Learn what your chargebacks are

Understanding the chargebacks specific to your business is the first step in creating a fraud strategy that reduces bottom line loss. Once you understand your chargebacks, you can build a plan to attack them with thoughtful chargeback responses while also creating a strategy that works to prevent them in the first place. Within this article, I explain how to isolate categories and create subcategories that will be crucial to all further steps in developing your unique fraud strategy. It all starts with digging into the chargeback codes, creating a database, and looking at trends. Once you have the big picture laid out that tells you what chargebacks you’re receiving, you can start to investigate what causes them. From there, you can start working to prevent.


1. Create high-level categories from the chargeback reason codes

Banks provide codes for each chargeback. Take these reasons and create high-level buckets. These will come in handy later in the process.
Chargeback codes are different depending on card type. You can read more about the specific meaning for each code from Chargebacks 911.
If you’re using a payment process rather than received chargebacks directly from a bank, they will generally already be categorized. Stripe, for example, uses the following buckets: general, fraudulent, product not received, duplicate, credit not processed, unrecognized, product unacceptable, subscription canceled.

2. Conduct hands-on analysis into individual chargebacks to isolate characteristics for each chargeback

Ideally, you should review as many chargebacks as possible, but this depends both on your team size and the number of chargebacks you’re receiving. The more you can review, the more likely it is that you will be able to pick up on trends. The goal here is to isolate user behavioral characteristics for each chargeback. You’ll also want to discern what is causing chargebacks: friendly fraud, malicious fraud, or product issues.

These characteristics should be identifiable to a trained eye, whether it be through external services such as Sift Science ML’s key fraud indicators or internal order data points like a customer review or email communication.
Once you have reviewed enough chargebacks, the characteristics should allow you to find patterns between chargebacks. With this data, you can move forward in building a strategy.

Some categories will be low in chargebacks. That means you don’t necessarily need heavy analysis on those. If a category is low in frequency, it’s likely not the right thing to focus on. 

Don’t try to group them together yet, simply note different characteristics. As you observe more unique cases, trends will emerge and allow you to create a more organized list of characteristics.

3. Start isolating patterns, let’s call these subcategories

An example pattern could be a chargeback for ‘product not received’. Characteristics include tracking that shows the order was delivered late and your order status update didn’t reach the customer.
The subcategories, in this case, could be ‘delivered late’ or ‘poor customer experience’.
Start asking questions about the holistic customer experience that could have caused the chargeback. Did you have poor customer service in handling a refund or cancellation? Did you charge for a subscription without clearly indicating so or following up? Does the user actions for fraudulent grouped chargebacks look like normal user behavior? Are there factors such as changing the shipping address, IP address, device, or time zone?
Once you have both categories and subcategories, two things happen:

  1. You can start fighting your chargebacks effectively with targeted templates. Read more on structuring your chargeback response.
  2. You can start preventing fraud by targeting subcategories. You must firmly understand your data before you begin this step.

Before we get to these next two steps, a little housekeeping to make this process iterative…

4. Set up tracking for chargeback codes

Create a data analysis template to take these chargeback codes. It should allow you to map them to categories easily and then monitor fluctuations over time. In addition, you’ll want to monitor your win rates over time for these codes. You will need to adjust the templates periodically as you gain learnings to create the strongest response.

5. Be sure to monitor bank decisions on chargebacks

The only path to success with chargebacks is iterating on the entire process; however, banks generally can take up to 60–75 days to resolve a chargeback once the evidence is submitted. Don’t forget to check back in to learn what worked versus what did not work. Sometimes this can be tricky. Two chargeback scenarios can look exactly the same with exactly the same response. You win one and lose the other. In these cases, be sure to step back and make sure your assumptions about the lost chargeback were accurate. If it’s fraudulent, the bank might know more than you do, meaning you could have diagnosed friendly fraud incorrectly. Remember — correctly analyzing and understand your chargebacks is a top priority in building a strong fraud strategy.