Amazon Fraud Detector customers

At AWS, we use Amazon Fraud Detector to protect our customers and our own business. Amazon Fraud Detector is particularly useful when we enter new markets or geographies, since the enrichment of inputs like IP Address and Email Address provides an immediate boost even when we do not have robust historical data. By using Amazon Fraud Detector’s automated Online Fraud Insights model, we started capturing 35% more fraudulent account signups in new geographies. Moreover, Amazon Fraud Detector allows us to train models for new parts of our business in half a day – the process previously took at least a week.

Twelve months ago, we were looking for a fraud detection solution that didn’t require us to invest heavily in our own machine learning expertise. With transparent pay-as-you-use pricing, Amazon Fraud Detector helped us to easily create and incorporate an effective and affordable new machine learning model into our existing setup.

Since introducing Amazon Fraud Detector, our abort rate has dropped below 2% (vs. 5% previously). Additionally, our chargeback rate is the lowest it’s ever been since the company’s inception. The business can now accept more checkouts that our past models would have flagged as risky and turned away. But perhaps the best thing is we’re getting these great results with roughly the same operational costs as before. All of this results in an increased number of bookings and revenues, along with a decrease in losses due to chargebacks

With Amazon Fraud Detector, we reduced fraudulent transactions by 6%. At the same time, we’ve been able to automate checkout fulfillment on more than 90% of the transactions that would have previously been flagged for manual review. Now, we’re manually reviewing less than 1% of our transactions—down from 10%.

Since we implemented this service, we’ve seen a significant improvement in our Trustpilot score, and we know it’s a result of this checkout detection automation, as well as additional enhancements we are consistently making on the website. Trust is a big part of our value to customers, so that’s a huge win for our business.

Amazon Fraud Detector has been a great addition to our Fraud detection and mitigation capability. The ability to write custom rules that apply to our unique situation, train ML models on-demand, and seamless integration with other AWS services has enabled us to make decisions quickly and intelligently while retaining complete control of the platform. AWS was very helpful during the proof of concept stage and has been adding new features to the platform inline with Fraud trends.

The rules-based fraud detection approach we developed for our web design platform has been successful at stopping bad actors, but as our business has grown we recognize the need to continuously improve. Some of the approaches we tried were preventing fraud but came at a cost to our customer experience. We started investigating machine learning (ML) based approaches that would help us balance improving fraud prevention with maintaining a painless customer experience that we could get up and running quickly. Amazon Fraud Detector is exactly what we needed: a cloud-based fraud detection service that would allow us to easily create customized ML models and integrate with our existing approach. We found the process of building and deploying a ML-based fraud “detector” was easy and could be accomplished with our existing resources. With Amazon Fraud Detector we have been able to further improve accuracy and seen a double digit drop in false negatives, enabling us to catch more bad actors. In addition, we have saved countless hours of effort that would have been required to build an ML based approach from scratch and integrate it with our existing solution. We see lots of potential to expand our use of Amazon Fraud Detector to help us provide the secure, painless web design experience that our customers expect.

GoDaddy is committed to preventing fraudulent accounts and we’re continually bolstering our capabilities to automatically detect such accounts during sign-up. We recently began using Amazon Fraud Detector and we’re pleased that it offers low cost of implementation and a self-service approach to building a machine learning model that is customized to our business. The model can be easily deployed and used in our new account process without impacting the signup experience for legitimate customers. The model we built with Amazon Fraud Detector is able to detect likely fraudulent sign-ups immediately, so we’re very pleased with the results and look forward to accomplishing more.

In Q1/Q2 2020 we experienced a spike in accounts being used for phishing attacks. As a result, we needed to supplement our existing homegrown solution with stronger transaction data and signals to identify bad actors sooner. A scalable solution based on predictive machine-learning was important to us as a growing business ourselves. Amazon Fraud Detector made it easy to build a model using our own data that accurately identifies account signups that result in phishing attacks. More importantly, we were able to get these results with a very low false positive rate, which means no additional work for our operations staff. Amazon Fraud Detector has a competitive pricing model and we can easily integrate the model into our existing workflow.

Amazon Fraud Detector has enabled us to drastically improve operations, increase our flexibility to respond to bad actors, and have greater control of systems and processes. Initially, we were exploring an in-house and 3rd party solution. When Amazon Fraud Detector was announced, we immediately changed course. We have been an AWS customer for many years and have great trust in Amazon’s products. With Amazon Fraud Detector, we are no longer bound by the conventional limitations of on-premise or SaaS offerings. Instead, we have the flexibility to adapt a Machine Learning powered service to meet our needs and the ability to use AWS’s rules-only option while easily scaling to full Machine Learning capabilities when needed. This saved Truevo 3-6 months in development! In fact, we deployed our first prototype model within 30 minutes.

Overall, we are operating with greater confidence in our ability to detect fraud in real-time. We are better equipped to deploy rule detections when we notice odd activity that we may not fully understand, but need to stop. We are able to respond and adapt to ever-changing regulatory and scheme requirements allowing us to stay on top of our game.

At Clearly we are constantly looking for ways to improve our unparalleled customer service while also better protecting our business. Our previous fraud detection solution involved reactively hard-coding rules based on past fraud attempts and manually investigating all suspicious transactions. This resulted in us missing a large amount of fraud, delaying customers’ orders due to manual reviews, and driving up our fulfillment costs. We were hesitant to explore an ML solution since we needed to move quickly and didn’t have a tenured ML team, but with Amazon Fraud Detector we were able to deploy an accurate and reliable fraud prevention solution in just a few weeks. Now, we capture more fraud and send fewer orders to manual review, saving us thousands of dollars per week.

After implementing this fraud detection solution, the ICONY support team saw the time they spent dealing with fake and spam accounts fall by 77%. This freed up the team to deal with individual user checks, which immediately improved quality on the platform and caused fraud reports from the community to drop by 63%. Moreover, the number of registered users returning to the platform has increased 4.13%. With less harassment from fake accounts and scammers, users feel more comfortable on the platform and enjoy using it.

By implementing Amazon Fraud Detector into the WuzzTrack system, Wuzzon now has a much more robust and reliable fraud detection solution, which can also detect more novel fraud techniques. The implementation was quick and easy and the results were even better than initially hoped. In some extreme cases there was a decrease in false positives of up to 43% (when compared to the previous rule-based solutions), while for other sources, the true positive rate increased by 11-14%.

Since we launched Amazon Fraud Detector into production, our results have been great. Approximately 94% of our claims are typically rated as low risk and for these claims, the turnaround time has already reduced from 48 hours in February with our traditional manual process to less than 6 hours by the end of August. This has resulted in improved customer experience. We have also seen a 36% increase in our Net Promoter Survey (NPS) scores between February 2022 to August 2022 since going live in production. We attribute this to the faster payouts due to safely automating approvals for low-risk claims. For the 6% of claims rated as high risk, we now have more capacity to interrogate these better than before. As a result, we are able to thoroughly investigate suspected cases and stop more claims with actual fraud. Overall our confirmed fraudulent cases that we have been able to identify before paying out a claim has increased over 100%, which has greatly reduced the business' exposure to risk.