked into
The provider had been live in eight states for two years, was scaling clinical capacity, and had paid search as its primary acquisition channel. Revenue was real and growing. The clinical and growth teams had been arguing for months about whether the marketing was working.
Five findings from the first-week audit:
The conversion event was a form submission with no downstream signal.
Google Ads had visibility into form fills only. No visibility into eligibility-screen-passed, no visibility into intake-call-completed, no visibility into first-paid-visit, and certainly no visibility into 90-day retention or full-course-of-care completion. Smart Bidding was optimizing toward the easiest possible patient to acquire, which is structurally not the most valuable patient to acquire.
No PHI hygiene in the data layer.
The previous agency had been firing standard GA4 events and the Meta pixel on patient portal pages that displayed visit history and clinical information. This is a HIPAA-policy violation that puts the provider at material regulatory risk and could result in Meta and Google enforcement actions against the ad accounts. The provider was effectively one audit away from a serious problem and didn’t know it.
Healthcare ad policy compliance was a coin flip.
The account had 14 active “Limited Ad Targeting” flags Google had applied to ad groups containing sensitive healthcare terms, and the previous agency had ignored them. Some ads were running with disapprovals that hadn’t been remediated for months, and one ad group had been suspended for a policy violation that nobody had been notified about because the account email routing was broken.
Every ad pointed to the same generic homepage.
A search for “telehealth for [condition] [state]” landed on the same page as “online doctor visit [state].” No condition-specific landing pages, no state-specific compliance copy, no eligibility pre-screen on the page itself, and a clinical intake form that took 11 minutes to complete and had a 38% abandonment rate.
No measurement of patient lifetime value tied back to ad data.
The provider’s billing system tracked patient revenue accurately, but those numbers had never been connected to which campaign, ad group, or keyword had originally driven the patient. The clinical team could tell you the average revenue per patient, but couldn’t tell you whether the patients coming from “online doctor” queries were producing the same revenue as patients coming from “[condition] specialist online.”
The previous agency was running the account the way you’d run a B2C ecommerce account — optimize the funnel toward the first conversion, ignore everything downstream. In healthcare that approach acquires patients who never become patients, and the clinical team eats the cost.
What we did
Compliance first. Patient LTV second. Landing page CRO third.
HIPAA and compliance triage
Before any optimization, the legal exposure had to be addressed. This was not negotiable.
- Removed the Meta pixel from all PHI-bearing pages. Patient portal, post-intake pages, and any page that loaded after a logged-in user authenticated had their tracking pixels removed. The growth team had not realized this was happening; the previous agency had treated all pages as equally trackable.
- Server-side GTM via Stape with PHI scrubbing. Implemented a server-side data layer that explicitly stripped PHI before any event was forwarded to Google or any other ad platform. Email and phone were hashed via SHA-256 before transmission. No clinical information, condition data, or identifying patient details ever touched the ad platforms.
- Business Associate Agreements verified. Confirmed BAA coverage with all platforms in the stack that touched any patient data, and removed two tools from the stack that didn’t have proper BAA coverage and weren’t HIPAA-compliant.
- Google Ads healthcare policy remediation. Worked through the 14 Limited Ad Targeting flags, restructured the ad groups where necessary, rewrote ad copy to comply with healthcare ad policy (no specific medical claims, no implied guarantees of outcomes, no targeting based on sensitive categories), and resolved the suspended ad group.
- Account email routing fixed.Multiple stakeholders now get policy alerts so suspensions can’t go unnoticed.
We told the client during week one that the HIPAA exposure was the most important thing in the engagement and that we’d spend the first three weeks on it before touching any optimization work. The growth team had to convince the founders. The founders agreed once we walked them through the actual regulatory exposure. We’ve turned down healthcare engagements where clients wanted us to skip this step.
Patient LTV conversion hierarchy
This is the core differentiator of the engagement. Most healthcare paid search optimizes toward form submissions. The honest move is to optimize toward patient lifetime value, with the conversion event hierarchy reflecting the full clinical funnel.
