Plaid LendScore 2 is a cash flow credit model that the San Francisco company says predicts a borrower’s ability to repay 42% better than traditional credit data alone. Plaid announced it on 6 October 2026 as part of its annual Fall Product Release, alongside LendScore Arc, an Instant Link consent feature, a fraud foundation model inside Plaid Protect, and a payment risk upgrade to Signal and Guaranteed Payments.

What does Plaid LendScore 2 add to cash flow underwriting?

Cash flow underwriting reads how a person earns, spends and pays bills, which matters for the millions of US adults whom Plaid says fall outside the traditional credit lens. Plaid LendScore 2 is the new core model in that approach, and Plaid built specialised versions on top of it. The auto version cut delinquency 26% for deep-subprime borrowers with approvals held steady. The home lending version approved 6.3% more borrowers while risk stayed unchanged. A third version targets short-term lending.

The companion feature is Instant Link. A borrower who has connected accounts to Plaid’s Consumer Reporting Agency can consent to share cash flow insights with lenders on later applications, and lenders receive them in under two seconds. It is also a distribution move: once a consumer has consented, the data travels with them from lender to lender through Plaid’s network, which reaches more than 12,000 institutions in the US, Canada, the UK and Europe, according to Plaid.

How LendScore Arc applies a transformer model to credit risk

LendScore Arc is described as Plaid’s first transformer-based credit risk score. It reads the sequence and timing of a borrower’s transactions instead of summary totals. In early testing Plaid reports a 20% lift on deep subprime borrowers and 24% on superprime borrowers over its own core model, and Plaid calls it its best-performing credit model.

The choice of architecture matters because Affirm described a transformer-based underwriting model in September, which Fintechbits covered in Affirm’s transformer underwriting model. Two lenders and data providers reaching for sequence models within weeks of each other suggests the useful signal sits in the ordering of transactions, and that bureau-style snapshots are losing ground as the default input.

How Plaid’s fraud and payment models extend the same foundation

Plaid applied the same sequence approach outside credit. A fraud foundation model powers Plaid Protect and strengthens the Trust Index, its scoring framework. Plaid says the model draws on hundreds of millions of data points from its network and showed up to 40% relative improvement over previous baselines in internal evaluations.

For payments, the sequential model feeds Signal, Plaid’s ACH payment risk score. In testing it let Signal avert 26% more ACH returns with no rise in false flags. Guaranteed Payments uses that sharper read to offer delayed release and partial guarantees instead of a single approve-or-decline answer.

What Plaid’s AI release leaves unproven

The headline figures are not comparable with each other. The 42% for Plaid LendScore 2 is measured against traditional credit data alone. The 20% and 24% for LendScore Arc are measured against Plaid’s own core model. The fraud figure is a relative improvement over undefined previous baselines, expressed as “up to”. The ACH figure compares against Signal before the upgrade. That is four comparisons against four baselines, so no single result tells a lender how much better its own book would perform.

Several gaps matter to a buyer. Plaid does not say how many borrowers were in each test, which lenders supplied the outcomes, or what the base approval rate was behind the 6.3% figure. A 26% delinquency cut among deep-subprime applicants could be a large change in absolute terms or a small one, depending on a starting rate the release does not give. The fraud result carries no false-positive rate, which is the number a risk team checks first. All results are Plaid’s own evaluations, and the release does not mention independent validation.

Fintechbits analysis

Why Plaid LendScore 2 matters for lenders outside the bureau system

My reading is that the model scores matter less than the data position behind them. Plaid says more than 9,000 companies use its products, and the network gives it transaction histories that a bureau cannot see. Plaid LendScore 2 turns that into a product that a lender can drop into an existing decision flow, and Instant Link reduces the friction that has kept cash flow data at the margins of consumer credit.

The sharper claim is the home lending result. Mortgage and auto lending have leaned almost entirely on traditional credit data, and Plaid is arguing that cash flow adds approvals without adding risk. If that holds under a lender’s own back-testing, regulated lenders gain a reason to run Plaid LendScore 2 beside the bureau score. If it does not, the models remain a tool for the subprime and thin-file segments where cash flow data already has a track record. For related background on the data layer underneath, see Fintechbits’ explainer, what is open banking.

What to Watch Next

The test that will settle this is a named lender publishing a delinquency or approval result for Plaid LendScore 2 on its own portfolio, ideally in auto or home lending where Plaid claims its specialised gains. A second marker is whether LendScore Arc leaves early testing, since Plaid gave no release date for it. Plaid’s Business Wire release, linked here, is the primary record of the claims.

Questions and answers

What is Plaid LendScore 2?

Plaid LendScore 2 is Plaid’s new core credit risk model for cash flow underwriting. Plaid says it predicts whether a borrower will repay 42% better than traditional credit data alone.

What results did Plaid report for the auto and home lending versions?

In Plaid’s testing, the auto version cut delinquency 26% for deep-subprime applicants while approvals stayed flat. The home lending version approved 6.3% more borrowers at unchanged risk.

What is LendScore Arc?

LendScore Arc is Plaid’s first transformer-based credit risk score, built on its sequential foundation model. In early testing Plaid reports a 20% lift on deep subprime borrowers and 24% on superprime borrowers over its core model.

How fast does Plaid Instant Link deliver cash flow insights to lenders?

Plaid says lenders receive cash flow insights in under two seconds once a borrower who connected accounts to Plaid Consumer Reporting Agency has consented to share them.