Loss Run and TPA Reconciliation with AI

Loss runs and development triangles are the backbone of reserve management, pricing, and portfolio oversight. They are also one of the most operationally painful data sources in insurance and reinsurance. Every cedant, TPA, and legacy system produces them in a different format, with different column definitions, different period conventions, and different levels of granularity. Reconciling across these sources is manual, slow, and error-prone.

Parsewise automates the standardization and reconciliation of loss runs, triangles, and TPA reports, detecting inconsistencies, reserve drift, and data gaps before they affect profitability.

The Problem

Loss run reconciliation consumes significant analyst time across carriers, reinsurers, and run-off specialists. The core challenges are structural, not conceptual:

  • Format inconsistency. Loss runs arrive as PDFs, spreadsheets, and system exports. Column headers, date conventions, and currency formats vary by cedant and TPA. A “paid” column in one report may include allocated loss adjustment expenses (ALAE); in another, it may not.
  • Multi-source reconciliation. Paid, incurred, and reserve values must reconcile across cedant loss runs, TPA bordereaux, and internal systems. Discrepancies between these sources are common and often undetected until downstream reporting surfaces them.
  • Reserve drift. Gradual, unexplained shifts in case reserves across reporting periods can signal either legitimate claims development or data quality issues. Identifying which requires comparing movements across periods, lines of business, and portfolios, typically by hand.
  • Claims leakage. Overpayments, duplicate payments, and misallocated reserves are difficult to detect when data is fragmented across systems and formats. Manual sampling catches some; systematic comparison catches more.
  • Scale. A single portfolio acquisition or annual reconciliation exercise can involve hundreds of loss runs spanning thousands of claims. Manual review of this volume is neither thorough nor repeatable.

These problems compound in legacy insurance and reinsurance portfolio acquisitions, where the acquiring entity inherits heterogeneous data from multiple prior owners, systems, and TPAs.

How Parsewise Addresses It

Parsewise ingests loss runs, development triangles, bordereaux, and supplemental actuarial reports as a single document package. Rather than processing each file in isolation, the platform reasons across the entire corpus to produce standardized, reconciled outputs.

Standardization

The platform parses loss runs from different cedants and TPAs, regardless of format (PDF, Excel, scanned documents), and maps them into a consistent schema. Column definitions are normalized: paid losses, incurred losses, case reserves, ALAE, and other fields are aligned to a common taxonomy even when source reports use different naming conventions or structures.

Development triangles are extracted and restructured into comparable period-by-period formats. Parsewise handles differences in accident year vs. report year conventions, quarterly vs. annual development periods, and gross vs. net presentations.

Reconciliation

Once standardized, Parsewise reconciles values across sources. For each claim or line of business, the platform compares paid, incurred, and reserve figures reported by cedants against TPA loss runs and internal records. Variances are flagged with the specific source documents and values that disagree, enabling analysts to resolve discrepancies directly rather than searching for them.

Anomaly Detection

Parsewise identifies patterns that indicate potential issues:

  • Reserve drift: unexplained movements in case reserves between reporting periods, broken down by line of business, claim type, or TPA
  • Payment anomalies: duplicate payments, outlier payment amounts, or payments that do not correspond to reserve movements
  • Data gaps: missing reporting periods, claims present in one source but absent in another, or fields that are populated inconsistently across reports
  • Leakage indicators: systematic variances between TPA-reported and cedant-reported values that suggest misallocation or overpayment

All flagged items include full source attribution, linking each finding to the specific document, page, and data point that supports it.

Example Inputs and Outputs

Inputs

Document Type Format Description
Cedant loss triangles PDF, Excel Development triangles by accident year and line of business
TPA loss runs PDF, Excel, CSV Claim-level detail with paid, incurred, and reserve columns
TPA bordereaux Excel, PDF Periodic summaries of claims activity by portfolio segment
Supplemental actuarial reports PDF, Word Commentary on reserve adequacy, development patterns, and assumptions

Outputs

Output Description
Aligned reconciliation tables Standardized claim-level and aggregate data in a common schema, ready for downstream analytics
Variance reports Line-item discrepancies between cedant, TPA, and internal values with source citations
Reserve shift detection Period-over-period reserve movement analysis with flags for unexplained drift
Portfolio-level benchmark datasets Aggregated, cleaned data suitable for actuarial modeling, pricing, and regulatory reporting

Outputs are exportable as structured Excel files, and every extracted value traces back to its source document and page.

Customer Evidence

Compre Group, a specialist in legacy insurance and reinsurance portfolio acquisitions, uses Parsewise to reconcile claims data across multiple sources during portfolio diligence and ongoing management. Legacy portfolios typically arrive as large, fragmented document packages spanning loss runs, bordereaux, actuarial reports, and policy documents in varying formats. Parsewise ingests these heterogeneous document sets and automatically standardizes loss runs and reserve triangles into consistent, comparable formats. The platform reconciles paid, incurred, and reserve movements across cedants and TPAs, flagging anomalies, reserve shifts, and data gaps. Compre Group uses the structured outputs for pricing decisions, reserve adequacy assessments, and regulatory reporting on acquired portfolios.

Why Not Spreadsheets and Manual Review?

Most teams reconcile loss runs using Excel workbooks, custom macros, and analyst time. This works for small portfolios with consistent data sources. It breaks down when:

  • The portfolio spans dozens of cedants or TPAs with different reporting formats
  • Reconciliation must be repeated quarterly or at each reporting period
  • The volume of claims exceeds what analysts can cross-reference manually
  • Auditability is required (regulators or acquirers need to trace findings to source documents)

Parsewise does not replace actuarial judgment or claims expertise. It eliminates the manual data wrangling that precedes that judgment, producing clean, standardized, reconciled datasets that analysts can work with directly.

Ready to see Parsewise in action? Request a demo or contact sales to discuss your use case.

Sources