Centene Recovery Platform
When complexity hurts people: redesigning recovery
Leading a $40 million dollar initiative, enabling hospitals to more effectively communicate with insurance companies, challenge costs, and speed-up claim processing.
Centene Corporation
Fixing healthcare at a Fortune 25
Centene Corporation is a leading healthcare enterprise specializing in government-sponsored and commercial healthcare programs, including Medicaid, Medicare, and marketplace insurance plans.
As of writing this, Centene is listed as number 22 on Fortune 500 with 74,000+ employees.
The Problem
The insurance claims process is broken
Healthcare analysts are burdened by manual reviews, high denial rates, and outdated workflows that cost payers time and money while frustrating providers and patients.
This process is time-consuming, prone to error, and ultimately leads to higher patient costs.
Impact
Measuring success, by the numbers
2 years later, the Centene Recovery Platform completed 149 business goals, 53 specific outcomes, and 182 hypotheses tested resulting in 31 published and tested iterations.
- Speed
- +34%Increase in claim processing speed
- Errors
- -62%Reduction in data errors
- Challenges
- -17%Reduction in reimbursement challenges
- Time
- 180 daysSaved from the adjudication process
Kick Off
Leading a new team
Our internal teams run a decentralized team formation to product development. I work as an embedded consultant, leading multiple product teams simultaneously to lead their design initiatives.
Defining the Approach
Building a foundation before anyone writes a line of code
Before the work could begin, the team needed a shared foundation. I introduced three frameworks to shape how we’d operate — Agile product design, Lean thinking, and a clear understanding of the end-to-end design process.
Gathering Team Alignment
The conversation that unblocks everything
Every new product team starts with the same problem: team chaos. Where should we start? Who is doing what? What are we even building?
I facilitate a Product Canvas exercise as a method of gathering team alignment. This exercise allows the team to form a shared basis of understanding.
Moving forward, we can build momentum knowing we all agree on where we currently are and the fundamental problems we’re trying to solve.
The Problem
Parking Lot
Business Goals
Users
User Needs
“
The current process is costing us $38 million dollars annually.
— Business Analyst
“It takes two to four months to process recoveries in our current environment.”
— Product Owner
“Every claim is overpaid by design. That's just how the current system works.”
— Business Analyst
“We heavily rely on 3rd party vendors to manage this process for us.”
— Senior Manager
Strategic Focus
Better quickly instead of perfect later
Fixing the insurance claim process could go many directions. Some of those directions are good, some not so good, and some completely unobtainable. It’s more important to be better quickly than perfect later.
NOW
focused effort
- Replacing Palantir
- Improving the current process
- Automating data flow
- Outdating the need for Excel
LATER
parked, not abandoned
- Solving a larger healthcare problem
- Innovating the adjudication process
- Inventing a new claim adjustment flow
Increase claim validation speed, minimize errors, and simplify the recovery calculation process.
Introducing Dual-Track Agile
How we stopped waiting on each other
To maximize momentum in an Agile framework, I introduced the team to Dual-Track Agile. This cadence enables the product team to know their involvement with the design process and when those involvements will occur.
This cadence cycles with each iteration, enabling the design momentum to stay ahead of development.
I further break down each sprint and introduce a design cadence. This enables the team to understand what is expected of them and when those expectations are required.
Prioritizing Effort
The fastest path to something that matters
With an agile mindset and an objective of publishing quickly, the goal is to determine where we can provide the most value with the least amount of effort.
The Hypothesis Prioritization process from Lean UX builds team alignment while prioritizing momentum. We can use the data collected from our Product Canvas to quickly build hypotheses.
We will achieve
Increased claims processing speed
Hypotheses were presented to the greater team for prioritization. Product Owners can evaluate the value of each outcome while developers and analysts can evaluate the effort involved with building the feature.
Value × effort
Hypothesis Prioritization
High value · Low effort
Quick wins
View an organized / grouped list of claims
through Claim Filtering
View claims organized by priority
through a timeliness indicator
Take action on a claim
through ‘Approve’, ‘Pend’, and ‘Deny’ actionables on every claim
View an organized / grouped list of claims
through Claim Sorting
View the edit history of a claim
through a ‘View history’ screen that displays historical data
High value · High effort
Strategic bets
Filter quickly
through applying multiple saved filters simultaneously
Measure denials
through specifying why a denial occurred via a dropdown of generic reasons
View specific and customized claim details
through customized dashboard layouts
Low value · Low effort
Fill-ins
PI Analyst
Export a list of claims
through CSV exporting
Prioritize which claims to view
through applying multiple sorts simultaneously
Low value · High effort
Later
Find claims quickly
through error prevention via form validation
Healthplan Analyst
Communicate claim actionability to the PI team
through the ability to write comments on claims
See relevant claims quickly
through saving specific filter criteria
Compare claims to contra claims
through a comparison table that flags differing fields
Use subtractive filtering
through ‘or’ filtering
View more information about a claim quickly
through expanding and collapsing rows with user-created notes
Find claims more easily
through conditional filtering (contains, between, more than)
Choosing a Design System
Why we started with a system
The team chose PrimeReact, an open-sourced, federated design system to meet our constraints with tech stack, legal, and maintenance requirements. This system excels at presenting and manipulating data — a mechanic paramount to our users’ workflows.
