Health plans run their marketing on Cost Per Enrollment. The business runs on member-years. That gap is where the budget gets spent on the wrong audiences, the wrong creative, and the wrong channels — every AEP, every OEP, every cycle.
Pick any Medicare Advantage plan in the country and look at the marketing dashboard. The hero metric is some version of Cost Per Enrollment. It might be called CPE, CPA, cost per app, cost per member acquired — same metric in different packaging. Total marketing spend, divided by total enrollments. The lower the number, the more the team gets praised.
Now look at the actuarial model the same plan uses to price its bids. Member-years. Risk-adjusted contribution margin per member-year. Expected tenure by acquisition cohort. Lifetime value calculated against the medical loss ratio, plan-level admin cost, and the share of CMS quality bonus revenue an enrollee unlocks. None of those numbers appear anywhere on the marketing dashboard. The marketing organization is optimizing toward a metric that is not connected to how the plan actually makes money.
This is not a small reporting nuance. It is the central reason MA marketing budgets get misallocated, why broker channels keep getting more rope than they should, why the cheap enrollments in October look like a win in November and a problem by April, and why Performance Max is pointed at the wrong audience for half the budget. The marketing KPI and the business KPI are not the same metric, and almost no plan has done the work to align them.
What Cost Per Enrollment actually optimizes for
Cost Per Enrollment treats every enrollee as identical at the moment of enrollment. A member who switches off the plan in month four is worth the same in CPE as a member who stays for six years. A 67-year-old healthy enrollee from a high-tenure ZIP code is worth the same as a 64-year-old age-in from a high-churn channel with three complex chronic conditions and a history of plan-switching.
Google's AI does not know that. The conversion signal you feed it does not contain that information. The platform — running on Smart Bidding — optimizes toward the cheapest, fastest path to whatever conversion you defined. Define enrollment, get enrollment. Define cheap enrollment, get cheap enrollment. The platform is doing exactly what it is told. It is the metric that is wrong, not the algorithm.
The downstream effect is predictable. The audiences that are cheapest to convert are not the audiences that produce the highest member-year value. The channels that report the strongest CPE are often the channels with the worst tenure economics. The creative that wins on CPE is the creative that pulls in the lowest-friction, lowest-commitment shopper — exactly the shopper most likely to disenroll the next AEP. The marketing team hits its number. The plan loses money on the cohort. Nobody connects the dots because the two metrics live in different systems.
Last-click attribution had this same shape, just for a different industry. The metric is cheap to measure, easy to defend, and quietly destructive to the business it claims to serve. Health plans are about a decade behind retail and DTC on figuring this out. Meridian MMM exists in part because retail finally got tired of paying for the Search-attribution illusion. MA marketing is still in the illusion.
The metric that actually runs the business
Cost Per Member-Year is the marketing investment required to acquire one expected year of enrolled membership, weighted by the contribution margin that member-year is expected to produce. Strip out the formal-sounding language and it is a simple correction: stop pretending every enrollment is worth the same, and start pricing each enrollment against what the plan actually expects to earn from it.
Three components matter. Predicted tenure — how many years this enrollee, given their acquisition source and demographic profile, is statistically likely to stay. Predicted contribution margin per year — what this member-year is worth net of medical and admin cost. And the marketing investment required to acquire that enrollee in the first place.
The math is not complicated. The data is. Most plans have all three components sitting in different systems, owned by different teams, scored on different cadences, and never joined into a single record at the level marketing actually needs to act on it. Acquisition source rarely makes it cleanly into the actuarial model. Tenure prediction rarely makes it into the marketing platform's conversion signal. The plan knows what a member-year is worth in aggregate, but cannot tell you what it is worth by audience, channel, or campaign.
That is the actual work. Not building a new metric — building the data plumbing that lets the metric exist at the granularity Performance Max can act on.
What changes when you optimize on CPMY
Run the same media plan with both metrics applied to the same conversion data and the budget allocations shift in three predictable ways.
Audience selection moves up the value curve
Cohorts that look expensive on CPE often look efficient on CPMY because their tenure is long enough to absorb the higher acquisition cost. The classic example is the brand-loyal age-in: more expensive to convert in October, but six-year tenure with low churn risk. CPE penalizes you for chasing them. CPMY rewards you for it.
Channel mix corrects the broker overweight
Most plans run direct channels and broker channels with two different math models — direct on CPE, broker on a fixed commission per enrollment. The two are not comparable, and the comparison usually flatters whichever channel is being defended at the moment. Restated in CPMY, both channels become commensurable. Plans that have done this work find that direct-to-consumer Performance Max often produces a better member-year value than the broker mix, but the comparison was hidden by metric mismatch. The 2024–2025 MA Final Rule restrictions on TPMO marketing make this rebalance unavoidable; CPMY is the metric that lets you do it cleanly rather than reactively.
