False declines are the target of a new partnership between Riskified, the ecommerce fraud and risk intelligence company, and Marqeta, the modern card issuing platform. The two announced on 5 August that card issuers on the Marqeta platform will gain access to Riskified pre-authorization risk intelligence. Nothing flashy attaches to this one. No funding round, no product launch with a splashy name. Yet false declines, meaning legitimate transactions rejected because a fraud model flagged them incorrectly, are an expensive and underdiscussed problem in payments. This is a direct attempt to chip away at it.

Fintechbits analysis

Why False Declines Happen in the First Place

The mechanics are well understood inside the industry even if consumers rarely think about them. Issuers set fraud thresholds conservatively because the cost of approving a fraudulent transaction is immediate and measurable, arriving as chargebacks, liability and reputational damage. Meanwhile the cost of turning away a good customer is diffuse and surfaces later, as reduced spend, card abandonment, or a switch to a competitor card. Industry estimates have long put false declines at several times the size of card fraud losses. J.P. Morgan research frames the gap as fraud accounting for roughly 7% of the total cost of fraud against 19% for false positives.

Scale is what makes that asymmetry so costly. Research from PYMNTS Intelligence and Nuvei put $157 billion of US ecommerce sales at risk from false declines in 2023, with $81 billion lost even after shoppers tried to complete the purchase through other payment methods. Issuers have therefore been over-rejecting legitimate spend for years, because the incentives inside any single transaction pushed them that way.

What the Integration Does About False Declines

Pairing Riskified ecommerce fraud data with the Marqeta issuing and authorization layer is a logical way to attack that mismatch. It gives the authorization decision more context about the merchant and the transaction pattern than an issuer working from card network data alone would hold. The announcement describes precisely that. Riskified supplies enriched intelligence drawn from its global merchant transaction network, giving issuers additional context on an order before it reaches authorization.

That timing detail matters more than anything else in the release. The signal lands pre-authorization rather than post-decline, feeding Marqeta Real-Time Decisioning and the AI-powered predictive risk score already running inside it. Consequently the architecture described is the deep version rather than the shallow one, which is the difference between reducing false declines and merely documenting them after the fact. Fraud models of this kind increasingly depend on AI-driven pattern recognition rather than static rules.

Why False Declines Matter to Marqeta Strategically

For Marqeta, this partnership matters beyond the immediate fraud benefit. Card issuing processing has grown commoditized as more platforms enter the space, and raw processing speed or uptime is no longer much of a differentiator when several vendors offer comparable reliability. Marqeta processed nearly $400 billion in annual payments volume during 2025 and holds certification in more than 40 countries. Scale is therefore real, though scale alone does not defend a position.

Data partnerships that measurably lift an issuer approval rate make a more durable moat. Switching away from a partner that is demonstrably reducing your false declines carries a quantifiable cost rather than a migration headache. That logic runs through much of the current embedded finance market, where differentiation has moved from infrastructure to the data layered on top of it.

The False Declines Numbers That Already Exist

Whether the partnership moves the needle still turns on execution, although less of it is unknown than the announcement first suggests. Riskified has published results from earlier issuer integrations. Over a 30-day period, a top-tier US card issuer drawing on the Riskified merchant network lifted authorization rates by 5.9% at a ticketing merchant, 1.4% at a gaming merchant and 1.6% at an online retailer, while reporting a 25% cut in false declines with certain Riskified merchants. Athletic apparel retailer Lorna Jane watched its bank authorization rate climb from 82% to 95%, alongside a drop in chargebacks of more than 90%.

Those figures come from other deployments rather than this one. So the benchmark exists and the mechanism has worked elsewhere, which moves the open question away from whether the approach functions and toward whether it holds at Marqeta scale, across a portfolio of card programs instead of a handful of merchant relationships. Applied AI in fintech tends to show exactly this gap between pilot results and portfolio results.

The number to watch is therefore the false declines reduction rate Marqeta issuers report once this is live. Neither company has committed to publishing it. That single metric separates a real product improvement from a partnership announcement with no measurable outcome attached.

Fintechbits covers card issuing, fraud prevention and payments infrastructure. Nothing here constitutes financial or investment advice. All analysis represents the editorial views of Fintechbits.