Hi Eyal,

I work on the risk team at Stripe and wanted to clarify a few points.

First, as the emails above show, we didn’t suggest “racial profiling”, to combat fraud or for solving any other problem…ever. That wouldn’t make sense morally or logically.

We flagged the first order you mentioned only because the shipping address (Algeria) was in a different country from where the credit card was issued (Greece). This kind of discrepancy often increases the likelihood that a charge is fraudulent. We would also have flagged a transaction from a British card shipping to the United States.

You’re definitely right to question the use of the phrase “gut check.” We actually take a pretty methodical approach to dealing with fraud, and that didn’t quite come across in our previous response.

What we should have said is that when businesses start out on Stripe, we assign a machine learning-based fraud algorithm that corresponds to the type of business they’re running (SaaS businesses get the SaaS model, etc). Our systems are built to detect patterns and decline charges automatically when we’re confident they’re fraudulent. Particularly when users are new to Stripe, we may not have enough of a sample size to detect fraud patterns unique to a specific business, so they’ll only benefit from fraud algorithms designed for their type of business. Input from you helps us accelerate accuracy. We have thousands of users working with Stripe in novel ways, and we’ve seen that a one-size-fits-all model just doesn’t work as well as a tailored one.

Having said that, we’re continuously updating our fraud screening to make it work better in cases like yours, and we’d be happy to work with you to further minimize fraud on your account. If you’ve any suggestions on how we can improve our fraud detection further, I’d love to hear them! You can reach me at anurag@stripe.com.