BFA #014 | Your fraud strategy is failing

This might sound bad, but I see it as an opportunity

What's happening Fraud Fighters?

Fraud strategy not performing? Let’s get testing.

When I joined a company, their rejection rate was through the roof and so was there chargeback rate. Now this might sound like some bad news, a bleak way to come into my new role, but I saw this as an opportunity.

Let's investigate further.

Read Time: ~3.5 Minutes

The House is Burning

I bet you saw this and thought to yourself “yup that’s me…”

When I first joined a company, our fraud strategy was a little over the place. It was all over the place because there was no fraud strategy. When I started, one priority was to get our rejection rate down while lowering our chargeback rate.

It felt like we were catching all the legit users and letting the fraudsters through. This heightened anxiety was understandable, given the potential consequences of failing to adequately address the fraud issue.

In the face of such challenges, it became crucial to establish confidence in our ability to effectively tackle fraudulent activities. Overcoming this tide of nervousness required a well-thought-out strategy and a comprehensive understanding of the fraud landscape.

The key here was to run small, low budget experiments, prove the value, and ramp gradually. You need to earn trust before you start to think about scaling. Start small.

You found out your fraud strategy is failing? Let’s get testing!

Experiment, Experiment, Experiment

The constant slack pings (or teams if you’re into that) from colleagues felt like a steady stream of accusations, as if you were personally responsible for each fraudulent transaction slipping through the cracks.

You reluctantly turn on the camera for your update to your boss. You need to balance immediate wins with laying down the foundations for your long-term fraud strategy.

You may be thinking you never want to find yourself in this situation, but I saw this as an opportunity. This was a great example where I could get into things and make an impact right away.

I started by

  • Getting our data in place

  • Auditing the transaction flow

  • Using the transactional, chargeback, and customer reported data to identify issues

I could’ve focused on onboarding and how the users got into our ecosystem to begin with, but all eyes were on the payments. Here’s how I began overhauling our fraud strategy with a new perspective.

  1. Gather data: Compile transaction data from the period you want to analyze, including chargeback data and any associated customer and transaction details. Be sure to include both fraudulent and non-fraudulent transactions.

  2. Data preprocessing: Clean and preprocess the data, addressing any missing or inconsistent information. Standardize the data to ensure accurate analysis and interpretation of patterns.

  3. Identify key variables: Determine the most relevant variables that may be indicative of fraud, such as transaction amount, location, payment method, IP address, time of day, and frequency of transactions. I could keep going but these are just a starting point.

  4. Perform exploratory analysis: Conduct an exploratory analysis of the data to identify any initial patterns or trends. Visualize the data using graphs and charts to better understand the relationships between variables.

  5. Segment the data: Group transactions based on common characteristics, such as transaction type, customer demographics, or geographic location. This will help you identify patterns within specific segments and improve the precision of your fraud detection rules.

  6. Analyze customer behavior: Examine customer behavior patterns, such as purchase frequency, types of items bought, navigation patterns on your website or app, and any communications.

  7. Detect patterns and anomalies: Analyze the data to identify patterns or anomalies indicative of fraud, such as sudden spikes in chargebacks, high-risk locations, or abnormal transaction amounts. Note any emerging trends that could be used to enhance your fraud detection rules.

  8. Develop new fraud detection rules: Based on your analysis, create new fraud detection rules to address the patterns and anomalies you've identified. Be sure to consider the balance between minimizing false positives and effectively detecting fraud.

  9. Test and refine rules: Implement the new rules in a test environment to evaluate their effectiveness in identifying fraudulent transactions. Use feedback from this testing phase to refine the rules and improve their accuracy. Best case scenario, you can test these with live data in “dark mode” or as an A/B test.

  10. Monitor and update: Continuously monitor the performance of your fraud detection rules and update them as needed based on new trends or emerging fraud patterns. Regularly review and analyze chargeback data to ensure your rules remain effective and responsive to the evolving landscape of fraud.

That meant I had to make sure communication was on point by doing a few things:

  • Making sure everyone could see what I was up to

  • Going all-in on changes that would make the user journey better

  • Keeping a close eye on everything and making sure our data was clean

Now, because of some bad experiences in the past, people had lost their faith in handling fraud. But after digging through all the data I collected, I quickly found ways to fix our fraud logic and user journey.

That was the proof I needed to show the bosses that we were finally on the right track. If you're trying to scale a fraud team without a lot of resources, my advice is to manage everything closely.

Remember, you can't just set things up and forget about them.

Fraud Can A/B Test Too

You know, there are so many ways to build and manage a fraud team. I've found that it's best to start with the high-value, low-effort stuff first. Once you've got that sorted, you can expand and tackle other challenges, always keeping in mind that you've got limited time and resources, of course.

Oh, and another cool thing you can do is run mini experiments throughout your user journey. I mean, everyone's into A/B testing, so why not apply it to fraud prevention too, right?

Here are some experiments I tried out myself:

  • Adding just one extra piece of data

  • Borrowing insights from our marketing team

  • Using labels from different stages of the user journey

These tweaks might sound simple, but they can make a big difference. Sometimes, it's the small changes that lead to the most significant improvements.

And guess what? We totally nailed it with these changes. Big wins all around!

After payments, we turned to account takeovers, but that’s a whole other story that I’ll save for another time…

What’s New This Week?

No new content this week besides this week’s batch of LinkedIn content.

As you know that fraud Job Board is a side project for our community, so I’ve been working on that little by little in my few spare moments.

Major improvements will be coming in time and are currently in progress.

We’ve sourced almost 200 additional jobs this past week.

Question for the Fraud Fighters

See you again next Friday in your inbox.

​Brian

Reply

or to participate.