Why is my CAC going up every month even though nothing else changed?
On July 1, 2026, a new $50 Medicare access path went live for eligible Part D members through the CMS Medicare GLP-1 Bridge. Your number can rise even when your team changed nothing because eligibility, patient demand, appointment supply, checkout friction, or the way results are counted changed around you; find the cause by locating the first rate that broke, not by treating the final ratio as a diagnosis.
Customer acquisition cost, or CAC, is the amount spent to add one new patient under a fixed definition. The formula is simple: eligible acquisition spend divided by new patients. The diagnosis is not. A change in either side of that fraction can produce the same higher CAC.
A higher CAC is a finding, not a cause
Four different problems can end at the same number:
- The price of reaching the same audience rose.
- Fewer people remained eligible or able to complete care.
- The business counted fewer of the people who did complete care.
- The path from first interest to completed first visit became less efficient.
Those causes belong to different owners. Auction pressure goes to the acquisition team. State coverage and appointment supply go to operations. A definition or event break goes to analytics and compliance. A mismatch between the promise, page, and intake path goes to the growth team. Changing ads before locating the first broken rate can turn one problem into two.
The cleanest diagnostic uses one cohort, one cost definition, and one new-patient event. If one month counts booked visits while the next counts completed visits, the comparison is already invalid. If one report includes agency fees and another includes only platform spend, the numerator changed even if the media budget did not.
First suspect: the market changed around you
An untouched account still enters a live auction. More buyers can compete for the same impression, a large platform can introduce a new offer, or a policy change can alter who is shopping and what price they expect.
The IAB/PwC Internet Advertising Revenue Report for 2025 reported $162.4 billion in programmatic revenue, up $27.6 billion, or 20.5%, from the prior year. The prior-year growth rate was 18.0%. Those figures are not a telehealth CAC benchmark. They are evidence that automated buying activity grew sharply while an individual operator's settings could remain still.
Check whether the first movement was in the cost to reach or bring someone to the site. If that cost rose across several campaigns in the same geography while later rates held, external auction pressure becomes more plausible. If only one audience, creative, or state moved, a market-wide explanation weakens.
Now check the eligible population. The Medicare GLP-1 Bridge began July 1, 2026, carries a $50 copay for eligible beneficiaries, and is scheduled to run through December 31, 2027. That single program changed an access and payment path on a known date. A telehealth operator serving that category should split the trend around July 1 and separate Medicare-eligible patients from cash-pay patients. Combining them can make a real demand shift look like an unexplained account failure.
Use the same logic for a clinician leaving a state, a payer changing coverage, or a capacity cap appearing in one region. These are not excuses. They are testable causes with dates and affected cohorts.
Second suspect: the counting changed
A reported CAC rises when the denominator shrinks inside the reporting system, even if the real count of new patients holds. Consent changes, a broken cross-domain handoff, a renamed event, or a different attribution window can all remove credit without removing patients.
Window settings alone can create large reporting differences. Google Analytics documentation on lookback windows says the default for acquisition events such as first_visit is 30 days, while the default for other conversion events is 90 days. If a patient takes longer than the selected window to complete the counted event, the event can remain in operational records while disappearing from the attributed total.
Health data creates another constraint. HHS says its Office for Civil Rights is prioritizing Security Rule compliance in investigations involving online tracking technologies in its tracking-technology guidance. Do not solve a reporting gap by sending more patient data to a vendor. Reconcile aggregated operational totals with privacy-reviewed analytics, and have counsel or the responsible compliance team approve what each tool receives. This is strategic guidance, not legal advice.
The practical test is a three-way comparison by week:
| Count | Source | What a break means |
|---|---|---|
| Completed first visits | Scheduling or clinical operations | Closest business denominator |
| Qualified intake completions | Intake system | Eligibility and form completion |
| Attributed new-patient events | Privacy-reviewed analytics | Measurement and attribution |
If completed first visits stay flat while attributed events fall on one date, repair measurement before changing spend. If both fall together, measurement alone does not explain the rise.
Third suspect: capacity or access reduced the denominator
People can arrive at the same rate and still fail farther downstream. Appointment inventory can tighten. Support replies can slow. A payment flow can reject more transactions. A clinician-coverage gap can remove one state. An intake rule can correctly route more people to in-person care.
Follow one weekly cohort through these steps:
- Site arrival
- Intake start
- Qualified intake completion
- Booking
- Completed first visit
- Appropriate referral or clinical ineligibility
The sixth line matters. An appropriate referral is not a failed acquisition path, but it also should not be counted as a new patient. Keep it visible as its own disposition. Otherwise, the business either punishes sound clinical routing or hides a real fit problem.
