Use cases
Whatever you're building, the data was the blocker. Not anymore.
Consultants who can't benchmark a population. Startups who can't demo without PHI. Data teams who can't backtest a model. They're all stuck on the same thing — access to healthcare data. Verism Health builds synthetic patient archetypes and renders their whole data trail — claims, eligibility, labs, quality, revenue — so the work starts in a minute instead of after a six-month data deal. Medicare Advantage is live today, with every line of business close behind.
Win the work before you have the data deal.
You're asked to benchmark a population, pressure-test a risk-adjustment or Stars approach, or model a value-based contract — across whatever lines of business you work with, often for a prospect, often before a contract is signed. The methodology is the easy part. The blocker is that you can't point it at any data. Medicare Advantage is the panel you can pull today; the rest of the book is on the way.
A national claims license runs $50k–$250k a yearand takes months of legal review. You can't expense that to a pitch you haven't won, and the license terms almost never permit speculative or demo work anyway.
So benchmarking decks get built from stale public PDFs, hand-wavy assumptions, or a single prior engagement's data you're not really allowed to reuse. Every new pitch starts from a cold, dataless start.
Download a calibrated panel and stand up real benchmarks in an afternoon — risk scores, MLR, PMPM, utilization per 1,000, dual vs. non-dual splits — all tied to published targets, with claims, eligibility, and revenue joined to the same members. Today that panel is Medicare Advantage, calibrated to CMS and MedPAC.
Build the deliverable, test the methodology end to end, and walk into the pitch with working numbers. It's synthetic and licensed for exactly this, so there's nothing to clear with legal first.
Real-claims licenses are slow, expensive, and contractually can't be used for spec work, demos, or anything outside a named engagement — exactly the moments where you most need data to win.
And generic "synthetic" or open datasets aren't calibrated to real benchmarks: the risk scores, payment factors, and utilization curves don't hold up the moment an actuary checks them against published targets. Verism is built to survive that check — every release ships a credibility audit with citations.
A boutique actuarial firm is pitching a regional MA plan on a risk-adjustment audit. With a 100k-member Verism panel, they reproduce the plan's likely V24-to-V28 risk-score shift, quantify the revenue impact, and bring a quantified "here's what we'd find" story to the first meeting — for the cost of a starter panel, not a license.
Build and demo the product before the real-data DUA closes.
You're building a risk tool, a care-management platform, or an analytics dashboard. To make it real, it has to run on healthcare data — claims, eligibility, maybe labs or quality — but your first customer's data is locked behind a DUA that won't close for months, and you can't demo, develop, or sell on data you don't have.
Getting PHI for development means HIPAA, BAAs, and a security review — months of work before a single engineer touches a row. Until then, the roadmap stalls and the demo is a slide.
Teams fall back on hand-built mock data, but it doesn't survive contact with a clinician or an actuary: the codes don't co-occur, the costs don't add up, and the credibility of the whole demo collapses in one question.
Develop and demo on realistic, linked healthcare data with zero PHI exposure — no BAA, no security review, no privacy risk. Claims, eligibility, and revenue render from the same members, so the whole product wires up to a real schema today. Medicare Advantage ships now; more lines of business are close behind.
Because the data is calibrated and clinically coherent, the demo holds up when a prospect's clinical or actuarial team pokes at it. When the customer's DUA finally closes, you swap synthetic for real — the schema is the same, so your code doesn't change.
Acquiring real PHI just to build is a multi-month compliance project — HIPAA, BAAs, vendor security review — that most early teams can't afford to start, let alone finish, before they need to ship.
And mock or toy data doesn't survive a clinician's scrutiny. Verism gives you the realism of real healthcare data without the exposure, on a schema that mirrors what production data looks like — so the swap to live data is a config change, not a rebuild.
Zero PHI, day one
Every row is synthesized — no real patient is ever represented — so there's nothing to de-identify, no BAA to sign, and no review board to clear before you start building.
Demo-grade realism
Eligibility, medical, Rx, and revenue all render from one coherent set of member journeys, so a diabetic-with-CHF member looks like one — across files and across months — when a buyer drills in. Labs and quality measures render from the same people next.
Same schema, real later
Build against the production schema now and swap in the customer's real data when the DUA closes. Your ingestion, models, and dashboards carry over untouched.
A care-management startup is six weeks from a design-partner demo, but the health-plan's data won't arrive for a quarter. They build their risk-stratification and outreach workflows on a Verism panel, demo a working product on realistic members, close the partner — then point the exact same pipeline at the real feed once the DUA lands.
Train and backtest with ground truth real data never gives you.
