Many online businesses are facing financial pressures as a result of the COVID pandemic for a wide variety of reasons. As the macro-economic view across many sectors looks gloomy after the impact of lockdowns, managing costs is at the forefront of many CEO’s minds. This is trickling down to all functions, and online fraud detection is no exception. I am seeing it less so in banks, who haven’t really lost business as a result of the pandemic. However, in ecommerce it’s more prevalent. Yes, ecommerce volumes may have gone up (a lot) in some sectors, but in many others they haven’t (have any of you booked a flight recently?). And increased volumes alone don’t always compensate for increased disruption and costs in supply chain, logistics and other back-end functions.
A mistake that some people make when looking at cost optimization in fraud detection is thinking that the goal is to reduce the amount of fraud that you’re suffering. True, that is one element of reducing fraud costs, but there is far more to it. You can think of the total cost of fraud (TCOF) as being made up of multiple elements:
- Fraud Losses – funds stolen from accounts; stolen goods and services; chargeback costs.
- Tools and Headcount – the costs of detecting and mitigating fraud; internal systems and vendor tools; human resources.
- Customer Lifetime Value Impact – customers who experienced fraud and decide to shop or bank elsewhere; good customers who experience friction due to fraud detection and decide to go elsewhere for a better CX; false positives.
All of these contribute to your TCOF. So when thinking about cost optimization in fraud detection, there are various knobs to turn and levers to pull to help bring your TCOF down in a way that best fits your business.
This can be a difficult message to sell internally, when other stakeholders may only be pressing you to lower your fraud rate. If you’re responsible for fraud management, the key message is this:
The goal is not to minimize fraud itself; rather, it is to minimize the total cost of fraud.
There is a notional model that I use with clients to illustrate to them that if they focus only on lowering fraud rate, it can actually push up the TCOF – as achieving an unrealistically low fraud rate usually involves pushing up your false positive rate and the costs of the tools you’re using. These increased costs can offset any savings made by lowering the fraud rate. Likewise, if you let the fraud rate get too high, then its contribution to the TCOF becomes dominant and your TCOF increases.
The key is to try and maintain your fraud rate at an optimal level that results in the lowest TCOF. There is no magic formula or spreadsheet that calculates all this for you, and as I mentioned, my model is notional and qualitative rather than quantitative. However, it does serve as a very useful tool in my client conversations, helping fraud managers to define the narrative on how they can support their business in optimizing costs. In this kind of environment, it’s better to be making proactive suggestions on how your function can optimize costs. The alternative is that someone else will optimize your fraud detection function for you.
For further insights into this topic, Gartner clients can check out my research Cost Optimization in Online Fraud Detection or get in touch with me.