The new conversion event structure:
| Stage | Event | Assigned Value | Treats Algorithm Signal As | How It Fires |
|---|---|---|---|---|
| 1 | Eligibility form submitted | $25 | Low signal | Pixel fire on form submit page |
| 2 | Eligibility screen passed | $80 | Real clinical fit | Offline conversion import from clinical system |
| 3 | Intake call completed | $180 | Committed intent | Offline conversion import |
| 4 | First paid visit | $420 | Real revenue moment | Offline conversion import from billing system |
| 5 | 30-day retained | $640 | Sticky patient | Offline conversion import from billing system |
| 6 | 90-day retained | $1,180 | High-LTV cohort fit | Offline conversion import from billing system |
On the assigned values:these are not the actual revenue per patient — they’re calibrated weights designed to teach Smart Bidding the shape of a high-LTV patient. The bidding algorithm now seeks queries, ad copy, and audience signals that produce patients who pass eligibility, complete intake, attend visits, and retain past 90 days. Patients who fill out the form and disappear are weighted at $25, which is to say, the algorithm treats them as nearly worthless and stops chasing them.
Offline conversion imports from the clinical and billing systems are the load-bearing piece. Without them, Smart Bidding can’t see downstream patient quality. We implemented the imports via the provider’s clinical operations system (which had a usable export) and the billing system (which required a lightweight custom export to a Zapier flow that pushed to Google Ads’ offline conversion API). No PHI moved through this pipeline — only hashed click IDs, conversion stage, and the assigned conversion value.
Campaign restructure around clinical intent
The account was rebuilt from scratch around clinically meaningful intent segments:
| Campaign | Example Keywords | Landing Page | Bid Strategy |
|---|---|---|---|
| High-intent condition-specific | “[condition] specialist online”, “[condition] telehealth treatment”, “[condition] online doctor” | /care/[condition] | tCPA → value-based bidding |
| Generic telehealth + symptom | “online doctor for [symptom]”, “telehealth [state]”, “virtual doctor visit” | /states/[state] | tCPA → value-based bidding |
| Branded | “[brand] login”, “[brand] reviews”, “[brand] pricing” | Branded LP | Manual CPC, capped |
| Competitor conquest | “[competitor] alternative”, “[competitor] review” | /compare/[competitor] | tCPA (high) |
- Exact and phrase match for the first 90 days. Broad match added only to campaigns that had accumulated reliable offline conversion volume and demonstrated consistent patient quality.
- Responsive Search Ads written per condition and per state. Headlines mirroring the clinical intent (“[Condition] Specialist Online in [State]”) with compliance-cleared language. Each RSA reviewed against Google’s healthcare ad policy before launch.
- Negative keyword lists organized by disqualifying intent. Job seekers, students, DIY symptom checkers, insurance-only queries, and sub-eligible-condition queries all suppressed at the campaign level.
The previous agency had run 1,200+ keywords across the account. We launched with 380, all tightly themed, and grew the keyword set selectively as offline conversion data validated which themes produced retained patients.
Landing page CRO program
The second distinguishing piece of the engagement. Paid healthcare without landing page CRO leaves most of the available conversion lift on the table. The 11-minute intake form alone was destroying the funnel.
- Biweekly A/B tests. Structured experimentation cadence: one test launched every two weeks, with clear hypotheses, minimum sample sizes, and statistical significance thresholds before calling a winner.
- Form length reduction. The intake form was reorganized into a multi-step progressive flow with conditional logic. Required fields at step 1 dropped to 4 (name, email, state, primary concern). Eligibility screening became step 2. Clinical intake became step 3 after eligibility was confirmed, with auto-save so users could complete it asynchronously. Total fields filled by a converting patient went from 23 to a comparable count, but completion rate moved from 62% to 84% because abandonment now happened cleanly after eligibility was confirmed, which is the right time to lose unqualified patients anyway.
- Trust signal additions. Provider credentials, state licensing information, patient testimonials (HIPAA-cleared), and insurance/pricing transparency added above the fold on condition-specific landing pages.
