New Ebook: How to Choose a Machine Learning Platform for Risk
Latest posts by Paul Bookstaber (see all)
- A New Timescale for Fraud Science: Insights from Our CSO - October 17, 2017
- Operationalizing Machine Learning for Fraud: Human + Machine Learning is the Best of Both - October 6, 2017
- Easing Into Machine Learning for PSD2: An Interview with Our VP of French Sales - September 13, 2017
If you’re a merchant, an acquirer, or a payment service provider, you might be searching for a unified strategy that can manage risk without creating customer friction. We can help you procure the right solution. The Feedzai platform powers 1 in 5 of the world’s top 25 banks, and we believe it can power your business, too. So if you’re looking for a machine learning platform, this guide is a good place to start.
The fact is this: poor fraud detection hurts your top-line revenue. It hurts to deny a legitimate customer, to incur chargebacks, and to have manual review bottlenecks that slow down your order fulfillments.
Good thing we’re witnessing the dawn of attainable AI. It’s introducing us to new systems that any business can access, identifying patterns that were invisible before, and turning big data into big insights. But not all machine learning platforms are created equal.
For example, some machine learning systems make their decisions inside a black box, which means humans can’t understand the whys behind their decisions, or leave an audit trail of documentation for regulators. Compare that to a platform that does whitebox processing, with a human-readable semantic layer on top of its underlying machine logic. The whitebox innovation comes to us thanks to the proprietary Feedzai Random Forest algorithm, which is made of tens of thousands of decision trees. A whitebox system weighs the top-most factors from this ensemble of decision trees, and communicates them to the human in a simple way.
For another example, a large, growing business should consider whether the platform is a picky eater. Can the platform ingest all data from all sources, no matter the channel or use case? Many platforms can only take in data from certain sources, for instance, point-of-sale data but not mobile devices. These platforms take an incomplete view of the customer. Compare that to a platform with omnidata and omnichannel capabilities, taking in every kind of data, whether it’s coming from an internal system, like a CRM system, or external data sources, like enrichers.
It’s also important to consider how your employees are interacting with the platform. Organizations can have many hundreds of people logging into the same system. Businesses at these large scales need a different kind of machine learning platform. Consider a global ecommerce seller with loss prevention teams, risk analysts, and fulfillment shipping centers, all operating in different timezones, 24/7. A case management tool needs to have self-configurable workflows so that everyone in this diverse user base can see relevant insights at the right times. Do you have a global case management tool that automatically distributes work based on different criteria, and prioritizes the most important orders to review?
Download the report below for a full checklist of things to consider when choosing a machine learning platform for risk.
Download the Feedzai report on How to Choose a Machine Learning Platform for Risk