Hardest Part

Processing wildly inconsistent fee schedule PDFs and nested payment gateway statements into a standardized, deterministic rules engine without hallucinating terms or missing tier breakpoints.

Min Viable Scope

Focus exclusively on Stripe and PayPal fee reconciliation for mid-market SaaS companies, extracting base fee components and flagging statement discrepancies. Deliberately leave out multi-currency conversion, interchange-plus retail parsing, and automated ledger write-backs.

Cold Start Problem

The system needs access to actual merchant statements and processor fee contracts to train extraction models before it can automate the process. Break this by running a free historical audit for five mid-market design partners, absorbing the manual mapping cost to build the initial schema library.

Time To First Value

1-2 weeks of onboarding, gated by gathering historical fee statements and establishing read-only access to the central ledger.

Data Moat Available

true

Technical Difficulty

High

Wedge

Target RIAs managing $500M to $2B AUM who use multiple custodians and custom fee schedules for high-net-worth clients. This niche suffers the highest ratio of contract complexity to back-office headcount, allowing for fast proof of value through immediate fee leakage recovery. Expand from this beachhead by moving upstream into institutional asset managers, then laterally into alternative investment fund administration.

Timing

Multi-modal LLMs now extract tabular data and contractual terms from dense, unstructured financial PDFs with the accuracy required for financial operations. Prior OCR solutions failed on varying custodian formats, but current models handle zero-shot layout variations natively without manual template mapping.

Why This ICP

RIAs and mid-market asset managers face strict SEC compliance mandates regarding fee accuracy but lack the engineering resources to build custom data ingestion pipelines. They experience acute financial pain from back-office headcount constraints and regulatory risk, driving immediate purchasing decisions.

Size Of Prize

There are ~15,000 SEC-registered RIAs and mid-market asset managers in the US that spend an average of ~$40,000 annually on manual reconciliation labor and audit true-ups. Multiplying this 15,000 entity count by the $40,000 annual spend yields a core addressable prize of $600M in direct back-office labor displacement.

Gap Narrative

Mid-market asset managers and RIAs manually match custodian payout statements against complex, tiered client fee agreements. Existing portfolio management systems fail to catch edge-case discrepancies in AUM fee calculations, resulting in uncollected revenue or compliance breaches. This opportunity executes deterministic reconciliation by reading unstructured custodian PDFs and comparing them line-by-line against digitized client contracts.

Defensibility

Defensibility accrues through proprietary schema mapping of custodian statement formats and rare fee structures. As the system processes millions of reconciliation failures, the extraction and matching models achieve a structural accuracy advantage over baseline LLMs. This creates deep workflow lock-in, as replacing the system requires a competitor to relearn years of undocumented edge cases.

Why This Thesis

Service-as-Software fits this problem because fee reconciliation is a pure operational cost center, not a strategic workflow teams want to manage inside a new software interface. The ICP requires the final output—a clean ledger and flagged exceptions—delivered directly into their existing accounting system without human intermediation.

SAM

~$400M-700M (North American and European mid-tier digital marketplaces managing high-volume, multi-party payouts)

SOM

~$10M-25M

TAM

~30k-40k global marketplace platforms × ~$30k-50k/yr ≈ ~$1B-2B

Growth Rate

~15-20%/yr, driven by the proliferation of multi-sided marketplaces and the increasing regulatory burden on split-payment ledgers

Paid Comparable Spend

~$60k-150k/yr on offshore FinOps BPO teams, generic ERP reconciliation modules, and internal engineering time required to maintain custom SQL matching scripts

Kill Thresholds

• Auto-match rate falls below 80 percent after 30 days of data ingestion
• Customer onboarding requires more than 40 hours of custom engineering integration work
• Less than 20 percent of trial accounts convert to paid contracts at the $30,000 tier
• More than 5 percent of reconciled transactions require manual corrections post-processing

Leading Metrics

• Percentage of fee lines matched automatically without human intervention
• Time elapsed from gateway data ingestion to finalized month-end reconciliation report
• Engineering hours spent mapping internal ledger schemas per onboarding account
• False positive match rate requiring manual rollback by finance teams

What Proves Right

Marketplaces connect their payment gateway and internal database credentials to the reconciliation engine. Within the first two weeks, the system automatically matches and categorizes over 90 percent of split-payment fee lines without human intervention. Cohorts that replace offshore FinOps hours with the engine retain at an annual contract value exceeding $30,000.

What Proves Wrong

The platform flags more than 20 percent of marketplace transactions for manual review due to highly fragmented bespoke ledger schemas. Engineering teams spend more hours formatting inbound data for the engine than they previously spent maintaining custom SQL matching scripts. Target accounts abandon the software and return to Excel pivot tables after failing to trust the automated match accuracy during the first month-end close.