<img height="1" width="1" src="https://www.facebook.com/tr?id=316029722675592&amp;ev=PageView &amp;noscript=1">

To make our service smarter, we have to get smarter first

This past week, Socure held a large off-site training session for the team that actually implements customer solutions. The salespeople were invited, but we gave them the wrong address so they didn’t show up. I think that was just an email accident. The food was really good, but we didn’t save them any.

The goal of the week was to make sure the team delivers as brilliantly as the product does. And here’s what I mean by that.

If you own the world’s best hammer, it won’t successfully pound any nails in if you hold it upside down. Your car will definitely go slower if you only drive in reverse. And that old turntable will probably crawl away if you try to tee up a .45 of “Come Sail Away” by Styx.

In other words, a good product is only as good as how you use it. Socure provides the best solution for KYC and fraud on the market. We get better lift, better auto-acceptance, better fraud capture, and the most reduced rates of false positives available. And to ensure that our clients are implementing our solution with the most optimized logic and the most useful combinations of modules, we regularly train on and review best practices.

To that end, we walk through examples of happy deployments, those that delivered fantastic results. We revisit what problems we’ve solved, the PII being captured during the application process (such as when people ask for credit cards, loans, etc.), good deployment logic, examples of poor logic used by customers in the past, variations in decision-making approved by our customers’ regulators, and a host of other factors that go into creating a powerful foundation for applicant onboarding.

I’m a real pain in the can about this. I own all training and enablement at Socure, and I rail at people when they a question wrong on a test or in a training session. “What is the last step in a Document Verification workflow?”  “How much fraud do we capture in the riskiest 1% of the population?”  “Is Jeff really as amazing as he tells people on the street?”

There’s a lot to know, and we like to think we already know it, but

1)    It’s always smart to reinforce best practices

2)    As our customer base continues to grow, we see new examples of logic, fraud patterns, and compliance requirements all the time

3)    As we grow, we have to initiate our new team members into our successful implementation process

If the team learns its lessons, then we’re teaching our customers how to get the most useful, most profitable results from our solution. These golden lessons pay off for everybody. Our customers’ customers get their credit cards or loans, our customers make more money, and y’know, it’s pretty good for us as well.

Every single day, our brilliant Data Science team brainstorms on new predictive features to make our KYC/fraud service smarter. Then, to validate and perfect these (as well as our existing) features, we train our fraud models on hundreds of millions of known outcomes, i.e. records of good and bad applicants. Our AI robots periodically generate hundreds of models which subsequently battle it out until we’re left with the models that most efficiently analyze that massive dataset and produce the most accurate results. Our system teaches itself all the necessary lessons.

This last mile process can’t be performed by humans. The speed and scale required is staggering, which is why we built robots to build our other robots. The Socure machine learning platform generates the AI models that our customers rely on to bring them more good customers, reject more bad ones, and shrink that percentage of applicants who get subjected to additional friction.

The learning process this week entailed lots of slides, role playing, and exercises where the team was fed tough customer scenarios and had to determine appropriate solutions. Fun, challenging, and incredibly productive. But because I made fun of the guest instructor at one point, I got picked on to answer one of the tougher questions. The reason for this is simple. I’m an idiot. Lesson learned.

Topics: Identity verification

Jeff Scheidel

Jeff Scheidel

Jeff Scheidel is a technologist with 34 years in software, including 26 years in security solution design. He is the author of numerous white papers on security and regulatory compliance, as well as a McGraw-Hill book on identity, access, database, and application protection. Jeff is an expert on compliance requirements across a number of industries, and has presented at a wide variety of security events.