You're training a risk-adjustment or cost-prediction model, validating a pipeline, or stress-testing infrastructure. Real healthcare data fights you at every step: no clean ground truth, locked behind access controls, and impossible to regenerate when you want to change one variable.
Real data has no labels you can trust. You never know a member's "true" risk or expected cost — only the messy realized outcome — so you can't cleanly separate model error from data noise.
It also can't be shared freely across a team, and it can't be regenerated. Want to test how your model behaves when coding intensity rises 5%? With real data, there's no knob to turn.
Because Verism panels are generated, they come with the generative ground truth behind each member — the underlying conditions and journey that produced the claims — so you can validate against what the model should have found, not just what happened.
The dataset is reproducible and freely shareable across your whole team (no PHI, no per-seat access controls), and you can regenerate it with controlled perturbations to backtest behavior under conditions you choose.
Real healthcare data has no clean ground truth — you only see realized outcomes, so true risk and expected cost are forever unobserved and your model evaluation inherits the noise.
Real data also can't be passed around a team without access controls, and it can't be regenerated with a controlled change — so reproducible experiments and deliberate stress tests are off the table. A generative source removes all three limits at once.
What you can do that real data won't allow
- Train and validate risk-adjustment and cost-prediction models against the member's underlying ground truth, not just realized spend.
- Backtest a model under controlled perturbations — shift coding intensity, prevalence, or utilization and regenerate to see how predictions move.
- Stress-test pipelines and infrastructure at 1M-member, multi-file, line-level scale before you ever touch production data.
- Share one canonical, reproducible dataset across the whole team — no DUA, no per-seat access review, identical inputs for everyone.
- Build reproducible benchmarks and regression tests that you can rerun release over release because the data regenerates deterministically.
A payer analytics team is rebuilding their cost-prediction model. They train on a 1M-member Verism panel, evaluate against the known generative ground truth to isolate model error, then regenerate the panel with elevated coding intensity to confirm the model degrades gracefully — all before the new model sees a single real member.
Calibrated to published benchmarks today — Medicare Advantage across eligibility, medical, Rx, and revenue; pattern-learning from licensed real data is on the roadmap. See the methodology →
Where we're going
RoadmapMedicare Advantage is the first line, not the last.
Today, Verism Health renders Medicare Advantage across four domains — eligibility, medical, Rx, and revenue — and that's what's live right now. But the engine builds a synthetic person and renders their whole data trail; it isn't tied to one line of business or one domain. Here's where it's headed. We label this honestly: the lines and domains marked below are expanding, not available today.
Lines of business
Medicare Advantage today · the rest expandingMedicare Advantage
Available nowOur first line, shipping today: HCC risk adjustment, MMR revenue, Part D, dual/LIS dynamics.
Medicare FFS
ExpandingTraditional fee-for-service Medicare — the benchmark population behind most ACO and VBC work.
Commercial / employer
ExpandingWorking-age commercial populations — different age mix, benefit design, and cost curve.
Medicaid
ExpandingManaged Medicaid with its own eligibility churn, demographics, and program structure.
ACA / exchange
ExpandingMarketplace populations with HHS-HCC risk adjustment and metal-tier benefit design.
Data domains
Four live · encounters, labs & quality expandingEligibility & enrollment
Available nowMember-month enrollment, demographics, plan/program, and benefit status — the spine every other domain links to.
Medical claims
Available nowLine-level institutional + professional claims: diagnoses, procedures, settings, allowed/paid, and realistic adjustment chains.
Pharmacy
Available nowNCPDP-grade drug fills with refill chains, benefit phases, formulary tiers, and net-of-rebate economics.
Revenue & payment
Available nowPayer-side revenue the way a plan receives it — capitation, risk scores, and the factors behind every dollar.
Encounters
ExpandingEncounter-level utilization independent of billing — the visit-and-service record health systems and value-based programs run on.
Labs & results
ExpandingOrdered tests with realistic result values trended to each member's conditions — A1c that tracks the diabetic, eGFR that tracks the CKD.
Quality measures
ExpandingMeasure-ready numerators, denominators, and gaps (HEDIS-style / Stars) rendered from each member's actual care.
We're building a synthetic-data platform for healthcare, but we'd rather be precise than grand: Medicare Advantage and four domains are what ship today, and every dataset is calibrated to published benchmarks— not trained on real row-level data yet (that's the v3 roadmap). If a line of business, domain, or panel you need is on this list, tell us — roadmap order follows demand.
Put it against your own use case.
Download the 1,000-member sample — full schema, full report bundle, no signup, no card. Benchmark it, demo on it, or train against it, and judge the fidelity yourself.