High-Fidelity Agile Iteration
Skipping the sketches, shipping the vision
Based on our two-week sprint cycles and values of a lean, agile workflow, value of effort was spent on quick, high-fidelity mockups over wireframes.
To communicate iterations and effort, we use a process called Slicing.
Test & Measure
What the users had to say
To validate the momentum of this iteration, I demo with various stakeholders and users to challenge if the output drives toward the intended outcome. Feedback was gathered enabling us to build new hypotheses as we begin tackling the next iteration of outcomes.
“I need to see the history behind the claim, the story it has gone through, and the decision-making that has been made along the way.”
— Business Analyst
“Would love to not have to open Amysis, Excellis, or other applications. It would be great to be able to see everything in one place.”
— Business Analyst
“I hate emails. We get far too many of them and things are getting lost.”
— Business Analyst
“A big part of this process is vendor management — who gets what and how much. Looking at trend data has helped in the past but we could use deeper analytics.”
— Product Owner
Prioritizing Outcomes
Making the roadmap make sense
To maintain communication with the broader stakeholder team, momentum is captured through Outcome Prioritization and Feature Rollouts.
Sprint × Hypothesis
Tracking outcomes across iterations
Enables any stakeholder to view a team’s discovery effort and areas of focus.
Dashboard
a single view for all recoveries
through a claim dashboard that displays both claim lists (master and Payment Integrity Exclusions)
Claim View
the ability to review specific claims
through a Claim Details page
We will know we succeeded when
Team alignment and a baseline for future innovation.
Dashboard
a customized, grouped list of claims
through claim filtering
access to a specific claim
through claim search
high-level information about a claim
through additional data columns on the claims dashboard
Claim View
a simplified view with only relevant details
through expanding / collapsing tables
We will know we succeeded when
Reduction in the excel documents needed to build a requested claim list.
Dashboard
an organized, grouped list of claims
through claim sorting
access to claims further down the list
through pagination
more specific information about a claim from the dashboard
through expanding a row to see additional information / notes
the ability to quickly filter
through quick-filtering table columns
Claim View
the ability to view a claim's edit history
through a claim audit log
We will know we succeeded when
Analysts no longer use excel to prioritize effort.
And so on…
a single view for all recoveries
through a claim dashboard that displays both claim lists (master and Payment Integrity Exclusions)
the ability to review specific claims
through a Claim Details page
Team alignment and a baseline for future innovation.
a customized, grouped list of claims
through claim filtering
access to a specific claim
through claim search
high-level information about a claim
through additional data columns on the claims dashboard
a simplified view with only relevant details
through expanding / collapsing tables
Reduction in the excel documents needed to build a requested claim list.
an organized, grouped list of claims
through claim sorting
access to claims further down the list
through pagination
more specific information about a claim from the dashboard
through expanding a row to see additional information / notes
the ability to quickly filter
through quick-filtering table columns
the ability to view a claim's edit history
through a claim audit log
Analysts no longer use excel to prioritize effort.
Feature Rollouts
Output momentum across iterations
Enables all stakeholders to view the output momentum of the application within each iteration.
Dashboard
Claims Dashboard
Master and Payment Integrity Exclusions claim lists in a unified view.
Claim View
Claim Details Page
A focused interface for reviewing a single claim.
Dashboard
Filtering & Search
Customized, grouped lists and quick access to specific claims.
Data Columns
High-level claim information surfaced on the dashboard.
Dashboard
Sorting & Pagination
Organized, navigable claim lists.
Inline Detail & Quick-filter
Row expansion plus column quick-filtering.
And so on…
Claims Dashboard
Master and Payment Integrity Exclusions claim lists in a unified view.
Claim Details Page
A focused interface for reviewing a single claim.
Filtering & Search
Customized, grouped lists and quick access to specific claims.
Data Columns
High-level claim information surfaced on the dashboard.
Sorting & Pagination
Organized, navigable claim lists.
Inline Detail & Quick-filter
Row expansion plus column quick-filtering.
Design
Bringing it all together
What started as a replacement for Palantir became a full-scale recovery platform handling everything from claim validation to vendor management.