Break every step out by state, device, payment route, appointment window, and new versus returning patient. If arrivals and intake starts hold but completed visits fall only where next-day appointments disappeared, the first repair is capacity. A nationwide creative change cannot create appointment slots in two constrained states.
Fourth suspect: the path itself lost efficiency
Only after the external, counting, and operational causes are tested should the acquisition path lead the differential. The audience may have seen the same creative too often. The promise may no longer match the page. More low-intent visitors may be entering the mix. The intake may ask for more commitment than the preceding page earned.
Find the first rate that moved. A lower click rate points toward audience or creative fatigue. Stable clicks with fewer intake starts points toward page-message mismatch or load friction. Stable starts with fewer qualified completions points toward eligibility, form design, or expectation setting. Stable bookings with fewer completed visits points toward scheduling, reminders, payment, or fit.
This is also where a static message can decay without a settings change. Competitors can copy the promise, patient expectations can reset, and a previously distinctive offer can become ordinary. The account stayed still. The comparison set did not.
Ineligible traffic is not an intake-length problem
From Pranay Parikh, MD:
When ineligible intakes rise, the reflex is to add questions. That is the wrong dial. A longer intake sheds the patients who did qualify, and you pay twice: once for the ineligible click, once for the eligible person who quit at question fourteen.
Three jobs, three places, and they are not interchangeable.
The ad can only do self-selection. Google lists health among its sensitive interest categories and bars advertiser-curated audiences for them. You cannot target eligibility even if you want to. The only filter available in the ad is the copy itself: name the disqualifier and the wrong person does not click. Most telehealth ads are written to maximize clicks, which is a very efficient way to buy people you cannot treat.
The intake carries the hard stops, and it carries them first. Only the two or three questions that can actually end the visit go at the top. Contraindication, state, age. A long intake is not the problem. A long intake before you have told someone whether they qualify is the problem.
The system records the reason, not the patient. No clinician in state. No slot inside the window. Clinician-directed referral. Contraindication. Those are written by the flow or the clinician at the point of exit, not asked of the patient in a form. It costs the patient nothing, it adds no length, and it is the only thing that tells you whether an expensive non-conversion was a correct rejection or a service failure you paid for.
Without that last field, an appropriate referral and a patient you lost to a three-week wait look identical in the ad account. Both look like traffic that did not convert.
Run the differential in this order
| First changed signal | Leading explanation | Next check |
|---|---|---|
| Cost to reach people rose; later rates held | Auction or market pressure | Geography, audience, and date of change |
| Attributed events fell; completed visits held | Counting failure | Consent, event firing, domain handoff, and lookback window |
| Arrivals held; qualified intakes or visits fell | Eligibility, access, price, or operations | State, payment route, capacity, and disposition |
| Reach cost held; early response rates fell | Acquisition-path decay | Audience mix, creative repetition, promise, and intake friction |
Then make five decisions:
- Freeze the formula. Write down every cost in the numerator and the exact event in the denominator. Use the same definitions across the comparison period.
- Mark change dates. Add policy launches, price changes, clinician departures, tracking releases, consent changes, and scheduling incidents to the trend line.
- Reconcile reports with operations. Compare attributed events with aggregated qualified intakes, bookings, completed visits, and appropriate referrals.
- Segment before averaging. Split by state, payment route, device, care path, and new-patient definition. A blended average can hide the only segment that moved.
- Repair the earliest break. Fix an event before judging new creative. Restore capacity before adding demand. Clarify the promise before pushing more people into a mismatched intake.
The result should be a cause with a date, a cohort, and an owner. "CAC went up" is none of those.
Get the Marketing Triage diagnostic by email to sort the first break into cost, counting, capacity, or patient follow-through.
If the result points to the acquisition path, use marketing triage to choose the next review without skipping the operational evidence.
Use the Telehealth Growth Score to lock the numerator and denominator before comparing periods. Put that result beside the telehealth growth metrics that locate the first changed step. Before adding budget, check the operational prerequisites in before you run paid ads; the broader readiness checklist before paid ads connects the full patient-acquisition path.
FAQ
Why can CAC rise even when spending has not changed?
CAC can rise when the same spend produces fewer new patients or when reporting captures fewer of the patients actually acquired. Auction pressure, a smaller eligible pool, appointment constraints, intake friction, and measurement changes can each alter the ratio without a budget change.
How can I tell whether higher CAC is a measurement problem?
Compare attributed acquisitions with a privacy-appropriate operational total for new completed visits. If the operational total stays stable while attributed acquisitions fall on a specific date, inspect event definitions, consent behavior, cross-domain handoffs, and lookback windows before changing campaigns.
Should I cut spending as soon as CAC rises?
Not before locating the first changed step. Cutting spend will not repair broken counting, appointment capacity, eligibility friction, or a weaker path from first visit to completed intake.