- Eligibility pre-screen on the page itself. Above-the-fold eligibility check (state, age, primary condition) that filtered out non-eligible patients before they wasted a form fill. Reduced raw lead volume by roughly 22% in the first month but improved eligibility-screen-pass rate by 70%, which is exactly the trade we wanted.
- Mobile-first redesign. 72% of traffic was mobile. The previous landing pages had been designed desktop-first with responsive afterthoughts. The redesign was mobile-first with desktop as the expansion, not the other way around.
Twelve A/B tests shipped over months 2–6. Eight produced statistically significant winners. Three were inconclusive (kept the control). One was a clear loser (reverted within 48 hours). The cumulative effect on landing-page-to-eligible-patient conversion rate: +47% over the baseline.
Value-based bidding maturation
Once 90 days of offline conversion data had flowed reliably, the bidding strategy matured:
- Max Conversions → tCPA.Initial campaigns launched on Max Conversions with the new conversion event hierarchy. After 60 days of stable offline conversion flow, migrated to tCPA targeting cost per “eligibility-screen-passed” event.
- tCPA → tROAS. By month 4, enough downstream conversion data had accumulated to shift the highest-volume campaigns to tROAS (target return on ad spend) using the weighted conversion values. The algorithm now optimized directly for the value mix, not just the cost target.
- Quarterly value recalibration. The assigned conversion values were reviewed and recalibrated quarterly based on actual patient LTV data from the billing system. The initial weights were educated guesses; the quarterly recalibration turns them into empirical reflections of real patient cohort economics.
- Audience signal layering.First-party data (existing patient lists, hashed and PHI-scrubbed) used as similar-audience seed lists and as exclusion lists to avoid bidding on existing patients. Demographic and geographic signals layered to prioritize states with the provider’s strongest clinical capacity.
Brand defense and scale
With the foundation stable:
- Brand campaign tightened. Branded search restructured with competitor-name exclusions, manual CPC with a hard ceiling, and explicit bid-only-when-competitors-are-present rules. Brand spend dropped 35% with no impact on branded conversion volume.
- Microsoft Ads launched.Mirror campaigns launched on Microsoft Ads targeting the same intent segments. Microsoft CPCs ran 30–40% lower than Google for comparable intent queries, and the patient quality was comparable once the offline conversion loop was established. Microsoft now drives roughly 15% of total paid patient volume.
- Scaling discipline. Spend increases only approved when offline conversion data confirmed the incremental spend was producing patients at or below the target cost per acquired patient. No scaling into campaigns without 90+ days of stable offline conversion history.
The math
| Metric | Baseline | Month 7 | Δ |
|---|---|---|---|
| Google + Microsoft Ads spend | $74K/mo | $82K/mo | +11% |
| Raw form submissions | 456/mo | 520/mo | +14% |
| Cost per raw form submission | $162 | $158 | −2% |
| Eligibility-screen-pass rate | 41% | 68% | +27 pts |
| Eligibility-screen-passed patients | 187/mo | 354/mo | +89% |
| First-paid-visit rate (of eligible) | 52% | 64% | +12 pts |
| Acquired patients (first paid visit) | 97/mo | 226/mo | +133% |
| Cost per acquired patient | $763 | $363 | −29% headline |
| 90-day retention rate | 38% | 51% | +13 pts |
| Projected 12-month patient revenue (paid search) | $2.4M | $5.3M | 2.2x |
On the −29% headline.This is the number we use publicly because it’s the most conservative honest interpretation. The full math gives a much larger improvement (cost per acquired patient closer to −52% on first-paid-visit basis), but that number includes structural effects from the eligibility pre-screen filtering that we want to credit to landing page CRO rather than to bidding. Holding the headline at −29% lets us defend the number against any auditor and lets the upside numbers (retention, projected revenue, 12-month revenue lift) carry the rest of the story.
