Oban MacTavish

The through-line

Everything MacTavish says traces back to a single observation he made while consulting for fintech companies during the pandemic: the data underpinning card payments is bad, and nobody seemed troubled enough by it. "I kept hearing the same problem everyone's like our data sucks," he recalls, whether the provider was "super modern one like Marquetta or Galileo or unit or you know something on the you know an old school one like an i2c" [1][2]. What confounded him was the scale of the neglect. "This is powering you know one of the largest profit pools in the entire planet financial services and all of its bad data that just is just crazy to me" [1][2]. That indignation became Spade, founded in November 2021 [1][2].

The company's trajectory tracks a shift in his own framing of the problem. What began as "a small API that would clean a transaction and give you a logo" became something that returned real business information, and then "a real Data Network that shares really valuable information with banks in real time" [1][2], an evolution from logo API to fintech infrastructure that he has described as the arc of the business [8][10]. The ambition has widened accordingly: he now positions Spade not as a card enrichment vendor but as infrastructure for money movement generally [8][10], and frames the founding logic simply, "if I don't do it someone else will" [1][2].

On why "the data sucks"

The core defect is that a bank frequently cannot tell what its own customer bought. "How is that a bank doesn't understand where we're actually spending our money like that just like makes no sense to me" [1][2]. The raw material is poor: a merchant category code may say restaurant when the business actually "sell firearms or other sort of like higher risk materials" or is a gambling site that "deserves a different sort of look from an agent" [1][2]. He has also discussed the structural roots of this, including the character limits imposed by the decades-old card standard and the way those constraints starve issuers of context [11]. His formulation of the stakes is blunt: "if you don't understand where your customers are spending you know it's sort of the old adage garbage in garbage out" [1][2]. The asset itself is not worthless, only unrefined. "We have this data it's a messy asset we can turn it into this really clean structured accurate information" [1][2].

On real data versus a pretty logo

Asked what separates Spade from incumbents like "entropy or MX person etics others that have been around for a really long time," MacTavish draws the line at provenance. Spade is "backed by real data," meaning "we're using State Registries and other information to give you verified and accurate information it's not like you know we just cleansed the name of this transaction and gave you a pretty logo" [1][2]. The network covers "99.9% of card accepting businesses in the US" [1][2], and he has described the resulting dataset on card merchants as among the richest in the country [11], delivering granular merchant, category and geolocation detail to the card ecosystem [8][10]. Firmographic data is combined with insights generated by Spade and its partners, and with whatever else an issuer wants to use in the authorization stream or after it [1][2].

On why latency changes the economics

Speed is not a feature for him, it is what converts data spend from expense to return. Spade's P99 is "under 50 milliseconds," which means "every single transaction you send to us is going to get this enhanced information" and that customers "can feel really comfortable leveraging our product in the off stream" [1][2]. That in-stream capability is what he calls the unlock: "issuers have been paying for better data for a long time but much of the time it's been for the purposes of like a user experience but there's only so much you're willing to pay for that because it's just a cost center" [1][2]. Real-time data instead touches the metrics that move a P&L: "we can actually help you stop more fraud we can help you cut fraud losses we can help you increase off rates like we're really driving really important kpis for your business" [1][2]. The pricing model follows the same logic: consumption-based, charged per transaction, with issuers "paying us you know fractions of a penny per card payment" for that improved performance [1][2].

On the use cases: fraud, authorization, attribution, analytics

MacTavish groups the applications into authorization, attribution and analytics [1][2]. Fraud prevention and authorization decisioning are the entry points, but he is emphatic that rewards are an underrated pressure point: card-linked offers are regulated, "you can't claim you're giving someone 5% cash back and then not give them 5% cash back," and inaccurate attribution becomes "a massive cost center for issuers when they're dealing with you know customer complaints" [1][2]. Better data means the payouts are right and provable [1][2]. Analytics is the growth area he sees in the existing customer base, as "every single issuer is becoming more and more focused on deeply understanding their customer base" [1][2].

The B2B side matters as much as consumer, and the examples are concrete. Real-time categorization lets business cardholders close their books faster: "we can tell you hey this is Facebook ads we can tell you this is businesses like Trucking repairs," feeding models that generate P&L and general ledger codes [1][2]. The same data enables vendor-locked, category-locked and geographically locked cards [1][2]. In logistics, working with what he identifies as the largest B2B Mastercard issuer in the United States, the target is fuel fraud: match the truck's location against the gas station, and if "the truck in the garage and you're saying you're filling up thousand bucks," send a push notification and challenge it [1][2]. His view of the addressable market is correspondingly flat: "if you issue a card you should probably be working with us to better understand where your customers spend their money" [1][2], across consumer programs, B2B cards, small business, fleet and fuel.

On letting the good guys in

The most interesting frontier, in his reading, is not fraud suppression but acceptance. Card programs "have been around for a really long time we're getting better and better at stopping fraud and fraud is always going to be a problem but the question is often these days is not just how you stop fraud but also how do you let more of the good guys in how do you make sure that the transactions that should be authorized are getting authorized" [1][2]. He puts Spade at the centre of that infrastructure, "driving increased payments acceptance rates for merchants by you know improving the ability for banks to understand what they do," and names direct merchant-to-issuer connectivity as one of the biggest waves he expects [1][2]. This is the ground on which he approaches payments performance optimization more broadly [3][4].