Creative shifts from urgency to fit
CPE rewards creative that maximizes conversion velocity. Fast pitches, hard urgency, simple benefit claims. The shopper most responsive to that creative is also the shopper most likely to disenroll once they encounter the actual plan. CPMY rewards creative that converts the right shopper, which is almost always the shopper who took longer to decide and arrived with clearer intent. The creative gets quieter, the member-year math gets better.
"The shopper most responsive to urgency is also the shopper most likely to disenroll. That is not a creative problem. It is a metric problem."
Cost Per Enrollment vs. Cost Per Member-Year, side by side
How to feed CPMY into Performance Max
The technical path is shorter than most teams expect. Performance Max already supports value-based bidding through custom conversion values. The piece most plans are missing is the model that produces a defensible per-enrollment value at the moment of conversion.
The shape of that model is straightforward. Build a tenure prediction by acquisition source, plan type, and demographic cohort using internal enrollment history. Build a margin-per-member-year estimate from the actuarial model, segmented at the same granularity. Multiply tenure by margin to produce an expected member-year value per cohort. Pass that value to Performance Max as the conversion value at the moment the enrollment fires. Google's AI will do the rest — bidding higher on cohorts with higher predicted member-year value and lower on cohorts whose tenure economics do not justify the spend.
This is the same value-based bidding pattern that retail and DTC have been running for years. Health plans have been late to it because the data infrastructure to predict member-year value at the cohort level was harder to build than retail's lifetime-value models. The infrastructure is now buildable. The Google ecosystem is set up for it. The constraint is no longer the platform — it is the willingness to do the data work.
Where compliance actually sits
CPMY is an internal economic metric. It does not appear in consumer-facing creative, member communications, or any surface the Medicare Communications and Marketing Guidelines (MCMG) regulate. The compliance work happens upstream — in making sure the audience modeling that feeds the conversion value does not use prohibited targeting attributes, that the enrollment data flowing into the value calculation is handled under the plan's existing CMS-compliant data architecture, and that any campaign flighting tied to CPMY thresholds respects MCMG-restricted dates and content rules.
Properly built, CPMY is more conservative than CPE on compliance, not less — because it forces the plan to think clearly about which audience signals are being used to predict member value and which are not. The plans that build this metric properly end up with a cleaner audit trail than the plans that did not, which is a useful side effect to mention to the legal and compliance teams whose sign-off you will need.
What this looks like on the P&L
The financial argument for CPMY is the same shape as every argument on this site. The marketing decision and the business decision become the same decision.
A plan running on CPE allocates budget to whatever produces enrollments cheapest in October. The cohort comes in. Some fraction of it churns by Q1. The member-year economics, calculated retroactively in the spring, look worse than the marketing reports suggested. The plan adjusts next year's plan design or premium to compensate. The marketing team gets credit for the original CPE number. Nobody is wrong. The business loses money anyway.
A plan running on CPMY allocates budget against expected member-year value at the point of decision. The cohort comes in already weighted toward longer tenure and better margin. The retroactive reconciliation in Q1 still happens, but the variance is smaller, and the variance that remains is information rather than damage — feedback that improves next year's tenure model rather than evidence that this year's marketing was misallocated. The same marketing dollars produce more member-years, more risk-adjusted margin, and a higher quality bonus contribution. Cost Per Decision drops because the marketing system is making the right decision the first time, instead of making the wrong decision quickly and correcting later. Consumer FinTech has the same shape of metric problem and so does retail — the lesson generalizes across any business where the customer's economic value is decided after the marketing event fires.
That is the structural margin lever. It is not a creative innovation, not a channel shift, not an AI tool. It is a metric correction, applied honestly, with the data infrastructure to support it. Every plan can do this. Almost none have.
This piece is one vertical instance of a broader argument — every marketing KPI in common use measures activity, not decisions, and agentic AI is the first technology that makes the correction operable rather than aspirational. Health insurance is just where the cost of getting it wrong shows up most cleanly.
The marketing team is not optimizing wrong. The metric is. CPE was the best a health plan could do without the data to predict tenure and margin at the cohort level. That data is now available. The plans that move first will spend the next two AEP cycles compounding a structural cost advantage the rest of the market is not yet measuring.
Sources & further reading
- About Performance Max campaigns — Google Ads Help. The campaign type plans now run their AEP and age-in budgets through.
- About Smart Bidding — Google Ads Help. The auction-time bidding system that optimizes toward whichever conversion signal the marketing team defines.
- Value-based bidding best practices — Google Ads Help. The mechanic for moving the platform's optimization from "cheapest enrollment" to "highest predicted member-year value."
- About conversion value rules — Google Ads Help. How to encode acquisition-source-specific tenure and margin into the bidding signal.
- About customer lifecycle goals — Google Ads Help. The framework for prioritizing retention and lifetime-value events over top-of-funnel ones.
- About data-driven attribution — Google Ads Help. The replacement for last-click attribution that uses account-level data to credit conversions across surfaces.
- CLV modeling for customer retention — Think with Google. The lifetime-value modeling pattern health plans are now late to apply at the cohort level.