How MCA Lenders Use AI to Score Concentration Risk Before Funding Big Deals

When Big MCA Deals Go Wrong, They Go Very Wrong

AI fraud detection for business lending has traditionally focused on catching forged documents and synthetic identities. But the most expensive losses in merchant cash advance don't come from fraud at all. They come from concentration risk on big deals that looked perfectly clean on paper.

A recent deBanked investigation detailed how a single conglomerate of car dealerships in California went out of business in November 2018, triggering the bankruptcy of 1 Global Capital, which had lost more than $40 million on that one deal. The dominoes kept falling from there. This wasn't an isolated incident. The pattern repeats across the MCA industry whenever funders chase large advances without understanding how concentrated the merchant's revenue truly is.

The problem is structural. Standard underwriting looks at deposit volume, average daily balance, NSF history, and existing positions. None of those metrics tell you whether 60% of a restaurant's revenue comes from a single catering contract, or whether a construction company's deposits all trace back to one general contractor. When that single source disappears, so does the merchant's ability to make daily remittances. The funder absorbs the full loss.

What's changing in 2026 is that AI systems can now analyze bank verification data, specifically recorded banking sessions, to surface concentration risk signals that human reviewers consistently overlook. This article breaks down how that scoring works, what signals matter most, and how funders can integrate concentration risk analysis into their verification workflow before writing large checks.

What Concentration Risk Actually Means for MCA Funders

Beyond Total Deposits: The Metrics That Matter

Concentration risk in MCA underwriting refers to the degree to which a merchant's cash flow depends on a small number of customers, revenue sources, or seasonal windows. A merchant processing $500,000 per month through their bank account looks healthy at the surface level. But if $350,000 of that comes from three customers, losing just one could cut deposits by 20% or more overnight.

Traditional bank statement analysis treats all deposits equally. A PDF statement shows a list of credits with descriptions and amounts. An underwriter scanning that document sees volume. They rarely have the time, or the tools, to categorize each deposit by source, identify recurring patterns, map customer concentration ratios, and flag seasonal dependencies. The math is straightforward; the manual effort is not.

This is precisely where concentration risk verification gaps expose funders to losses that better tooling would prevent. The gap isn't knowledge. Every experienced underwriter knows concentration risk exists. The gap is workflow. There's no efficient mechanism in most shops to measure it during the verification stage.

How AI Scores Concentration Risk from Bank Verification Data

Modern AI systems applied to bank verification recordings can perform several analyses that manual review cannot match at scale. The first is deposit source clustering. By examining transaction descriptions, amounts, and timing patterns across 90 to 180 days of banking activity visible in a recorded session, machine learning models can group deposits into likely source categories. Repeated deposits of similar amounts from the same originator get tagged as a single revenue source.

The second is a Herfindahl-Hirschman Index calculation applied to deposit sources. This is a standard concentration metric borrowed from antitrust economics. A merchant with deposits spread evenly across 50 customers scores very differently from one with deposits concentrated in three. AI can compute this in seconds from categorized transaction data.

Third, temporal dependency analysis identifies whether certain revenue sources are seasonal, cyclical, or one-time. A landscaping company that shows strong deposits from April through October but near-zero from November through March has a concentration risk that's temporal rather than customer-based. AI models trained on MCA portfolio data can flag these patterns and assign risk multipliers accordingly.

None of this analysis is possible from a static PDF bank statement. It requires visibility into the actual banking portal, where transaction details, running balances, and date ranges can be examined in context. This is why recorded bank verification sessions, where the merchant navigates their live banking portal on camera, provide the raw data that AI concentration scoring needs.

Why Async Recordings Enable Better Analysis Than Live Calls

Live verification calls have a fundamental limitation for concentration risk analysis. The underwriter on the call is focused on confirming identity, verifying specific transactions, and checking for obvious red flags. They're processing information in real time, under time pressure, with the merchant waiting on the line. Asking an underwriter to simultaneously perform concentration risk scoring during a live call is unrealistic.

Asynchronous bank verification changes the dynamic entirely. When a merchant records their banking portal session through a platform like Exact Balance, that recording becomes a permanent, replayable asset. The underwriter can review it once for standard verification, then hand it to an AI system for deeper analysis. The AI can pause on each visible transaction, extract description text through optical character recognition, cluster deposit sources, and generate a concentration score before the underwriter even opens the file.

This is the operational advantage that async verification creates for large deals. As we've explored in our analysis of how AI detects fake banking sessions in screen recordings, the same recording that validates authenticity can simultaneously feed concentration risk models. One recording, multiple layers of analysis, zero additional merchant interaction.

Building a Concentration Risk Workflow for Large MCA Deals

Implementing concentration risk scoring doesn't require rebuilding your entire underwriting stack. It requires adding a decision gate for deals above a certain threshold, typically $250,000 or more, where the potential loss justifies deeper analysis.

The workflow starts with the standard Exact Balance process. Your team creates a verification request with custom instructions specifying the date range you need to see, typically 90 to 180 days of transaction history. The merchant receives a secure link, records their live banking portal in their browser without installing any software, and submits the recording. Your underwriter reviews for authenticity and standard red flags.