On the eligibility-screen-pass rate jump. This is where the work compounds. Better targeting via patient LTV bidding produces more eligible-fit patients. Eligibility pre-screen on the landing page filters out the rest before they consume a form fill. Together, the pass rate moves from 41% to 68% — patients who fill out the form are dramatically more likely to be actual clinical fits, which means clinical operations spends less time on dead-end intakes and more time on patients who become real revenue.
On the retention improvement.Bidding for patient LTV produces patients who look more like the existing high-LTV cohort. That’s the structural insight: Smart Bidding finds patients in the shape of the signal you give it. Give it form-fill signal and you get form-fillers. Give it 90-day retention signal and you get patients who retain. The 13-point retention lift is not a clinical operations improvement; it’s an acquisition quality improvement that the clinical team didn’t have to do anything to earn.
On the projected 12-month patient revenue. This is a forward-looking number based on actual realized retention curves on month 7 patients applied across the full acquired cohort. We mark it explicitly as projected and we share the assumptions in any pitch follow-up. Healthcare CFOs care about projected lifetime revenue more than raw acquired-patient counts, so we lead with it; honest sophisticated readers will check the math.
What we’d flag to anyone reading this
This worked because four things were already true. They aren’t always.
The clinical product was strong and patients retained when they got proper care.
Acquisition optimization can find more high-LTV-shaped patients, but if the underlying clinical service is mediocre, the retention signal will be flat and Smart Bidding has nothing to chase. We’ve turned down healthcare engagements where the clinical retention numbers told us the product wasn’t ready for scaled acquisition.
The clinical and billing systems had usable conversion data.
Offline conversion imports require an extractable record of which patients hit which milestones. Some healthcare providers run on EMR systems that lock down clinical data so tightly that even compliant export pipelines are impractical. The provider’s stack allowed for HIPAA-compliant conversion exports, which is the prerequisite for value-based bidding. Without it, we can run the campaign architecture work and the landing page CRO, but the bidding stays at “eligibility-screen-passed” rather than full patient LTV.
The growth and clinical teams were willing to talk to each other.
A meaningful share of the engagement was structured collaboration between growth (which owns CPA) and clinical (which owns patient quality and retention). Some healthcare providers run growth and clinical as siloed organizations that don’t share data or strategy. That structure doesn’t work with this playbook. We have this conversation in discovery and we surface it as a prerequisite, not a nice-to-have.
HIPAA exposure was acknowledged as a real risk, not a compliance checkbox.
Some prospects want us to skip the compliance triage step because it doesn’t produce headline numbers in month one. Those prospects don’t become clients. The legal exposure of running healthcare paid acquisition without proper PHI hygiene is real and we won’t sign engagements where the client tries to negotiate it away.
And a fifth thing worth naming: healthcare paid search has a slower performance ramp than non-regulated categories.
The offline conversion learning loop takes 90–120 days to mature because the patient journey from click to first-paid-visit averages 18–25 days, and 90-day retention data takes 90 days to accumulate. Operators who want week-to-week conversion lift won’t get it from this playbook. The compounding effect arrives in months 4–7 and accelerates from there. Healthcare clients who want to see the model perform need to commit to a 6-month minimum; faster timelines aren’t honest.
Engagement details
- Team on the account
- 1 strategist, 1 paid search specialist (healthcare-experienced), 1 CRO specialist, 1 landing page designer, 0.25 compliance / privacy reviewer
- Tool stack
- Google Ads, Microsoft Ads, Stape (server-side GTM with PHI scrubbing), Unbounce (compliance-cleared landing pages with HIPAA-aware tracking), client clinical operations system, client billing system, GA4 (configured with PHI-safe parameters), Looker Studio
- Reporting cadence
- Weekly campaign and CRO test review, monthly executive performance review with clinical and growth leadership, quarterly QBR including conversion value recalibration
- Contract structure
- 6-month minimum (healthcare requires this for the learning loop to mature), month-to-month after that
Ready for the same teardown on your patient acquisition program?
We’ll audit your Google Ads, your landing pages, your HIPAA-compliance posture, and your conversion measurement, and show you where qualified patient revenue is leaking before you commit to anything.