#18 - Is your fraud team working backwards?

Here's something I see far too often:

A Fintech builds sophisticated fraud prevention systems, invests in automation, and implements cutting-edge tools.

But when it comes to their fraud operations team? They're still working cases on a first-in, first-out basis.

Or even worse, they think they're being smart by prioritizing by transaction amount.

Why is that an issue? Best case, it creates very costly inefficiencies. The higher the traffic - the more you need to hire additional team members.

Worst case, it actually slows down your growth.

So today I want to share one of my top strategies for FraudOps: Queue Prioritization.

Let’s talk about it.

FraudOps teams need to be optimized for ROI

How much do you invest in your fraud team right now? Let’s consider the combination of gross salaries, hardware, software, etc. How much does it cost you to run a single FraudOps person-hour?

Side note: we’re only taking into consideration direct costs. I’m not even going into the “missed opportunities” cost of operational inefficiencies, as we want to keep it simple and manageable. Point is, treat the above number as a base minimum.

Now let’s talk about the return side of the equation: How much in fraud losses is your team blocking? How much savings per person-hour do you generate?

Here’s the thing:

While you cannot easily change your cost metric, you can influence how much you manage to save. In fact, there are three dimensions that influence it:

Exposure: This is a combination of risk level and amount. A high-risk, high-amount case obviously deserves more attention than a low-risk, low-amount one.

Accuracy: What are your false positive (blocked good users) and true negative (unblocked fraud) rates? Mistakes don’t only cost spent resources, but also create additional negative business impact.

Resolution time: How long does it take to investigate this type of case? Some cases might require extensive documentation review, while others can be resolved in minutes.

Side note: The above is also the internal priority order between the different dimensions. High exposure cases are more urgent to review than cases you can resolve quickly.

Each of the above dimensions should be evaluated per case, before investigation even begins. And by prioritizing accordingly, you can now make sure that every FraudOps hour you spend yields the highest return.

Queue Prioritization: Beginner to Pro

Let’s look at how to practically approach Queue Prioritization, depending on how advanced you are as an organization.

Beginner: Your FraudOps team is your fraud organization. You have little to no automation tooling and/or knowledge. Assigning and tracking KPIs doesn’t come easy to you.

Approach: Take a day off and map which processes flag cases for review. These might be AI score thresholds, rules, or even hardcoded logics inside your product.

Review the last 6 months of data, segmented by the different flag sources. The more granular you get - the better.

Finally, try and calculate the above KPIs per flag source:

• Exposure: What was the average loss per case?

• Accuracy: What were the False Positive and True Negative rates? (Not sure how to calculate them? ​Check out this issue of TSFS​).

• Resolution time: What was the average resolution time per case?

Found segments that show good metrics on 2-3 dimensions? Prioritize them for review first. Likewise, downgrading the priority of sources that show poor metrics can increase your ROI.

What if you only have one flag source, say AI score? Segment it yourself.

For example, if every case that gets a score >70 goes to review, segment the cases that got scores of 70-79, 80-89, and 90-100 separately, and measure them as different sources.

Now all that is left to do is to track your FraudOps over time and see how it improves while you optimize your prioritization.

Intermediate: You have a fraud analytics team and feel comfortable writing fraud rules. You have some dashboards where you track basic KPIs in real time.

Approach: After you’ve taken all of the above steps, it’s time to focus on improving your underlying logics. There are two ways of doing that.

The first, is to improve the source logics that flag events for review. If in the past they were mainly based on risk, now you can start optimizing them for accuracy and resolution time as well.

Secondly, it’s time to start introducing “quick resolution” logics to your real-time flow. Every FraudOps team has them: details that would make the investigator go “oh, this one’s fine” within less than a minute.

Curate these logics, make sure there’s consensus around their strength, and then find ways to exclude them from being flagged in the first place.

As we all know, it might take one minute to decide on a case, but to actually resolve it, you might spend 5 more minutes. We want to avoid this altogether.

Pro: You have your own data science team and employ your own Machine Learning (ML) models for fraud detection, even if not in real time.

Approach: It’s time to fully automate the prioritization process with a dedicated ML score. There can be several approaches to it, but a good place to start would be clustering.

The “closer” this case is to other past cases that generated high savings - the higher priority it should get. Notice that we’re labeling based on savings, not risk.

Admittedly, the initial investment in rolling out a new model can be quite high. Consider, though, that this is not a real-time model, and it’s also not likely to need the same cadence of re-training.

In return, you’ll be able to streamline your Queue Prioritization process, without caring too much about the different flag sources. This would also mean that changing or adding more sources would become easier.

Side note: Do not use your fraud score for Queue Prioritization! There’s a difference between predicting what is fraud and what would generate higher investigation savings.

The secret benefit of measuring accuracy and resolution time

Ever wondered how to evaluate a vendor for your FraudOps?

Now you can easily do that: is the overall improvement they generate in accuracy and resolution time greater than their cost?

This will make it easier for you not only to decide if to take up a new vendor, but also to compare the competition based on ROI.

That’s probably easier to justify to your CFO than “they have a cool UI”, right?

Have questions or feedback? Reply to this email, I read all messages.

In the meantime, that’s all for this week.

See you next Saturday.

P.S. If you feel like you're running out of time and need some expert advice with getting your fraud strategy on track, here's how I can help you:

Fraud Strategy "Power Call" - Book a consultation call with me to get clear, actionable recommendations that fit your budget. Guaranteed.
​Book a Call Now >>​

Fraud Strategy Workshop - are you an early-stage Fintech that needs to move fast and with confidence? Book this 1.5-hours workshop to get instant insight into your vulnerabilities, optimization opportunities, and get clear actionable recommendations that won't burn through your budget.
​Book Your Workshop Now >>​

Fraud Strategy Transformation Program - are you a growth-stage Fintech in need for performance optimization or expansion of your products offering? Sign up to this 6-8 weeks program, culminating in a tailored made, high-ROI roadmap that will unlock world-class performance.
​Schedule a Call Now >>

Enjoyed this and want to read more? Sign up to my newsletter to get fresh, practical insights weekly!