Hear horror stories and tips on how to optimize manual review processes from experts with first-hand experience of fighting fraud on the frontline.
Manual reviews of suspicious transactions remain to be a key component within fraud prevention – yet it’s likely costing your business more than it should.
Ideally, 90% of all transactions should be either auto-declined or auto-approved but how do you determine which you need to look at manually?
Furthermore, if your anti-fraud system is too strict you might be losing serious money so it’s vital to choose the most suitable model for your business.
This webinar will help you with your manual review horrors, featuring members of our senior team who hold years of experience on the front line in the fight against fraud and an expert in the field!
Agenda
- Manual Review: Effective But Not Foolproof
- Key challenges with manual reviews:
- False positives
- Fixed outcomes
- Hard to scale
- Machine Learning as a solution?
- How can SEON help?
- Your questions answered
Speakers
- Ronald Praetsch – Co-Founder – About-Fraud
- Gergo Varga – Senior Content Manager & Ex-Fraud Analyst – SEON
- Bálint Patkós – Head of Customer Success & Ex-Fraud Analyst – SEON
The host was our very own co-founder and COO, Bence Jendruszak.
Key Answers From the Q&A
We often hear the victim claimed innocent on sharing OTP for the unrecognized transactions. Is Fraudster capable of hacking a user’s account to get OTP?
I just saw a couple of weeks ago an article where a fraud analyst had actually fallen victim to this exact scam. The fraudster had built a smart workflow and learned a loop hole on how the bank treats you when you added a new payment method but yes it’s perfectly possible. It just takes dedication from their side to figure it out and they will basically reverse engineer how it works to find that one hole where they can get in. OTPs are unfortunately not the end to be. – Gergo
In emerging markets, the manual review may work better BCS the customers aren’t that tech-savvy. How do you score those with limited/even zero digital footprints?
In the early days of SEON, we would speak to many financial institutions in the APAC and LATAM markets. In those specific markets, there’s not a lot of existing data available and IDs can easily be compromised meanwhile mobile wallets/payments are some of the highest in the world. Companies that work with us who issue loans rely heavily on our analysis tools as an email address can be viewed as a type of digital passport because you use it to register for social media and other websites. With the right tools, you can reveal this person’s trail of online information within a matter of seconds. – Bence
I agree, there is always a footprint and if there isn’t, you should investigate why. In general, use data as much as you can and ensure that your processes, security questions and truly understanding the context around the cardholder. – Ronald
If there would be one suggestion that you could tell all the Fraud analysts grinding through manual reviews day by day what would that be?
Again for leaders it is that idea of talking to your team as humans. Your team needs to be motivated, get your team talking perhaps in a standup session on a daily basis to talk about their experiences. I have seen many times where teams become isolated as they work on individual cases but I think the interaction between the team, sharing their knowledge instead of just doing their job for 8 hours without sharing helpful insights. – Ronald
Try to have fun. We know that it’s a heavy job but if you are good at what you do, you can get into a trance and sometimes you can have a good combination of people alongside you. The fun things like sharing your taste in music whilst everyone is grinding through the shift – it can massively improve morale. – Gergo
Machine Learning is often not enough because we can see several types of fraud been placed. Do you agree with that? If yes, which other layer you like to use (ex. 2factor auth, Face Recognition, etc)?
Short answer is yes. There’s a couple of angles to tackle this, firstly do you have all the data across all use cases? Is the team responsible across all use cases instead of separate silos? Having different data and different tools, with machine learning trained across the use cases is the biggest challenge. The biggest challenge is the organisation of data across the departments and training your machine learning model for each use case, for account opening, login, transactions etc. – Ronald
Not just from the machine learning aspect but different types of fraud require different expertise or methodologies. If we’re talking about other layers, there are legacy processes like taking a selfie with your ID but they can add a lot of friction so only maybe when a customer is deemed risky by your machine learning fraud tool but the user wants to proceed then you embed such actions. These however can be costly so that is another challenge. – Bálint
I would add that it isn’t just about machine learning, it’s about having a holistic security mindset. A lot of times your fraud analysts only looks at the transactions but to minimise risk you need to be monitoring the entire customer journey because that’s where a fraudster will trip. Maybe they have perfected how to beat your sign up page but will later be lazy and log in via mobile which reveals their true IP is in a different country. If the data is siloed and not accessible, your analyst will be chasing databases instead of having one system that tracks all information in real time. – Gergo
What was your longest shift, and what happened?
Personally, the longest shift that I worked started as an 8 and a half hour day then as I headed home, upper management called to say the next shift was almost empty so I had to jump in from home… and altogether it was almost 17 hours. This for me clearly shows the issue that if we don’t have the option of automating processes, the human element of fraud prevention is more prone to illnesses, uncontrollable problems, and other absences. – Bálint
It is worth keeping in mind though peak season is coming soon, with Black Friday and Christmas around the corner. I would say it is important to make sure you get the team ready and all your resources ready as it’s important to handle this in the most human way possible to avoid these ridiculously long shifts which only demoralize your workforce. – Ronald
Watch our webinar to learn about how to prevent horror stories!
Watch our webinar to learn 5 reasons to use social KYC and digital footprint analysis!
You might also be interested in reading about:
- SEON: KYC Verification: How to Save Costs & Reduce User Friction
- SEON: AML Fraud Detection: How to Choose a Solution
- SEON: Identity Proofing: What Is It & How Can It Prevent Fraud?
- SEON: Alternative Credit Scoring: How it Works?
Learn more about:
Data Enrichment | Browser Fingerprinting | Device Fingerprinting | Fraud Detection API
