Spade

Take the AI away and what remains is merchant category codes, precisely the legacy standard the product exists to replace. The company markets against those codes by delivering verified categories independent of them. A proprietary matching engine is driven by agents that continuously scan the web and external sources to fill metadata gaps and eliminate duplicate records. The company describes the result as self reinforcing, growing more capable with every transaction processed. Resolving an arbitrary payment string to a specific business with precise geolocation is inference, not lookup.

The data feeds automated decisions by design. Authorisation decisioning is the first named workflow and the sub 40 millisecond tail latency exists precisely so enrichment can sit inside a live authorisation path, where no human reviews anything, and the chief executive frames the ambition as banks moving toward fully automated agentic workflows built on this data.

Nothing describes what happens when a match is uncertain: no confidence score is mentioned, no fallback behaviour is defined for an unresolved merchant, and no guidance is published on how a customer should treat a low confidence enrichment inside a decline decision.

Three performance figures are published, and one of them is chosen in a way that reveals seriousness. Coverage is stated at 99.9 percent of United States and Canadian merchants with matching accuracy above 99 percent, both scoped to a defined universe rather than asserted generally. Latency is published as a 99th percentile figure under 40 milliseconds rather than as an average.

That is the honest measure, because it describes the worst case customers actually experience inside an authorization path rather than a flattering mean. For a data layer whose entire value is correctness and speed, publishing both scoped and tail metrics is the right disclosure. No confusion matrix or error breakdown accompanies it.

Five customers are named, and they are substantial. They are a global payments company, a corporate payments group, a business banking provider, a banking as a service platform and a rewards focused card program, described as using the data to authorize more transactions, prevent fraud and build new features. A $40 million Series B closed in March 2026 with a specialist fintech growth investor.

The performance figures are stated precisely, at 99.9 percent coverage of United States and Canadian merchants, over 99 percent matching accuracy, and tail latency under 40 milliseconds. Founded in 2021, the company describes customers arriving for authorization decisioning, fraud prevention, rewards attribution and analytics. It notes these workflows were historically considered too critical to entrust to something resembling data cleaning.

The flywheel is disclosed as the central advantage, and it is exactly the stewardship concern. The system is described as self reinforcing, becoming more intelligent with every transaction processed, which means each customer's volume improves the database delivered to all the others.

Those others include direct and adjacent competitors, since the named customer list spans a global payments company, a business bank, a banking as a service provider and a corporate payments group operating in overlapping markets. The company states this openly as a strength. Nothing describes what is contributed, whether a customer can decline, or what happens to learning derived from a departing customer's traffic.

The payload is a complete record of where individuals spend their money, and no data protection agreement, retention schedule, subprocessor list or deletion commitment is published. Enrichment establishes exactly which business a person transacted with and where, at precise geolocation. The company describes the objective as detailed, verified and structured consumer behavior.

Downstream uses named by the company include behavioral segmentation and loan targeting, so purchase histories shape what individuals are offered. Nothing published states what is retained, whether enriched consumer transactions persist beyond the query, or what a person is told.

There is no public attestation, certification, trust center or enumerated framework. A global payments company and several banking platforms have completed vendor assessment and placed this service inside their authorization flows, which is diligence at a demanding standard. For a vendor handling consumer transaction data in real time at that position in the stack, a published control set is what every subsequent enterprise buyer will request.

No supervisor, statute or instrument is named. Two named use cases sit close to regulated territory: behavioural segmentation and loan targeting derived from a consumer's transaction history engage rules on the use and sharing of financial information, and where such data informs credit offers the consumer reporting framework becomes relevant. Nothing published addresses either, nor any payment network or data handling standard despite the product sitting inside card authorisation flows.

Two exposures matter. The first follows from the stated use cases. Enriched purchase history drives loan targeting and behavioral segmentation, so what a person buys shapes what credit and offers they are shown, through an enrichment layer they never see and cannot correct. The second is a coverage inversion. At 99.9 percent merchant coverage, a residual tenth of a percent goes unresolved, and unresolved merchants will disproportionately be small, new, informal or rural businesses.

The customers of exactly those merchants are the ones whose transactions look unrecognizable to a fraud engine and are likeliest to be wrongly declined. No breakdown of coverage or accuracy by merchant size or type is published.

Beyond the published accuracy figures there is no guarantee, indemnity or falsifiable commitment, and those figures are marketing claims rather than warranted terms. The institutional customer can measure performance against its own dispute volumes, which the company identifies as the problem poor enrichment causes. The consumer has no route at all. A misattributed transaction shows the wrong business on a statement, which is what generates the dispute in the first place. Nothing describes how a wrong merchant match is corrected in the database once identified.

Sources are described only as the web and external data sources scanned continuously by agents to fill metadata gaps, with the resulting database described as proprietary and ground truth. That leaves the provenance of the merchant data unstated. It matters because the accuracy claim rests entirely on what the underlying references contain and how they are licensed. No model provider is named for the matching or agent components, and no subprocessor list or hosting arrangement is published.

Delivery is through interface infrastructure engineered for the most demanding environments. The evidence of integration depth is the customer list rather than a connector catalog, since a global payments company and a banking as a service platform have embedded this inside their own transaction paths. Coverage across card, automated clearing house and wire rails means one integration serves all three. No named platform, processor or core system integration is published, and there is no public developer documentation, which interface first products usually provide.

Where the platform runs is not published, with no hosting provider, region selection, residency commitment or private deployment option named. The North American footprint simplifies exposure. The platform nonetheless processes consumer transaction records inside live authorization paths for regulated institutions, whose own examiners would expect the processing arrangement to be documented.

No pricing, packaging or basis of charge is published. The company is explicitly moving from an interface priced on enrichment volume toward a platform supporting workflows. A commercial shift like that would ordinarily change how customers are charged, and nothing describes either the old model or the new one. Whether charge falls per transaction enriched, per workflow or as a platform fee is undescribed.

Buyer types span banks, fintechs, payment companies, banking as a service platforms, corporate payments providers and card programmes, which is broad within the payments ecosystem, and rail coverage is complete across card programmes, automated clearing house and wire transfers rather than cards alone. Use cases now extend well beyond enrichment into authorisation decisioning, fraud flagging, rewards attribution, notification triggers, behavioural segmentation and loan targeting. The clear limit is geographic and stated plainly: coverage is United States and Canadian merchants, so this is a North American product.