Want to know more?
Are you interested in this project? Or do you have one just like it? Get in touch. We’d love to tell you more about it.
Beating ecommerce fraud at scale
Global marketplace
How a global retailer outwitted criminals to save millions
Our client, a global ecommerce marketplace managing millions of SKUs and working with over 18,000 suppliers, was facing a growing problem in tackling online fraud. This was exacerbated by a surge in digital transactions during the Covid-19 pandemic, and an accompanying move by fraudsters away from bricks and mortar theft in favour of targeting ecommerce sites.
The company engaged Equal Experts to help modernise its fraud detection infrastructure, implement intelligent detection systems, and build scalable, future-proof solutions to combat the increasingly sophisticated threats while preserving impeccable customer experiences.
Outcomes
$7M
savings in false positive order cancellations and chargebacks
90%+
reduction in fraud intelligence licensing costs
15%
increase in fraud detection
ABOUT
Our client is a multibillion dollar revenue company in the retail and wholesale industry. Thanks to the size and nature of its online e-commerce operation, it is a tech company in its own right, as well as a furniture and home goods marketplace. The business currently offers for sale 14 million items from more than 18,000 global suppliers. It has offices and warehouses throughout North America, Ireland and the United Kingdom.
Industry
Retail & Ecommerce
Organisation size
14,000+ employees
Location
Global, with headquarters in the US
Length of project
3 years
Challenge
Keeping pace with evolving fraud techniques
It is generally recognised that ecommerce is underprepared to deal with increasingly sophisticated attacks. Average fraud rates are typically 1-2% of revenue, with scammers tailoring their approach for each merchant. This creates a problem that is difficult to solve with a standard approach.
Our client was experiencing significant financial losses due to increased fraud, with a sharp rise in use of stolen credentials, and scams exploiting customer return and refund policies. Criminals requested refunds or replacements for goods they claimed were undelivered or damaged – often sending back empty boxes or irrelevant items.
Existing fraud checks were embedded in a monolithic system, making them slow, inconsistent and unreliable. This led to poor customer experiences, increased operational overhead, and an inability to respond quickly to new fraud tactics.
Due to its platform model, with thousands of suppliers, unchecked fraud posed not just a financial risk but also a reputational one. Suppliers relied on this marketplace to operate, and continued exposure to fraud could erode trust and competitiveness. The company needed to evolve its fraud strategy, shifting from reactive monitoring to intelligent, proactive prevention.
Solution
Building agile fraud defenses with modern tech and real-time insights
Our team took a two-pronged approach.
1. Modernising the architecture to scale fraud prevention
We began by upskilling the client’s fraud engineering teams in modern technologies, microservices, and cloud-native architectures. We then led the decoupling of fraud services from the legacy PHP monolith into a scalable, maintainable microservices architecture hosted on Google Cloud Platform (GCP).
We also integrated an API gateway to streamline interactions with multiple external payment processors – breaking free from a pattern of proxying third-party requests – to reduce latency and improve response times. In parallel, we improved Kafka retry handling, including a new service to store failed messages. This has enabled reinjection and reprocessing through a queryable endpoint.
An automated end-to-end testing framework replaced manual testing in pre-production, improving reliability and deployment speed. We also created a real-time analytics dashboard to surface the company’s true fraud cost and support better decision-making.
2. Delivering fraud-specific products and tools
Alongside the architectural improvements, we also delivered key features to improve fraud detection workflows. For scam-specific fraud, we developed a custom Scam Review Tool for fraud analysts. This incorporated early detection logic to flag suspicious orders, improving analysts’ ability to act on likely fraud.
We also created an account association model that clusters seemingly unrelated accounts based on shared characteristics such as names, email addresses, IPs, and payment methods. This allowed the client to identify organised fraud networks. In its first release, the model flagged over 33,000 fraudulent accounts, equating to $13M in potential losses.
Results
Big savings, faster detection, stronger trust
This work led to some notable financial and operational benefits for the company:
- Better visibility of B2B customers helped teams adjust fraud checking levels to reduce false positive order cancellations and chargebacks. This saved the company $7M annually.
- Operational improvements reduced the fraud team’s custom metric footprint by 50%, saving $30K per year. The new analytics dashboard helped reduce query time from 12.5 seconds to less than two.
- The automation of the scam review process helped reduce manual review times by 25%, increasing scam prevention rates by ~15% and cutting monthly costs by $200K.
- Optimising infrastructure reduced the team’s reliance on third party tech, bringing a reduction of over 90% in licensing costs for fraud intelligence tooling. This saved the company up to $160K per month.
The project also increased internal alignment, enabling fraud prevention initiatives to scale effectively within product teams.
“Throughout the engagement with Equal Experts, I have been consistently impressed with their expertise, the way they own problems from beginning to end, and also their collaboration abilities. From day one, they felt like core members of the team, and I was very pleased with how they followed our coding patterns (pair programming, TDD), didn’t cut corners to get a project out the door, and always were building code and services with the future use cases in mind”.
Conclusion
By rearchitecting fraud detection systems and embedding intelligence-led solutions, we helped the marketplace transform its ability to manage fraud at scale, without compromising customer experience. This collaboration proved that a cloud-native, data-driven approach is essential for ecommerce companies looking to stay ahead of increasingly organised and agile fraud networks.