Why Marketplaces Should Think Twice Before Hiring A Machine Learning Team
Users of marketplace platforms such as Uber, Kickstarter and Airbnb stay on the platform because they trust the platform. These marketplaces are also targets for fraudsters whose actions erode trust and cause user attrition that is difficult to reverse.
To maintain the user base, marketplaces like Airbnb have purposefully designed trust systems as fundamental components of their technology stack and operating infrastructure. They’ve created dedicated teams and systems to design, engineer and build trust into their platform.
To engineer in this trust, it requires Airbnb to think carefully about every pixel on the page and every step in the customer journey. For instance, after seeing data showing that having host profile pictures displayed on room listing drove guest bookings, they made profile pictures mandatory for each and every host member. The key step here is to first establish identity and thereafter trust.
But they didn’t stop there; they built a safety net for their guests and hosts, such as a 24/7 customer support operation that spans across multiple geographies, as well as a $1 million guarantee for hosts in case anything goes wrong during a stay. On top of this, they added a recommendation system that lets both hosts and guests rate their experiences. This system retains feedback history and fosters a sense of community that self-polices behavior. These interlocking components work together to give Airbnb complete end-to-end control of who joins their platform and to keep bad actors out.
Other marketplace-based businesses like Uber and Amazon have embedded their own version of trust into their products. They have all succeeded because a system of trust based on machine learning is engineered into their platform. These machine learning systems monitor, learn and react to situations in real-time in order to protect the platform.
Why You Don’t Need To Start From Scratch
It used to be that these machine learning risk systems were expensive and had to be engineered from the ground up by internal teams. However, the quantum leap in computing power, cheap memory storage, bandwidth and big data technology within the last few years — has now made ready-made machine learning systems available.
Just as a company today would not consider building an in-house CRM solution but instead buy Salesforce, marketplace platform providers can buy machine learning risk systems that are affordable, easy-to-use, big data-capable and real-time:
- Affordable: Using Hadoop, Cassandra and open-source technology, today’s machine learning systems can be deployed on-premise using more affordable commodity hardware or accessed via the cloud. For marketplace platforms that are starting up and testing out new business models or market traction, this affordability and flexibility to drop in a machine learning function helps them get to market faster.
- Easy-to-use: Advances in machine learning algorithms such as Random Forest and Deep Learning now are integrated with a semantic layer that outputs clear, human-readable text to demystify the underlying machine logic and explain the machine decisions. This transparency is vital in order to provide customer service excellence to the community of users on marketplace platforms.
- Big data capable: 90% of the world’s data has been created in the last few years and modern machine learning systems are built for this “expanding dataverse”. The ability to stream in additional tera- or petabytes of data into machine learning systems enables marketplace platforms to develop complete, 360 degree “uber-views” of their user base activity.
- Real-time: Along with big data, handling “fast data” is also a requirement for today’s marketplace platforms because they serve users that are always connected, use more mobile, and that expect 1-click simplicity. As such, new machine learning systems can make decisions that evaluate millions of variables in under 20 milliseconds (it takes 250 milliseconds to blink).
These new capabilities enable companies to maintain command and control of their customer experience. There is no substitute for trust, and to keep and serve a healthy community of customers, marketplace platforms cannot outsource their risk decisions. They need to have systems internally that are engineered for building trust.
Fortunately, the advances in machine learning systems in recent years now enables any business that connects with users or buyers to procure machine learning risk systems that are drop-in ready, just like buying an off-the-shelf CRM system. At Feedzai we already power numerous such platforms, including a top taxi hailing service, a luxury retailer, a money transfer service and more. Contact us to learn more about how we can help you.
Latest posts by Ken Bui (see all)
- EU Fifth AML Directive: How Banks Can Prepare for Five Key Changes - November 18, 2019
- AI Best Practices to Improve Enterprise Risk Outcomes - October 22, 2019
- Why It’s So Hard for Challenger Banks to Fight Financial Crime - September 24, 2019
Subscribe to stay infomed