Problem statement

The current claim journey, its failure mechanisms, alternatives and measurable business boundary.

Product decision
Treat document quality, provider identity and transaction continuity as distinct product problems.

Document type

Problem statement

Status

Portfolio sample · written from my healthcare claims experience

Product decision

Treat document quality, provider identity and transaction continuity as distinct product problems.

Portfolio sample from my healthcare product experience. Requirements, thresholds and review records are illustrative; they are not employer documents or historical sign-offs.

In this document 7 sections

Business problem

Supported paper claims require repetitive entry before electronic processing. A clean extraction result still creates limited value if provider evidence is incomplete, the transaction is rejected, or downstream completion is invisible. Operators then repeat work across intake, matching and recovery. Customer care inherits an ambiguous status, while product reporting can overstate improvement by counting successful intermediate steps instead of completed claims.

1. Supported paper claims require repetitive entry before electronic processing
2. A clean extraction result still creates limited value if provider evidence is incomplete, the transaction is rejected, or downstream completion is invisible
3. Operators then repeat work across intake, matching and recovery
4. Customer care inherits an ambiguous status product reporting can overstate improvement by counting successful intermediate steps instead of completed claims

Evidence and attribution

I led the conversion of paper claims to EDI 837 transactions for automated adjudication. More than 400,000 paper claims were converted to EDI 837, an 80-person team was redeployed and the business saved about $1M a year. Separately, I designed the end-to-end provider flow against the provider master and proved it with a 99% match in a proof of concept. A third experience replaced a legacy claims engine with an in-house engine. These are separate projects, not one deployment; their outcomes are not combined or annualized.

1. More than 400,000 paper claims were converted to EDI 837 an 80-person team was redeployed and the business saved about $1M a year
2. proved it with a 99% match in a proof of concept the end-to-end provider flow against the provider master
3. replaced a legacy claims engine with an in-house engine

These are separate projects, not one deployment; their outcomes are not combined or annualized

Current journey and friction

The following journey expresses the problem this product contract addresses.

Step	User need	Failure mechanism	Consequence
Receive form	Know the source is supported and readable	Malformed or incomplete source	Manual clarification before entry
Extract and correct	Trust required values	Uncertain field looks complete	Error moves downstream
Resolve provider	Use supported identity evidence	Missing name or variable address	No match, false match or repeated lookup
Generate and submit	Meet receiver requirements	Wrong variant or invalid field	Rejection and rework
Track completion	Know what happened after handoff	Acknowledgement confused with disposition	Misleading status or unsafe retry

Causal hypotheses and validation

The first hypothesis is that eligible transcription can be removed without removing interpretation. The second is that a portion of matching friction comes from evidence completeness and representation, rather than slow infrastructure alone. The third is that hidden intermediate states create repeated work and inaccurate reporting. Validate each by tracing a recent failed task and a comparable successful task, inspecting source/correction history and segmenting exception reasons. These hypotheses must remain distinguishable: a faster API cannot compensate for the wrong provider, and more OCR capacity cannot fix a receiver-contract error.

1. eligible transcription can be removed without removing interpretation
2. a portion of matching friction comes from evidence completeness and representation rather than slow infrastructure alone
3. hidden intermediate states create repeated work and inaccurate reporting
4. tracing a recent failed task and a comparable successful task inspecting source/correction history and segmenting exception reasons

Alternatives and chosen scope

Manual-process improvement reduces disruption but leaves repetitive entry. Unrestricted straight-through automation maximizes apparent throughput while concealing unresolved evidence. The chosen product scope is supported-form automation with an explicit provider contract, validated 837 handoff and accountable review. Platform replacement is a separate investment because consumer continuity and cutover risks require additional evidence. Medical 837 processing does not include retail pharmacy NCPDP transactions or change benefit/clinical decisions.

1. Manual-process improvement reduces disruption but leaves repetitive entry
2. Unrestricted straight-through automation maximizes apparent throughput while concealing unresolved evidence
3. The chosen product scope is supported-form automation with an explicit provider contract, validated 837 handoff and accountable review
4. Platform replacement is a separate investment because consumer continuity and cutover risks require additional evidence

Medical 837 processing does not include retail pharmacy NCPDP transactions or change benefit/clinical decisions

Measures and decision boundary

BR1–BR6 connect the business problem to FR1–FR10. Compare receipt-week cohorts by form and channel, keeping eligible, all-received and pending counts visible. Measure net automation, cost per completed claim, p50/p95 completion time, corrections, false matches, EDI rejects and reconciliation. The sample provider API target is p95 ≤500 ms at 50 requests/second, with source-query/cache paths and errors reported separately. Operations and customer care validate useful recovery; claims/provider specialists validate interpretation; legal, compliance and relevant patient-safety partners review consequential boundaries. Expansion requires end-to-end quality, not a better isolated rate.

CMS: professional claims and CMS-1500 ↗ · CMS: institutional CMS-1450 / 837I ↗ · CMS: adopted transaction standards ↗

1. Compare receipt-week cohorts by form and channel keeping eligible, all-received and pending counts visible
2. Measure net automation, cost per completed claim, p50/p95 completion time corrections, false matches, EDI rejects and reconciliation
3. The sample provider API target is p95 ≤500 ms at 50 requests/second with source-query/cache paths and errors reported separately
4. Operations and customer care validate useful recovery claims/provider specialists validate interpretation

Expansion requires end-to-end quality, not a better isolated rate

Decision to proceed

Proceed with a bounded cohort only if the complete path and its exceptions can be represented in one reconciled inventory. Keep a manual baseline for unsupported cases. If measured benefit comes from excluding difficult claims, the hypothesis has not succeeded. If identity errors rise, resolve that cause before interpreting lower processing time as an improvement.

1. Proceed with a bounded cohort only if the complete path and its exceptions can be represented in one reconciled inventory
2. Keep a manual baseline for unsupported cases
3. If measured benefit comes from excluding difficult claims the hypothesis has not succeeded
4. If identity errors rise, resolve that cause before interpreting lower processing time as an improvement