Frequently Asked Questions

SPARK Predict works across Medicare Advantage, Commercial, Medicaid, ACO/MSSP, and ACA Exchange. More importantly, it understands the distinct economics, quality measures, and risk adjustment logic of each — so analysis reflects how each business actually operates.

The platform supports CMS-HCC, HHS-HCC, RxHCC, ESRD, and related models, and adapts to the specific program context — including nuances like state-level variation where the data supports it.

We train machine learning models on large longitudinal datasets (~1.9M members, 5 years, ~40M claims) using techniques like gradient-boosted trees. Models are validated on independent holdout data before being applied to your population to generate forecasts and surface emerging risks.

We also use traditional statistical methods (regression, time series, survival analysis) where appropriate. Our Supervisor agent validates model assumptions and discloses methodology in each response.

Our predictive models are trained on large internal datasets (40M+ records) and applied to your data to generate population-specific insights. We do not use your data to train our models or support analyses for other clients. For customers interested in custom model development on their own data, we offer that as an advanced capability — always opt-in and governed by clear data use agreements.

Metric definitions are configurable and transparent, aligned to your internal standards. For ambiguous metrics, our Supervisor agent asks clarifying questions before generating results.

Yes. Users can define logic, cohorts, and metric calculations directly through natural language.

SPARK Predict works alongside your existing tools, but represents a fundamentally different way of interacting with your data. Rather than building dashboards or submitting requests, users ask questions in natural language and get answers with full methodology. Over time, teams often shift analytical work into Spark Predict or build integrations with their existing tools.

Yes. Results, data tables, and visualizations can be exported in multiple formats (CSV, Excel, PDF, PNG) for use in emails, presentations, or reports. We're also building AI-generated presentations, summaries, and draft communications.

For initial deployment, integration is not required — we start with data files via SFTP to minimize IT lift. As the partnership evolves, we can integrate with your data warehouse, cloud storage, or analytics platforms.

No. Your data is yours. We do not use your data to train models that benefit other clients unless you explicitly opt in. All data use is governed by clear agreements that specify exactly how your data can and cannot be used.

Only users within your organization with appropriate role-based permissions. We do not access your data or monitor your activity except to provide maintenance and support.

Each customer environment is fully isolated within our multi-tenant architecture. There is no cross-customer data access.

We operate under HIPAA-compliant infrastructure and execute a BAA with every customer. We can use your standard BAA or provide ours. PHI is protected through encryption, access controls, and audit logging.

Your data is stored in a secure, HIPAA-compliant cloud environment with encryption at rest and in transit. Access is governed by strict role-based controls and monitored through comprehensive audit logging. We are SOC 2 Type II certified and conduct regular security assessments to continuously validate our security posture.

Upon termination, we will securely delete all of your data from our systems within a defined period, and provide written confirmation of deletion. If you need a data export before termination, we can provide that. Retention and deletion terms are specified in the contract and BAA.

Every analysis goes through our validation system: the Analyst agent generates results, the Supervisor agent reviews methodology and checks for errors, and both are governed by quality guardrails.

Yes. For any result, you can view the underlying logic, assumptions, and methodology used to generate it.

The system recognizes uncertainty. When data is insufficient or a question is ambiguous, it says so, explains what's missing, and suggests next steps rather than guessing.

SPARK Predict is a decision-support tool — it informs your team's judgment, it does not make decisions on its own. Every response includes the methodology and assumptions used, so your team can evaluate the output before acting on it. Our Supervisor agent catches errors, flags uncertainty, and discloses limitations. When the system is unsure, it says so. That said, no analytical tool is infallible, and we encourage users to apply their domain expertise when interpreting results — just as they would with any analyst's work product.

Yes. You can review the methods, assumptions, and data sources behind results for audit and compliance.

Minimal IT lift means you don't need backend integrations, API development, or database connections to get started. What you do need: someone who can export your claims data files and upload them via SFTP. That's it for the technical side.

Typically 1–2 weeks from data receipt to first insights. Week 1 is data upload and processing; week 2 focuses on validation, onboarding, and running your first queries. More complex setups may take up to 4 weeks.

We start with a kickoff to align on goals and key metrics, followed by data intake and validation. We then set up users and access, provide a brief training session, and support your team through initial use.

The trial grants full platform access to a limited number of users. Your team can run real analyses across cost, risk, quality, and utilization. The goal is to answer your actual business questions, compare outputs to your current workflows, and validate value.

We can start with as little as one quarter of data and expand from there. For the full range of analytics, 12 months of recent data is ideal.

Trial terms vary depending on the engagement. We'll discuss pricing and structure during our initial conversations to find an arrangement that works for both sides.

If the trial is successful, we move into full deployment with expanded access across your organization. This includes onboarding additional users, broadening data coverage, and aligning on ongoing workflows. There is no reimplementation — your existing environment and analyses carry forward.

SPARK Predict acts as an accelerator for your analytics team. It handles ad hoc questions, reduces backlog, and allows analysts to focus on more strategic work such as deeper investigations, custom modeling, and advising the business.

Minimal. The interface is conversational, so most users are productive in their first session. We provide a short intro session to cover key use cases.

No practical limit for typical enterprise usage. Access is managed through role-based permissions, so different teams — executives, analysts, care managers, operations — can work simultaneously with appropriate data controls.

We provide direct access to our team, regular check-ins, and responsive troubleshooting. Early on, we stay closely engaged to support adoption. Over time, support becomes more on-demand.

Updates deploy continuously with no disruption to your environment. We communicate clearly when changes are relevant to your workflows.

We are focused on expanding three core areas: (1) Intelligence Hub, which delivers proactive, personalized insights based on each user and enables more effective monitoring; (2) Analyst Mode, which enhances transparency and collaboration by making methodology more accessible to analytics teams; and (3) Member Action Center, allowing care managers and operational teams to quickly act on insights and drive real-world impact on a member-by-member basis.

Customer feedback directly shapes our roadmap. We incorporate input from active users, prioritize high-impact use cases, and iterate with customers to refine features before broader release. During the trial, we schedule monthly check-ins to review your experience and gather feedback.

Pricing is typically structured at the organization level, based on scope of use, data coverage, and number of users. This allows broad access across teams without limiting usage.

Spark Predict reduces time to insight from days or weeks to seconds, which accelerates decision-making and reduces reliance on manual analysis. Customers see value through faster iteration, improved decision quality, and more efficient use of analyst time.