TAMID club brings co-founder of Israeli startup
Riskified founder and CTO, Assaf Feldman, shared insight into the Israeli startup ecosystem.
TAMID Group at Brandeis, a “business organization that develops professional skills through hands-on interaction with the Israeli economy” as described on the club’s website, virtually hosted a guest speaker, Assaf Feldman, on Oct. 4. Feldman is the co-founder and Chief Technology Officer of Israeli security company Riskified. Feldman talked about his life and the circumstances that prompted him to found the company, lending insight into the realities of the Israeli entrepreneurial ecosystem.
Feldman had a relatively late start to programming. After a frustrating experience at age 10 with a lagging Massachusetts Institute of Technology coding program that moved a robotic turtle, he “never touched a computer” until his mid twenties. Instead, he explored a variety of subjects in high school and at university, particularly philosophy, math and film. He recounted viewing a lecture in New York City’s Lincoln Center for the Performing Arts about the ways in which video can be combined with computer-generated graphics to track actions such as the movement of a dancer, shaping film in new ways. This lecture inspired Feldman to pursue studies in computer science, seeing computer graphics as the future of film.
Eventually, “I realized I would not be the new … hope of Israeli cinema,” Feldman admitted. Around the time he graduated from MIT, the internet boom was in full swing, and Feldman joined a startup of friends from film school. They were working on a proto-YouTube style online system that would allow users to share and comment on videos online. However, while working on these ideas, computer technology was not yet advanced enough to support what would eventually become the domain of YouTube. Feldman remarked that in hindsight, the failure of their ideas was a good example of the vital importance of proper timing for entrepreneurial success. Feldman moved with his fellow entrepreneurs to New York, but quickly had to extricate himself with the burst of the internet bubble.
Uncertain about his next move, Feldman applied to MIT’s Media Lab for graduate school on a whim and got in, getting in to his surprise. At the Media Lab, he said that he fell in love with machine learning, “the idea that you're teaching a machine to do something and it does it at scale.” After graduate school, he applied his new skills to a series of startups that would consistently not get enough traction within the market due to the lack of a market fit.
At this point, Feldman had married and returned to Israel for his wife’s career in the film industry. Continuing to work at startups, he struck up a friendship with coworker Ido Gal, who had his own idea for a startup for e-commerce security. Feldman remarked that this was a case in which the timing worked out perfectly for him. Just as his wife’s career in Israeli film was taking off, Feldman started Riskified with Gal in 2013, seeing that the market fit was perfect for their ideas.
Feldman then shared about Riskified in greater detail. According to Feldman, Riskified’s main function is fraud detection for e-commerce. He explained that “every e-commerce system basically screens the orders to see if the order is fraudulent or legitimate,” or in other words, if the “order is done by someone who doesn’t own the payment information.” If they fail to catch an unauthorized payment, then the liability is on the merchant, who must pay a chargeback fee in compensation.
“We realized though, early into the game, that the problem is not the fraud. The problem is the fear of fraud and the lost money due to being overcautious about it and not accurate enough,” Feldman said. Because of this fear, “merchants were declining orders and letting go of legitimate users.”
While previous fraud companies generally offered tools for companies to deal with the fraud by themselves, Riskified promised to take on the entire burden. Merchants could accept risky orders with the assurance that Riskified would screen the order and provide compensation in case of mistakes.
The “second lesson I learned about execution was to try to bring out a service as soon as possible,” Feldman said. Although their security program took some time to be completely finalized, Riskified had a product out in three months. Because of the nature of machine learning, they needed data to train their model. So in the beginning, the founding members had to start out with “manual work,” Feldman explained, showing an image of the members hard at work in a cramped apartment.
Nevertheless, within four months, Riskified was able to train a stable screening model, “and from there it just grew and grew,” Feldman said. The company went from handling the orders of companies on Shopify worth $10 million to merchants worth $100 million, and soon enough Riskified was managing the orders of multinational companies such as Footlocker, Macy's and Wayfair.
“We really caught e-commerce when it was booming,” Feldman remarked. According to him, Riskified started in an era when companies were starting to take larger risks in order to stay competitive with Amazon, and thus required the use of a reliable security system. They were one of the first in the field, giving them a high degree of leverage for business activities such as mergers and acquisitions.
After Feldman’s lecture, he answered several questions about working as an entrepreneur. He said that in searching for a business idea, he finds it essential that “the passion in the early days should come from the passion to build a product,” and it is “best to think of something you have experience in … that will make sure you will build something that is to the point.” Good business ideas and startups are formed when “you solve a problem for someone and you build a good product,” Feldman said. He explained that even in the early days of Riskified, his commitment to the startup was such that the possibility of liquidation never crossed his head.
Feldman said of his experience at Riskified that “for me this whole journey was one big learning experience of solving one problem after another.” Every stage has a different set of problems, he explained. According to Feldman, in the beginning it was not clear that they could build a sufficiently accurate machine learning model to be financially sustainable — in other words, a model that was wrong once in a thousand decisions. After the stable model was developed and Riskified grew larger, there were thoughts of scaling up and establishing a proper company hierarchy.
In response to a question of his motivation to work in startups, Feldman explained that “startups are a great way to learn a lot of new things, [because] you can wear many hats and you can really figure out what you like the most. There’s also the side of being close to the product, so you’re working on the most important stuff.” He added that when he was at MIT, he worked in Nokia research for a summer, and this experience convinced him that he did not want to work for a large corporation again.
Regarding his relationship with the other Riskified executives, he explained that because of the need for founders to perform many tasks in the beginning years, their eventual position in the company is “what they make out of it.” For a company to progress, it is necessary to bring in new members at every level, even at the top executive level. Nevertheless, Feldman stressed the importance of maintaining a solid basis of teamwork for a company to run smoothly, and argued that this teamwork must start from the top.
— Editor's note: Jacklyn Goloborodsky is an editor on the Justice and a member of TAMID club at Brandeis. She did not contribute to the reporting or editing of this story.