Hacker News Reader

The product looks interesting but the hand picked examples are literally useless. 3 examples where you approved a transaction that a competitor opted to decline. Based on your examples your model could literally be "approve every transaction".

It's not a fair comparison unless you compare approval/decline rates and total fraud per platform.

View on HN

Yep, totally get it. We wanted to show specific examples, but obviously not disclose PII. So, this was a middle-ground although not idea.

For every client we measure order approval and decline rates before/after. Here are several case studies with those results: https://bolt.com/case-studies

And we have more that we can send over if you're interested in learning more! rb [at] bolt.com.

Reply on news.ycombinator.com