Martin Sweeney and the problem that wouldn’t wait

A taxi booking is a poor place to ask for patience. Someone wants to get somewhere. A driver wants a fare. Between them sits a payment, and perhaps a stolen card. At Hailo, where Martin Sweeney was a founding engineer, that last possibility became a problem worth leaving the company to solve.

Existing fraud tools did not offer the speed and accuracy his team wanted. The gap gave Sweeney and his colleagues a business: Ravelin, founded in 2014. Its starting point was pleasingly practical. They had encountered a problem at work, looked for a satisfactory answer, and decided to build one.

The resulting career has kept returning to a difficult question: how do you stop a dishonest customer without making an honest one wish they had stayed at home? Sweeney now leads a company owned by Worldpay. The taxi booking has become a much larger argument about whom the internet should trust.

The engineer on the merchant’s side

Sweeney grew up near Liverpool and studied at the University of Leeds. His background combines physics and software; before Ravelin, his work included Hub Culture and Hailo. The sequence matters because he entered fraud prevention through the business accepting payments. He knew what it felt like to need an answer while a customer was still waiting.

Hailo belonged to the early smartphone taxi market, when booking and paying inside an app was still a novel proposition. Sweeney remembers competing with Uber when it was called UberCab. The technology changed the pace of commerce. A payment decision became part of a live service, rather than an administrative task conveniently left until later.

In his telling, Ravelin’s inheritance from Hailo is the experience of having been a merchant. A prediction has to help someone run a business. Blocking a sale can be a mistake, too. That gives the engineer a rather unforgiving brief: make the right decision, make it quickly, and let the customer get on with the journey.

An impatient founder, a patient business

Hailo became an early anchor customer. Passion Capital supplied institutional backing, with a term sheet arriving on Christmas Eve. For a young company, that is a useful sort of seasonal correspondence.

Sweeney has described wanting answers quickly while learning that enterprise customers move through procurement and integration at their own pace. The lesson extended to the product: payments security required care from the beginning. His early code was still part of the system years later.

Even the travel had its little jokes. He recalled going from a sweltering São Paulo to a bitterly cold New York, with luggage apparently better suited to the former. The business crossed climates rather more comfortably than its founder’s wardrobe.

His advice to prospective founders was equally earthbound: examine your motives and have a lasting interest in the work. A quick fortune was a poor reason to begin. There is an appealing consistency here. The person impatient for results had chosen a business that would keep charging him tuition in patience.

“What I really get my kick from is building something that solves a problem.”

Martin Sweeney

A small fortification with a growing perimeter

The company’s name is an architectural clue. A ravelin is a defensive fortification associated with a fortress; Ravelin’s starfort logo incorporates several. It is a compact metaphor for a business trying to make transactions defensible while commerce continues around them.

Sweeney built it with Nick Lally, Mairtin O’Riada and Leonard Austin. The founding group’s work expanded from payment fraud into account security, refund abuse, marketplace fraud and authentication. Each addition widened the territory under inspection. A questionable payment might be the final visible act in a much longer sequence.

In the foreword to Ravelin’s 2025 fraud report, Sweeney described the original premise as understanding people and their intentions. The useful unit of investigation was the customer in context. A transaction on its own could leave too much of the story outside the frame.

That also explains the company’s habit of asking merchants about their experience. Sweeney presents listening as part of the work. Fraud prevention requires a model of human behaviour, and humans have a tiresome habit of revising the plot.

When the numbers acquired weight

By 2023, Ravelin reported ten billion scores across its products, including two billion orders. Those are different measures: an order is a purchase, while scoring can happen at other points in a customer’s relationship with a merchant. Confusing the two would make an already large number tell the wrong story.

The company’s results also described more than four million authenticated payments across 13 merchants, following none in 2022. Authentication had become another part of the business. The work was moving beyond identifying questionable activity into helping decide how a payment should proceed.

