Embracing automation: How Kinguin transformed its approach to fighting fraud

Kinguin is an online marketplace for gamers, where developers and wholesalers sell games directly to customers. They’ve achieved rapid growth, growing its transaction volumes by x3 since 2021, as well as reaching over 17 million customers.

During this extensive period of scaling up, Kinguin’s Fraud and Risk Intelligence Team fundamentally transformed their approach to risk management.

In 2021, Kinguin employed an in-house rules engine and large manual review team to tackle fraud. As a business enabling instant delivery of digital goods, having approximately 1 in 8 transactions sent to manual review was creating a bottleneck – which would have only worsened considering the scale Kinguin was ramping up toward.

The internal manual review team was dealing with insurmountable workloads. This resulted in poorer decision accuracy and an unhappy team, which ultimately led to poor performance.

The team was stuck in a vicious cycle. Internally, Fraud and Risk Intelligence were seen as a cost sink.

Egemen “Ege” Ertop, who by 2022 had moved from a frontline risk analyst to Head of Fraud & Risk Intelligence, knew that a fresh approach to tackling fraud was required. One that allowed them to do more with less, and build a resilient set-up that would scale.

“We wanted a set-up that enabled us to handle as much as possible automatically,” said Ege.

“What do I love most about working with Ravelin? It’s the service. I have the best account manager I could hope to work with,” explains Ege.

Partnership has underpinned Kinguin’s epic transformation. Ege led his thinly stretched team away from an approach that relied heavily on human intervention to one that centered on automation, all the while onboarding a new fraud solution. Ravelin’s integration and support teams spent hours in the trenches learning about Kinguin’s processes and brainstorming with Ege to identify what Kinguin’s new approach to risk management would look like – and how Kinguin could get the best out of Ravelin.

The central technology enabling the shift to automation is the custom machine learning model Ravelin built for Kinguin. This has been a huge mindset shift for the Kinguin team, who have historically relied so heavily on human evaluation.

Today, the vast majority of transactions receive an automated decision based on the machine learning score, with only a few critical transactions going into manual review. And throughout this period of fewer and fewer transactions being manually reviewed, Kinguin’s fraud rate kept decreasing.

All the while, the lives of Kinguin’s fraud analysts have changed. Analysts have drastically more time to review a transaction when this is needed, leading to more accurate decision-making (and in turn improving the machine learning model), a happier team, and more capacity to focus on proactive activities such as strategy, investigations and skill development.

As a team manager, Ege notes the resilience his team now has:

“Two years ago, if an analyst in my team was out sick, or left the company, I’d be dreading how to manage their workload. Now we have technology to lean on if operational capacity shifts.“

Kinguin’s transformation has been nothing less than remarkable. Across key performance metrics, Kinguin has seen these results:

• 87.5% decline in fraud rate

• 57% reduction in block rate

• 90% reduction in manual review rate

In a matter of two years, Kinguin’s Fraud and Risk Intelligence Team has pivoted from being a weak link in the chain to a crucial function that powers merchant growth on the platform.

For a seller on any marketplace, trust is a crucial component. As Ege highlighted, “sellers want peace of mind that the platform they’re selling on doesn’t really have a problem with fraud”. For Kinguin, this robust risk management strategy is central to their seller value proposition.

Thanks to Ege and his team, sellers on Kinguin's platform can have that peace of mind.

• On-demand webinar: "How to get the most out of your fraud solution – The evolving fraud needs of a growing business" featuring Egemen from Kinguin
• Article: How has the perception of the fraud team changed?
• Article: Machine learning and fraud prediction: What data is required to make it work?
• Ravelin Insights: Machine learning for fraud detection