#32 - STOP obsessing over accuracy, START optimizing efficiency

Fraud management is a discipline that requires us to spin many plates simultaneously.

One of the most talked about balancing acts is between catching fraud and not impacting user experience.

I myself have written about it numerous times in the past, but the fact is that this separation is a bit artificial. After all, these are two sides of the same coin.

And the coin’s name? Accuracy.

If you improve your accuracy, you’ll improve both your losses and approval rates.

But the knot that is really hard to untangle? That is the one between accuracy itself and scalability.

It’s all around us. Think about it:

Should I take time with my review to be sure in my decision? Or should I hurry up to review more cases?

Should I optimize my rule’s coverage to catch more fraud? Or should I optimize it for accuracy to avoid false positives?

Should I train a single AI model to streamline how much resources it drains for me to maintain it? Or should I opt for training a dedicated model for each segment for better performance?

Having to choose between accuracy and scalability sows doubt everywhere we look.

Or does it?

What is often misunderstood about accuracy and scalability

Here’s the misconception I’d like to break: I often hear teams claim that system accuracy leads to system scalability.

Meaning if I’d like my system to scale, I need to work on high accuracy first.

But I believe that in most cases, the opposite is true.

I believe that in most cases, low accuracy is the result of a broken system that is having difficulty scaling and keeping up with the business’ needs.

Here’s an example:

Say I issue credit cards. If I only receive one application per day, I can probably create a very robust fraud prevention system for very cheap.

Heck, as the CEO I can call the customer, have a 15-min chat and approve the application myself.

Easy, cheap, and bulletproof.

But if tomorrow I get 200 applications? Then my process will fail, and fraud will explode.

This is obviously an extreme theoretical example, but you get the point: the accuracy of a rushed, half-baked process is always better than having the process break down.

And so the way I see it, accuracy is the result of scalability and not the other way around.

Here’s another misconception I’d like to address:

To improve scalability we need to expand horizontally by distributing work between parallel workers.

In the above example, it would mean hiring 10 employees to conduct the applicant interviews instead of the CEO. Clearly, that’s not a viable strategy as I'll need to linearly increase my costs as the business grows.

But what if I could reduce the processing time? What if instead of a 15-min call, I can have a 3-min fraud review? Now, I only need two employees to manage the same task.

And what does that tell us?

Scaling can come from horizontal expansion, but it’s much better if it’s driven by efficiency. And by “better” I mean cheaper and with a greater ROI.

Our fraud strategy should maximize our potential

As fraud strategists, we need to focus on building efficiently scalable systems.

Accuracy is a by-product of such systems.

In fact, when I encounter leadership teams that obsess about accuracy or how it can be improved, my first task is to help them stop thinking tactically and start thinking strategically.

Side note: Accuracy and tactics do matter. When fraud is on the rise or when we experience a dip in approval rates, it’s imperative to deal with that. But this is the exception to the rule, at least when it comes to the leadership level.

Think about it this way: let’s say you’re a runner training for a competition and you’re worried about your results. What sort of advice would you seek from your trainer?

Sure, he can probably comment on how you utilize your energy: how you swing your arms, how wide your steps are, etc.

But what would have a bigger effect is what they’ll have to say about how you maximize your energy: what you eat, how much you sleep, how you warm up, and so on.

With fraud as with sports: when discussing strategy, we want to focus on maximizing our potential.

Everything else is obsessing over symptoms.

Grounding theory in practical tools

How do we make sure we’re speaking about the right stuff?

Simple: We speak about how fast our processes run.

I’ve already discussed in-depth ​the concept of Reaction Cycle​, but to briefly reiterate: to become faster at dealing with fraud we need to measure and manage the time it takes us to react to it.

Our Reaction Cycle is comprised of two main stages, each made of 3 phases:

1. Learning: detecting, scoping, and analyzing emerging threats

2. Acting: designing, testing, and deploying solutions to meet those threats

Now let’s imagine a scenario where the Director of Fraud Prevention requests a budget for implementing a new AI score.

The CEO’s (or whoever owns the budget) role isn’t to ask “how much loss would it save us?”.

Instead, they should work together to understand how would the new AI score solve a bottleneck in one of the above six stages:

1. Would it help us detect fraud attacks much faster than today?

2. Would it help us assess their scope faster?

3. Would it save us time in analyzing the patterns we’re seeing?

4. Is it easier for us to design a fix with an AI score?

5. Will it be easier to test?

6. Will we be able to deploy fixes much faster?

The point is not only to better understand the levers we want to pull, but also to validate it’s the most cost-effective option we have at our disposal.

Another thing to consider is that the effect might be indirect in nature.

For example: It might be that training, testing, and deploying an AI score will take us much more time than doing the same with a single rule. But having an AI score would reduce our dependency on rules by half.

In that case, even if it seems slower in comparison, relying on an AI score might still make our system faster as a whole.

The bottom line

I know, this all might seem complex. Especially measuring how much faster processes would be before we actually get a chance to try it out.

But look at it like this:

The point is to have a strategic, long-term discussion on how we increase potential.

Strategic alignment is all about matching visions, so when fraud hits the fan the entire leadership team has the same view on things.

If your conversation about fraud focuses on accuracy, you’re not speaking about strategy. You’re stumbling over tactics.

Teams that focus on long-term strategy don’t fiddle with technology or innovation.

They go back to basics and fix their efficiency.

Because efficiency leads to scalability, and scalability leads to accuracy.

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:

Free Discovery Call - Unsure where to start or have a specific need? Schedule a 15-min call with me to assess if and how I can be of value.
​Schedule a Discovery Call Now »​

Consultation Call - Need expert advice on fraud? Meet with me for a 1-hour consultation call to gain the clarity you need. Guaranteed.
​Book a Consultation Call Now »​

Fraud Strategy Action Plan - Is your Fintech struggling with balancing fraud prevention and growth? Are you thinking about adding new fraud vendors or even offering your own fraud product? Sign up for this 2-week program to get your tailored, high-ROI fraud strategy action plan so that you know exactly what to do next. ​
Sign-up Now »

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

<