Beyond BI: How a Healthcare Scaleup Built a Data App Platform That Cut Costs 70% and Shipped 5x More Applications
From one dashboard a week to a self-service data application platform
Tony Zeljkovic
2025-05-22
- Industry: Healthcare & Technology
- Duration: ~3 months
- Team: 1–2 Narona Data consultants
- Stack: Streamlit, Kubernetes, AWS (S3, ElastiCache), OAuth2, React, VS Code devcontainers
- Key Results: 5x more data applications | 90% faster deployment | 70% cost reduction vs. BI platform
Executive Summary
A rapidly growing U.S. healthcare scaleup was shipping one dashboard every one to two weeks — and leadership was questioning whether the data team's spend was justified. Business units needed interactive data applications, not more dashboards, but the team lacked the application engineering skills to build and deploy them. Narona Data delivered a Streamlit-on-Kubernetes platform in three months that enabled 5x more data applications, cut deployment time by 90%, and reduced costs 70% compared to the existing BI stack.
Situation
A high-growth healthcare company had invested heavily in a modern data stack — cloud data warehouse, BI platform, growing analytics team. The infrastructure was solid and the team was technically capable.
But the data team's output was measured in dashboards, and dashboards alone weren't generating the returns leadership expected. Business units across the organization needed data-driven automations and interactive tools — things that couldn't be built inside a BI platform.
The team had the Python and SQL skills to build these tools. What they didn't have was a way to deploy them securely, at scale, to end users across the company.
Complication
Three pressures converged:
ROI scrutiny. Leadership viewed additional BI platform investment as diminishing returns. The data team needed to demonstrate value beyond dashboards — fast.
Demand outpacing capacity. Multiple business units were requesting data-driven automations. At one dashboard per one to two weeks, the team couldn't keep up, and dashboards weren't what was actually needed.
Cost spiral. BI platform licensing and raw warehouse query costs were growing faster than the value delivered. The client needed a cheaper path to interactive data applications — one that didn't require hiring a full application engineering team.
Previous attempts to deploy data apps internally had failed. The team had experimented with various approaches but lacked the traditional application development expertise to build secure, scalable deployment infrastructure.
Resolution
Narona Data delivered three interventions in a 12-week engagement with a two-person team.
1. Secure Data Application Platform on Kubernetes
The first objective was a platform that matched or exceeded BI capabilities while running on the client's existing Kubernetes infrastructure — no new infrastructure spend.
Within a month, Narona Data deployed Streamlit applications on Kubernetes with a multi-level security architecture:
- OAuth2 authentication connected to the client's federated identity provider
- IdP group-based access control for application-level permissions
- Dynamic data masking tied to data warehouse roles for HIPAA/HITECH compliance on PII/PHI
Why Streamlit: Python-native, lightweight, and aligned with the data team's existing skills. No new language to learn. The security layer was the hard part — Streamlit itself was the easy choice.
2. Development Environment and Component Library
Raw platform access wasn't enough — the data team needed to be productive on it. Narona Data built three enablement layers:
React component library — A custom repository for Streamlit React components, packaged as stable pip packages. This let the team extend Streamlit's native UI without dropping into raw JavaScript.
Application templates — Polished, high-quality Streamlit templates that new applications could be scaffolded from, cutting boilerplate and enforcing consistency.
VS Code devcontainer — A custom development environment for interactive testing of front-end and back-end components simultaneously. Each team member received the environment and hands-on training.
Finally, a CI/CD pipeline automated deployment across local, staging, and production environments with split testing and blue-green deployments.
3. AWS Caching Layer for Sustainable Scale
After initial rollout and user testing, the client wanted to expand further. Streamlit server resources and warehouse query costs were the scaling bottleneck.
Narona Data extended the platform by offloading caching and session management to AWS S3 and ElastiCache. This decoupled application performance from warehouse query volume — applications scaled sustainably without proportional cost increases.
Results
| Metric | Before | After | Business Impact |
|---|---|---|---|
| Data applications shipped | 1 dashboard / 1–2 weeks | 5x more applications | Business units got the interactive tools they needed |
| Deployment time | Days to weeks per app | 90% faster | Data team could iterate and ship without ops bottlenecks |
| Platform cost | BI licensing + warehouse queries | 70% reduction | Freed budget and silenced ROI questions from leadership |
All metrics are directional based on team productivity and cost comparisons against the prior BI-centric workflow.
Ready to Talk?
Facing pressure to demonstrate data team ROI beyond dashboards? Narona Data offers a free consultation to help you find the highest-leverage path forward.