Platform
Nine detection layers. One executive summary.
Built for the patterns MCA portfolios actually default on.
Every credit passes through nine independent detectors and an agentic synthesis step. The output is a single risk score, an adjusted-revenue figure, and a ranked list of flags an underwriter can defend in committee.
Key takeaways
- Vyaso re-underwrites every bank statement in 90 seconds per file. ✓ verified
- Nine detection layers cover loan stacking, account kiting, circular transactions, window dressing, statistical anomalies, counterparty risk, daisy-chain inflation, graph-cycle laundering, and PDF integrity.
- The output is a confidence-weighted adjusted-revenue figure, not a gross-deposits total.
- An agentic synthesis step produces an executive summary that underwriters paste into the credit memo.
- Detection is industry-tuned. The same statement reads differently for a restaurant than for a construction firm.
Nine detection layers.
Each layer is independent. Each maps to a pattern that has defaulted MCA portfolios for years.
Layer 01
Loan stacking
Hidden lender debt
Multiple lender deposits arriving in a short window. Identifies deposits from MCA funders, fintech lenders, factors, and other fund providers, even when they're disguised as customer revenue.
Layer 02
Account kiting
Phantom deposit volume
Self-transfers between the merchant's own accounts that double-count the same money. Catches both keyword-obvious cases and fuzzy-name relationships across multiple statements.
Layer 03
Daisy-chain inflation
Long-horizon round-tripping
Counterparties whose money flows in a circle over weeks or months. The slow-burn version of circular fraud, invisible to short-window detectors.
Layer 04
Symmetric transactions
Recycled money
Debit-credit pairs of similar amount with the same counterparty that net out to nothing real. The simplest form of revenue inflation.
Layer 05
Graph-cycle laundering
Multi-hop circular flows
Money moving through three or more counterparties in a tight loop and returning to the merchant. Catches sophisticated schemes that no single transaction reveals.
Layer 06
Counterparty risk
Concentration and related parties
Single-source revenue dependencies, fuzzy-name matches against the merchant, bidirectional flows that suggest related-party transactions.
Layer 07
Window dressing
Period-boundary inflation
Large deposits at the end of a statement period that withdraw at the start of the next, designed to inflate average daily balance for underwriting snapshots.
Layer 08
Statistical anomalies
Numerical fingerprints of fabrication
Distributions that violate Benford's Law, excessive round numbers, deposits clustered just below reporting thresholds, and total deposits that exceed declared revenue by suspicious margins.
Layer 09
PDF integrity
Forgery and alteration
Forensic checks on the file itself: creation tools, font consistency, running-balance continuity, cross-page balance breaks. When this fires it fires with high confidence.
Each layer is described in detail in the glossary. See the glossary →
We don't subtract fraud. We re-underwrite.
The outcome
Most bank-statement tools give underwriters a binary view: this transaction is fraud, that one isn't. Vyaso doesn't.
Every credit passes through the nine detection layers and is weighted by how much of it represents real customer revenue. Some deposits are obviously revenue. Some are obviously not. Most fall in between.
The output is a single number, adjusted revenue, that reflects the weighted certainty of every credit on the statement, not a fraud-or-not flag. It is the number underwriters can defend in committee and in collections.
When a deposit is genuinely ambiguous, an agentic synthesis step reviews it with full context (description, amount, counterparty, industry) and resolves it.
Revenue waterfall
The revenue waterfall is one view in the dashboard. Every subtraction is reproducible and exportable. ✓ verified
One detector. Many industries.
The same statement reads differently for a restaurant than for a construction firm. Restaurants settle in daily card payouts; construction settles in milestone draws. Healthcare settles in payer reimbursements that look concentrated but aren't. Transportation looks like loan stacking when freight factoring is normal. Vyaso is industry-tuned: detection thresholds adapt to each segment automatically, classified from the company domain when one is provided.
What Vyaso isn't yet.
Here is what Vyaso doesn't do today.
Direct API integrations with Ocrolus, Heron Data, Validis
Today, raw PDF and CSV upload work. Direct API integrations with leading parsers ship in the pilot program.
Single sign-on and per-user accounts
Today, access is gated by a shared credential per pilot engagement. SSO, role-based permissions, and audit logs are next.
SOC 2 Type II
Compliance posture is GLBA-aligned in our handling practices. SOC 2 Type II audit is in progress; certification target communicated during pilot conversations.
What the underwriter actually sees.
The dashboard is the deliverable. Eight surfaces, every one exportable.
Risk score
82
High
Merchant: ACME CORP
Executive summary
Merchant shows 2 lender deposits within 5 days (QUICKFUND CAPITAL, VELOCITY LENDING) consistent with active stacking. A $30,000 self-transfer inflates monthly volume. Adjusted revenue lands at $48,000 of $158,000 gross deposits. Recommend decline pending merchant clarification on QUICKFUND balance.
Gross deposits
$158,000
Adjusted revenue
$48,000
Counterparties
4
Risk gauge, executive summary, and the cluster of topline numbers: total deposits, adjusted revenue, flagged amount, unique counterparties. The view a credit officer pastes into the committee deck.
Sample
See a full analysis.
Request a redacted sample Vyaso analysis: risk score, ranked flags, network graph, revenue waterfall, and the executive summary that pasted into a real credit memo. PDF, 8 pages.
Real Vyaso output. Merchant identifiers redacted by mutual agreement with the design partner.
Frequently asked.
Run Vyaso on your portfolio.
Free 30-day pilot. Bring 50–100 files. We'll show you what the model would have flagged, what would have been approved, and how the adjusted revenue compares.
No commitment. No setup fee.