For deals above your concentration risk threshold, a second analysis layer activates. The recording is processed through AI models that extract transaction-level data from the visible banking portal. Deposit descriptions are parsed and clustered. A concentration score is generated along with a breakdown showing the estimated percentage of revenue attributable to the top one, three, and five deposit sources. Seasonal patterns are flagged with historical comparisons.

The output isn't a pass/fail decision. It's a risk adjustment. A merchant with a concentration score of 0.35 on the HHI scale, indicating moderate concentration, might warrant a smaller advance amount or a shorter term than their deposit volume alone would suggest. A merchant scoring 0.70 or above, indicating severe concentration, should trigger a manual review of whether the deal economics justify the risk at any advance size.

This scoring also feeds portfolio-level analysis. If your fund already has exposure to the same industry vertical or geographic market as the merchant, the deal-level concentration risk compounds with portfolio-level concentration. AI models can cross-reference new deal data against existing portfolio composition to flag compounding risks that deal-by-deal underwriting would miss.

A Real-World Scenario: The Construction Subcontractor

Consider a construction subcontractor applying for a $400,000 merchant cash advance. Their bank statements show $1.2 million in monthly deposits. Average daily balance is strong. No NSF transactions in 90 days. Two existing positions totaling $180,000. By standard metrics, this looks like a solid deal.

But the bank verification recording tells a different story. When AI analyzes the visible transactions, it identifies that 73% of deposits trace back to payments from a single general contractor. The remaining 27% is split across eight smaller clients. The HHI score comes back at 0.54, which is solidly in the high-concentration zone.

What does this mean practically? If the general contractor delays payment by even two weeks, the subcontractor's daily deposits drop by nearly three-quarters. Their ability to service the MCA remittance collapses. If the general contractor goes out of business or switches subcontractors, the loss is total.

An underwriter reviewing the same recording without AI scoring might notice that many deposits have similar descriptions. They might not. The volume of transactions in a 90-day banking history makes manual pattern recognition unreliable, especially under production pressure. The AI catches it every time because it's analyzing every visible transaction against every other visible transaction, looking for exactly this pattern.

This is the scenario that played out at scale in the 1 Global Capital case. Concentration risk that was knowable, measurable, and preventable went undetected because the verification workflow wasn't designed to find it. The Federal Reserve's guidance on concentration risk applies the same principles to bank lending that MCA funders should be applying to merchant verification: measure it before you commit capital, not after.

Frequently Asked Questions

What is concentration risk in merchant cash advance underwriting?

Concentration risk in MCA underwriting refers to the danger that a merchant's revenue depends too heavily on a small number of customers, a single industry, or a narrow seasonal window. When a merchant's deposits are concentrated in one or two sources, losing that source can immediately eliminate their ability to make daily or weekly remittances to the funder. Standard underwriting metrics like deposit volume and average daily balance do not capture this risk, which is why dedicated concentration analysis using bank verification data is increasingly important for deals above $250,000.

How does AI detect concentration risk in bank statements?

AI detects concentration risk by extracting transaction-level data from bank verification recordings or documents, clustering deposits by source based on description text, amounts, and timing patterns, and calculating a concentration score using metrics like the Herfindahl-Hirschman Index. The models identify what percentage of total deposits come from the top one, three, and five sources, flag seasonal revenue dependencies, and generate a risk-adjusted score that underwriters can use alongside traditional metrics. This analysis requires detailed transaction visibility, which is why recorded banking sessions provide better data than summary-level bank statements.

Why is async bank verification better for concentration risk analysis?

Asynchronous bank verification produces a recorded, replayable session of the merchant navigating their live banking portal. Unlike a live verification call where the underwriter processes information in real time under pressure, an async recording can be reviewed multiple times and fed through AI analysis tools that extract transaction data, cluster deposit sources, and score concentration risk automatically. The same recording that validates the merchant's identity and transaction authenticity can simultaneously power deeper risk analytics without requiring any additional interaction with the merchant.

What HHI score is considered safe for an MCA deal?

There is no universally agreed-upon safe threshold, but general guidelines adapted from financial concentration analysis suggest that an HHI score below 0.25 indicates diversified revenue sources and lower concentration risk. Scores between 0.25 and 0.50 indicate moderate concentration that may warrant adjusted deal terms such as smaller advance amounts or shorter repayment periods. Scores above 0.50 indicate high concentration where a single source loss could materially impair the merchant's ability to repay, and deals in this range should receive enhanced manual review before funding.

Conclusion

The most expensive losses in MCA don't come from fake bank statements or synthetic identities. They come from perfectly real merchants with perfectly real revenue that happens to depend on one or two sources. When those sources disappear, so does the funder's capital.

AI-powered concentration risk scoring transforms bank verification from a binary authenticity check into a multidimensional risk assessment. The same async recording that proves a merchant's banking session is legitimate can simultaneously reveal whether their revenue structure can withstand the loss of a single customer.

For funders writing large advances, this isn't optional analysis. It's the difference between sustainable growth and catastrophic single-deal losses. Visit exactbalance.ca to see how async bank verification gives your underwriting team the recorded, AI-analyzable banking sessions that concentration risk scoring requires.