Working together at Uber’s risk team, Yifu Diao and Ming Fang came to realize that with the hypergrowth the company was seeing came fraud challenges. Diao’s first project at the company after starting in 2014 was building a rules engine called Mastermind to fight fraud. That engine, he says, allowed risk analysts to build and rollout rules without the help of engineers.
“While it was originally built for fraud, it came to be used for other cases such as safety and customer support also,” Diao said. “It was making 10,000 decisions per second by the time I left.”
When Fang joined Uber’s risk team in 2016, there was a need to use Mastermind at dispatch time — when drivers and riders are matched. However, dispatch was one of Uber’s most critical systems, and it had a much bigger scale than previous use cases while having a very stringent latency requirement.
“Ming [Fang] did enormous optimizations for Mastermind, which enabled this new use case and brought down fraud significantly,” Diao recalls. Mastermind is now used by hundreds of analysts and ops, and it makes real-time decisions throughout hundreds of user touch points.
Diao spent his last year at Uber working on a lending product, which also made use of Mastermind for underwriting. He then joined a credit card startup and realized that product also needed a rules engine to manage fraud, credit and compliance risks. However, that startup didn’t have resources like Uber to build such an engine in house.
“What’s more, explainability is important for fintech companies so that they can show regulators how decisions are made,” Diao said. “It became apparent to me that there was a need for a rules engine as a service.”
He immediately thought of Fang, who by that time was working at Google leading its Cloud AI Feature Store. The pair had been discussing a few startup ideas in recent years and they decided to team up to form Sperta this June (Sperta means “expert” in Italian) and build their own rules engine as a service. Its mission is to help financial services and tech companies automate decisions and manage fraud, credit and compliance risks.
Many of the current rules engine offerings tout a no-code UI to build rules with a few mouse clicks the duo believes that won’t work out well “when the logic gets more complex,” Diao said.
Sperta’s main differentiator from others in the space (other players who have recently raised funding include Alloy and Unit21), according to Diao, is its expression language.
“Our target users are analysts and data scientists, who already know SQL,” Diao said. “So we decided to build an expression language with similar syntax to SQL. We had great success with this approach in Mastermind. It made it possible for analysts to onboard in just one week.”
Rules won’t perform well if they are not tested, explains Diao. And high false positives can have a real negative impact on growth. False positives can mean a loss of customers for financial institutions and companies.
“We are keenly aware of this,” Diao said. “That’s why we are enabling rule unit tests for sanity checks, backtesting for measuring the rule performance and percentage rollout for safely applying rule changes.”
To solve its customers’ “end-to-end” needs, Sperta is building a risk decision platform with the rules engine at its core. Sperta also integrates with data vendors, and it also allows analysts to transform features obtained from those vendors.
Just two months after its formation, Sperta raised $3 million in seed funding in a round co-led by Kindred Ventures and Uncork Capital that included participation from some angel investors. It is just now talking about publicly.
“Customers can bring their own models too, and Sperta provides a clean interface for integration,” Diao said. “If decisions can’t be made automatically, the cases can be sent to Sperta’s case management tool. While models give us probabilistic predictions, rules give us explainable decisions and deterministic actions. We are really excited about helping the internet make better decisions faster.”
Sperta, Fang said, makes it easier for customers to integrate models with its UI.
“We can structuralize the data they want to use to detect risk and make decisions,” he told TechCrunch. “We can structuralize the way they use data to make sure their decision making is safe and cleaner.”
For fintechs especially, a rules engine is very important at the onboarding stage. Fraud is top of mind for all financial institutions, fintech or traditional. And with more transactions conducted online than ever, the possibility of fraud is also higher than ever.
So naturally, fintechs and financial institutions in general are target customers.
Competitors, Diao said, charge on average about $1 a decision. While he wouldn’t get specific, Diao said Sperta will charge less than that.
The company plans to use its new capital mostly toward hiring. It’s still developing its product but hopes to have an MVP by year’s end.
For Kindred Ventures’ Kanyi Maqubela, Kindred Ventures believes Diao and Fang’s work at Uber was “groundbreaking” in the area of fraud prevention.
“A generalized, powerful version of a decision engine is sorely needed in today’s market, where software companies are making millions of underwriting decisions every day,” he wrote via email. “Software companies need a solution that is sufficiently sensitive to allow both customization, yet maintain compliance… Successful architecture is flexible enough to be modular, but secure enough to maintain compliance and ease across use cases. Sperta has already achieved this, within only months of being in market.”
Andy McLoughlin, Uncork Capital managing partner, agrees that the founders’ insights from having built a rules engine at Uber give them “an immediate advantage.”
“In our diligence we heard repeated dissatisfaction at existing solutions which aren’t able to offer some of the more advanced features that Sperta delivers in v1.0,” he wrote via email. “The competitors have done a good job at priming the market but we see a huge opportunity to deliver the correct solution.”