Where Margin Quietly Leaks in Healthtech: A Unit Economics Teardown

Direct answer: In digital health, margin usually leaks below the topline in four predictable places: care-delivery labor (clinician time per member), payer and reimbursement friction (denials, retro-adjustments, slow collections), tech-plus-support cost that scales with usage instead of price, and churn that shortens the payback window on high acquisition costs. The fastest way to find the leak is a unit economics teardown at the level of one served patient or member per month — not blended company averages, which hide the loss inside a healthy-looking aggregate.

Healthtech margin problems rarely show up as a single big number. They accumulate quietly across a per-unit stack that looks fine until you segment it. This article walks through how to run that teardown, what "good" looks like, and where a tool like Percision (the strategic intelligence platform we build) helps versus where a spreadsheet or a fractional CFO is the better call.

Define the unit before you count anything

The most common Unit Economics error in digital health is choosing the wrong unit. A telehealth company, a chronic-care management platform, and a B2B SaaS-for-providers business all have different atoms.

Pick the unit that matches how you actually earn and spend:

• PMPM businesses (chronic care, virtual-first primary care): the unit is one enrolled member per month.
• Fee-for-service / per-encounter telehealth: the unit is one completed visit.
• B2B SaaS for health systems: the unit is one contracted account, with a secondary lens on per-seat or per-covered-life.
• Marketplaces (labor, referrals): the unit is one matched transaction.

Then build the per-unit P&L:

1. Unit revenue — the realized, not contracted, amount. In healthtech this gap is enormous: net of denials, downcoding, retroactive eligibility terminations, and bad debt.
2. Direct cost to deliver — clinician and coordinator time, pharmacy or device pass-through, lab, and any variable third-party APIs (eligibility checks, e-prescribe, payment processing).
3. Contribution margin = unit revenue − direct cost.
4. Customer acquisition cost (CAC) — fully loaded, including the sales cycle length for enterprise deals.
5. Retention / lifetime — measured as months of continued enrollment or contract, net of involuntary churn (eligibility loss is churn too).
6. Payback period and LTV:CAC.

If you can't fill in line 1 with realized revenue, stop — that's often the leak.

Where the leaks actually hide in digital health

Run each of these questions against your per-unit numbers:

Is realized revenue meaningfully below contracted revenue? Denial rates, downcoding, and retro-eligibility terms silently shave PMPM and per-visit revenue. If your model uses contracted rates, your contribution margin is fiction. Ask: what is our clean-claim rate, and what is our days-in-AR trend?

Does direct cost scale with usage but revenue doesn't? This is the classic PMPM trap. You're paid a flat per-member fee, but high-acuity members consume 5–10x the clinician minutes of low-acuity ones. Blended margin looks fine; the top decile of utilizers is deeply underwater. Segment contribution margin by acuity cohort.

Is "support" being counted as fixed when it's variable? Care coordination, patient onboarding, and clinical support often grow linearly with volume. If they sit in "G&A," you're overstating unit margin.

Is CAC payback longer than retention in your worst cohort? In consumer digital health, a 14-month payback against an 11-month median retention means you lose money on every acquired user in that segment. Enterprise deals hide the same problem inside long implementation timelines that push contribution positive out by quarters.

What does "good" look like? There's no universal benchmark, but directional targets most healthtech boards accept: contribution margin positive at the unit level within the first year of a relationship; LTV:CAC trending toward 3:1 as the model matures; CAC payback inside the observed retention window with a comfortable buffer. The specific numbers matter far less than the slope — is the leak widening or closing as you scale?

Turning the teardown into an execution plan

Finding the leak is analysis. Fixing it is a sequence of decisions: reprice the underwater cohort, tier care intensity to acuity, renegotiate reimbursement terms, cut CAC in the low-retention channel, or narrow the ICP. Each carries a scenario you should model before committing.

This is where a strategic intelligence platform earns its place. Percision — full disclosure, this blog is ours — runs your business context through structured Unit Economics reasoning as part of its 83-step analysis, producing a per-unit contribution breakdown, scenario comparisons (e.g., "what happens to blended margin if we reprice the top-acuity decile 12%?"), and a board-ready deck with an audit trail behind the numbers. It's built as a co-pilot: it surfaces where the leak likely is and models the fixes in minutes rather than weeks, but your clinical and finance leadership makes the call. For CFOs benchmarking rapidly or corp-dev teams pressure-testing a target's margin story, that speed is the point.

When you don't need us: If you have a clean data warehouse and a strong FP&A analyst, a well-built spreadsheet segmented by cohort will find most leaks — and you should build it regardless. If the problem is deeply payer-specific contract law, a healthcare-specialized consultant or actuarial firm is the right hire. And if your leak is operational (a broken eligibility-verification workflow), that's an execution fix, not a strategy analysis. Broadly, AI tools are strongest on well-structured reasoning work; a 2023 study by researchers at Harvard Business School and Boston Consulting Group found consultants using GPT-4 completed tasks faster and at higher quality within the tool's competence — and produced worse results outside it. Match the tool to the task.

FAQ

What's the single most overlooked margin leak in digital health? The gap between contracted and realized reimbursement — denials, downcoding, and retroactive eligibility loss. Models built on contracted rates routinely overstate contribution margin by double digits.

Should we measure unit economics per member or per encounter? Match the unit to how you're paid. PMPM businesses use per-member-per-month; fee-for-service uses per-encounter. Using the wrong unit produces confident, wrong conclusions.

How fast can we get a defensible unit economics teardown? With clean cohort-level data, a strong analyst can build one in days; Percision produces a first-pass, board-ready version in roughly 7–15 minutes for you to refine. Neither replaces leadership judgment on which fix to pursue.

If you want to run a Unit Economics teardown against your own numbers and get a board-ready scenario deck, you can try Percision here. Use it as a co-pilot — bring your clinical and finance context, and keep the decisions with your team.