Below: a short tour through four notable views, pulled directly from the working Figma file and annotated using the documentation that shipped with each slice. Together they trace how the analyst’s view evolved across 31 tested iterations.
The dashboard
The analyst's home. A left-rail navigator across all six recovery-platform surfaces, a persistent claims toolbar (batch actions, filter, settings, search), and a sortable table that doubles as the link out to a single claim view. Every claim arrives here after the exclusions process.
- Sidebar — Service Lines, Post Pay DRG Selection, Post Pay Audit Results, Rules Management, Direct Bill Claims, and Value Simulator. The platform grew well beyond a single dashboard; the sidebar is how analysts pivot between surfaces without losing place.
- Claims toolbar — Batch Actions for multi-row workflows, Filter for narrowing the queue, Settings for column visibility, and Search for jumping straight to a known claim. Always-visible above the table so the workflow doesn't dead-end.
- Sortable claims table — Claim Number, Vendor, Business Line, Adjustment Amount, Program, Recovery Sent, Provider, HP Deadline, TIN, and Status. Every column carries a sort affordance; the column set grew well beyond Slice 1's four as analyst hypotheses shipped.
- Status — one of four states (Approved, Denied, On Hold, Pending). Non-approved tags hover-reveal which exclusion stage caught them (Master Exclusions, PI Lookback, or Exclusion Set) plus any analyst comments, using the Disclosure component.
- User profile + feedback — surfaced at the bottom of the sidebar so analysts always know whose context they're in, with a one-click feedback link out to the team.
Anatomy of a claim
A single claim's full story on one scannable page. The header carries at-a-glance status, audit-log access, and a lookback timer; below it, the Healthplan Manager's verdict, prior adjustment comments, and the full structured property set.
- Header section — Provider, Claim Number with status tag, plus the at-a-glance metadata: Submitted On, Lookback Period (with a warning icon when the deadline is tight), Adjustment Product, and Adjustment Amount. The Audit Log button opens the full action history over the page (next screen).
- Health Plan Review — the Healthplan Manager's verdict on the claim. When approved, the card collapses to a header plus the date. When denied (as here), it expands to surface the analyst, the reason, and the full comment. Change Status triggers a confirmation modal.
- Adjustment Comment — surfaces any prior comments left on the claim so context isn't lost between handoffs, with the vendor reference inline.
- Properties — the full structured data table about the claim itself: claim metadata, provider details, adjustment modifiers, and Medicare / OIC fields. "No Value" placeholders keep the rhythm scannable when fields are unset.
The Audit Log
The Audit Log modal opens over the claim view so the analyst doesn't lose their place in the queue. Every action on the claim, indexed and searchable — the team's answer to "how did this claim get into this state?"
- Audit Log table — every item is organized in a table with three columns. Item Number simply numbers each edit so analysts can reference a single line even when dates or notes look similar. Date is when the action occurred. Notes describes the action in plain language.
- Bolded items — anything tied to passing or failing an exclusions step renders bold so the eye lands on what changed.
- Modal scrim — the table is scrollable so history can grow without bounding the page; the scrim keeps the underlying claim view in soft focus.
Saved filter Setups
By the latest 🟢 Advanced Filter slice, the right-rail filter rolls up into named Setups — analysts can save the exact configuration they use on a Tuesday morning, then reload it on Wednesday without rebuilding it field-by-field.
- Default Setup — always present in the Setups list, can't be edited or removed, and reloads the original pre-customization filter. The escape hatch when a saved setup goes sideways.
- Save as New — names and saves the current setup without overriding existing ones. Switches the user to the Setups tab on save so they can confirm the new entry landed.
- Save (existing) — overwrites the currently loaded setup with the new field selection. Disabled when the loaded setup is the Default, forcing Save as New for any modification.
- Setup management — the kebab menu offers Rename or Delete on any user-created setup. Delete loads an inline confirm so it can't fire on a misclick.
Component details
The small craft — modals, controls, and brand moments that make the platform feel like a single, considered system.
Team Health
Measuring the team, not just the product
While not as paramount as measuring impact, these softer metrics assist in validating application and process momentum. Less about user outcomes and more about team health.
- Complete Business Goals
- 149
- Outcomes Delivered
- 53
- Hypotheses Tested
- 182
- Hypotheses Developed
- 125
- Developed Iterations
- 31
- Usability Tests
- 64
- Unique Personas
- 17
- Completed User Needs
- 207
Closing
The American healthcare system needs help
The American healthcare system isn’t perfect — and I know that more personally than professionally. I got into this work because financial barriers have a way of outweighing human health, and I’ve felt that cost firsthand.
Projects like this one don’t fix everything, but they move the needle. Faster claims, fewer errors, less time lost in a broken process — that’s real impact for real people. It’s not the whole problem solved, but it’s progress. And progress is worth building.