For Sweeney, that expansion made the original engineering problem more complicated. The business had to keep serving individual merchants while operating through larger payments relationships. Its 2023 results named partnerships with Elavon, Fiserv and Mastercard. A tool first shaped by on-demand commerce was becoming part of other companies’ infrastructure.

The following May, Ravelin reported that its key account base had doubled during 2023. Its examples ranged from Glovo to LNER and KFC. The taxi-app problem had found relatives on railway journeys, delivery platforms and promotional offers.

The customer who knows the loophole

Some of Sweeney’s more revealing observations concern people who already have a relationship with the retailer. In January 2024, Ravelin published research finding that 37% of consumers considered taking advantage of gaps in retailers’ policies legitimate. Sweeney’s reaction was that some attitudes toward this behaviour were surprisingly brazen.

That is awkward territory for anyone selling things. Retailers want customers to use promotions, return unsuitable purchases and ask for help. The same systems can be exploited. A generous policy has to survive contact with someone who reads it as a set of instructions for extracting money.

Sweeney argued for approaches that adapt to behaviour rather than treating everyone alike. The appeal is clear enough. A customer who deserves a refund should receive one without being made to feel like a suspect. A customer inventing a grievance should present a different problem to the system.

His professional subject is full of such uncomfortable distinctions. The familiar customer can create risk; the unfamiliar customer can be perfectly legitimate. A tidy blacklist has limited conversational skills.

Worldpay buys the company. The founders keep working.

On 4 February 2025, Worldpay announced an agreement to acquire Ravelin. Sweeney pointed to the buyer’s reach: approximately $2.5 trillion in payments volume and more than 50 billion transactions during 2024. Those were Worldpay’s figures, and their relevance was access to scale.

Later that month, he announced completion and said he, Lally and O’Riada would continue leading the company day to day. He also emphasised autonomy. The deal brought a larger owner to a business whose value depended partly on the team continuing to do its work.

His first-anniversary account made that concern measurable. Sweeney reported 93% staff retention, headcount growth of 16%, and an employee engagement score rising from 82% to 85%. He also said the five largest merchants had renewed. These were his reported results, rather than independent verdicts on the acquisition.

The choice of measures is revealing. Alongside selling more, he was discussing whether colleagues stayed and customers continued to trust the company. A deal has a completion date. The relationships it inherits require more maintenance.

The programmer away from the payments slide

There are smaller windows into Sweeney’s interests. His public GitHub profile lists Edinburgh and includes a Slack bot for one-to-one meetings, an audio-trimming command-line utility and a project described as a state engine for the home. They are modest, specific things: software with a job to do.

His speaker biography lists fishing, cricket and rock climbing among his interests, alongside time with his three children. The combination supplies a useful correction to an otherwise payments-heavy portrait. There are occasions, apparently, when the desired outcome involves a ball, a rock face or a fish.

His professional biography returns to the pleasure of making something useful. That is a less ornamental account of motivation than the usual founder mythology. It also fits the public evidence: code, a problem encountered inside a taxi business, and years spent extending the answer. The continuity is more interesting than any attempt to make the career look inevitable.

Now the shopper may be software

In May 2026, Sweeney was addressing another change in how people buy: AI agents shopping on their behalf. Ravelin’s research found 44% of surveyed enterprise merchants already integrating agentic protocols, with another 32% planning to follow within six months. Only 29% felt very prepared for the associated issues.

He warned that changes in online commerce have repeatedly opened opportunities for fraud. The old questions had acquired new participants. A merchant might need to understand the customer, the software acting for them and the credentials authorising the purchase.

June brought another warning. Ravelin’s survey of 1,504 fraud and payments professionals across ten countries found 64% had experienced AI-enabled fraud or abuse during the previous year. Sweeney argued for understanding behaviour across purchases and returns, rather than looking narrowly at payment events.

There is a recognisable thread running back to Hailo. Make buying easier, and someone will test the opening. Sweeney’s work is to help the business answer that test while keeping the service usable. Somewhere beyond the models and acquisition announcements, an honest customer is still waiting to go somewhere.