Customers usually ask how our solution is different from the competitors’. When I talk about TCP/IP stack, SOCKS(4/5) proxy detection, and verification principles it’s hard to make the sale. Apparently, everyone has the best technology, and hearing me talk about the internet protocol suite is as much fun as a root canal.

This forced me to better articulate why Riskified is different.

We’re different because of our unique pricing structure:

  • A merchant receives a transaction and decides to decline it due to risk-related reasons
  • Instead of declining the transaction, he goes ahead and submits it to Riskified
  • We review the transaction and make a final decision
  • We charge a fee only if we were able to approve the transaction
  • Transactions we decide to approve are guaranteed by us in the event of fraud

This model forces us to build a platform capable of discerning good transactions from the high-risk pool. This is difficult, so we designed advanced detection technologies.

The statement “an order placed behind a proxy is 5X more likely to result in fraud” is correct but useless. If you decide to decline all orders placed behind a proxy, around 80% of these declines will be false positives.

Most risk indicators have an inherent duality: A proxy may indicate a malicious attempt to hide one’s identity. It may also be used for privacy concerns – someone connecting through a workplace VPN or maybe someone streaming Netflix from outside the US who forgot to switch back to his regular connection.

Current risk management is broken because the business incentive of the service provider is not aligned with your incentive to sell more. Algorithms developed to just stop fraud are different from algorithms developed to enable high-risk transactions.

It’s hard to understand if a proxy is good or bad by studying it in isolation. You need to take a look at the bigger story derived from the full order data. That’s why we do everything in-house. Not because it was especially important for us to offer a turnkey solution, but because we needed to build and control the relationship between analytics and technology in order to offer our solution.

A company prioritizes building technology that enables its business model. I’m happy that our business model makes us build the best analytics, algorithms, and machine learning technology available. It’s fun and what the Riskified team enjoys doing.