On trust, PII and security

His answer to data protection is architectural rather than procedural: Spade holds no personally identifiable information at all. "We require no pii to be sent to us for us to operate," and the contracts say so explicitly. "We don't need any bins or pans," nor tokenized payment information, only "non-identifying Merchant focused information that comes out of the off stream" [1][2][9]. On top of that sits SOC 2 Type 2 compliance and, in his account, "hundreds and hundreds of questionnaires pent tests" for every large enterprise engagement [1][2][9]. He ties this to the industry's founding condition: "banks are built on trust issuers are really built on trust and you want to make sure that you know the data you're leveraging is high trust" [1][2]. Hence the operating rule, "security has to be top of mind at all times because you know this is you don't want to become a fraud vector somebody" [1][2].

On AI, and on data being usable before it is intelligent

He treats AI as inseparable from the data conversation but subordinate to it. Banks and fintechs increasingly recognise that the opportunity in their data "is massive," but "the better question is like how do you make sure that data is usable effective valuable clean structured Etc" [1][2]. The opportunity he claims is upstream of the models: ensuring the data large issuers feed into them is accurate, verified and free of PII [1][2].

On new rails, crypto and the price of protections

MacTavish is openly aware that he is "selling my own book," and says so before arguing that "I don't think cards are ever going away" [1][2]. His reasoning is about hidden infrastructure rather than nostalgia: consumers take chargeback rights and the bank's responsibility for clearing dollars for granted, particularly on credit [1][2]. "Every time a new form of payment appears I think people forget the enormous amount of infrastructure that exists around older forms of payments," he says, so comparisons that make stablecoins or RTP look cheap are comparing a full product against an unfinished one, because "many of these protections or pieces of infrastructure that your bank and broadly you or the merchant are paying for don't exist yet" [1][2]. He therefore expects "a convergence of the price of many of these things," with prices on newer rails likely rising from very low starting points, already visible in the baked-in pricing of crypto products emerging from PSPs [1][2]. He calls that healthy: "the more competition we have for our dollars and where we process is you know the better," and notes that processing is "an incredibly efficient industry" where middlemen are under intense pressure to cut prices [1][2]. Spade's own posture follows the same principle that "we should understand more about our payments data" regardless of rail: it is not touching crypto, but is moving into RTP, ACH and wires [1][2].

On building an infrastructure company from nothing

His account of starting Spade is unglamorous. He had just moved in with his then girlfriend, now wife, into their first apartment, had "very little money to my name," and was doing fintech consulting on business development, pitch decks and messaging to pay rent when the pattern in customer complaints emerged [1][2]. The founding impulse predates the idea: he fell in love with Silicon Valley during college summers, and the lesson he took was that "you can just wake up and say I'm gonna try to build a company and if you're lucky enough and if you sometimes put in the work you can end up in a position where you're really you know changing an industry" [1][2]. He describes his own path as scrappy, with the difficulties of his first venture feeding directly into Spade [8][10].

He is candid about the specific hardness of selling infrastructure early, including overcoming skepticism as an early-stage infrastructure company, and about the need to keep iterating on the product against customer pain points and feedback rather than a fixed roadmap [8][10]. That extended to go-to-market, where he has discussed adapting sales strategies toward mid-market and enterprise buyers, and to the wider question of fintech overvaluation [8][10]. The results he points to are asymmetric: roughly 18 people after a Series A raised at the end of 2023, serving three publicly traded companies, Stripe and issuers "processing you know over hundreds of billions in card payments every single year," with a US focus reflecting US data coverage and expansion into Canada and other geographies underway [1][2][9]. Outside work, he cooks, in a domestic division of labour where his wife bakes, and he misses the era when he had hours to spend on molecular gastronomy and long braises [1][9].

Takeaways

• The founding insight came from repetition, not analysis: consulting clients across Marqeta, Galileo, unit and i2c all said the same thing, "our data sucks," and nobody had fixed it [1][2].
• Enrichment only pays for itself when it is real time. Spade's P99 is under 50 milliseconds so the data can sit inside the authorization stream, turning a user-experience cost center into fraud losses avoided and higher approval rates [1][2].
• The differentiator he claims is provenance: verified information from state registries covering 99.9% of US card-accepting businesses, as against vendors that "just cleansed the name of this transaction and gave you a pretty logo" [1][2].
• The next frontier in card risk is approval, not decline: "how do you let more of the good guys in," with direct merchant-to-issuer connectivity as the wave to watch [1][2].
• Spade holds no PII by design, needs no BINs, PANs or tokenized payment data, and layers SOC 2 Type 2 on top, because "banks are built on trust" [1][2][9].
• New rails look cheap because their protections do not exist yet; he expects prices on stablecoins, RTP and similar to converge upward, and considers the resulting competition good for the industry [1][2].
• B2B use cases are as important as consumer: real-time categorization that speeds up book closing, vendor, category and geographic card locks, and location-matched fuel fraud detection for logistics fleets [1][2].
• Pricing is consumption-based at fractions of a penny per transaction, aligning cost with the volume of decisions improved [1][2].