(no title)
rbres | 7 years ago
Half a dozen points to answer this question:
- First off, our #1 north star metrics are order approval rates and minimal false positives. We measure this for every merchant. Clients sign up with us under the clear expectation that they'll see a spike in order approval rates and all of our case studies (compared to tools like Signifyd, Riskified, Stripe Radar, and many others) have shown significant spikes in order approval rates.
- We have some of our case studies published here https://bolt.com/case-studies and will be posting more in short order.
- We are more incentivized than siloed fraud detection providers to approve more orders. Payment processors are aligned to maximize volume. Fraud providers are aligned to minimize fraud. We play both roles. We want maximum volume through our pipes just as badly as merchants do.
- Furthermore, fraud rates are simply FRAUD / TOTAL VOLUME. We always aim to maximize TOTAL VOLUME to keep fraud rates low vs squeezing the numerator (which is what most companies do).
- Finally, as elaborated in the most, we have unparalleled data visibility by powering checkout. We use no rules. So, just because something looks bad (e.g. a user using a VPN, coming from a certain country, etc) our models just factor that in as one of hundreds of variables.
- We try to not profit and maximize fraud for the first several months to root out false positives in our models. If you only approve things that are clearly good and reject all else, your models will never learn.
- Our human review team rigorously processes every machine decline before it's officially declined to once again maximize order approvals.
A lot more that I could elaborate on here. But, just a few quick tidbits on what we do to minimize false positives as our #1 priority. All of the above and more lends itself to, what we believe to be, the best results in the industry.
rbres|7 years ago