arrow left facebook twitter linkedin medium menu play circle
December 8, 2020 - Steve Knopf

Building a Better Rules Engine: The DataVisor Way

For the past 15 years, I’ve been working in the consumer-to-consumer E-commerce industry, specifically in the areas of trust and safety and fraud detection and prevention. In all that time, one thing has been a constant – the need for a real-time operations platform with a robust rules engine that can deploy rules and actions quickly based on activity that is either not caught by models or is a long-term concern for the business.

When You Need an Operations Platform with a Robust Rules Engine

A rules engine may not be the first piece of software you think of when starting an E-commerce business, but from my experience, it is essential. There aren’t many companies out there developing rules engines. Many of the companies that are have put a lot of time and effort into building rules engines for large sectors such as the insurance industry, but they are often difficult to integrate and maintain, and they can be very expensive.

For those and other reasons, sometimes the companies for which I worked would consider building rules engines internally, but we frequently did not have the engineering expertise to build them or the available resources. We needed to maintain our focus on the business at hand of helping our customers connect with other people to sell their products and not divert resources to building a product from scratch.

That’s why DataVisor’s rules engine and overall product suite was a perfect match for our E-commerce business. When I first spoke with DataVisor at my previous company, we were  looking for a machine learning solution to help stop mass-scale fraud and spam on the platform.

But we also mentioned to DataVisor that we were looking for a rules engine. It turned out to be the perfect timing and the right fit. DataVisor was working on a rules engine solution, and when I asked if we could help with the design and feature set, DataVisor agreed.

DataVisor Meets the Two-Pronged Challenge of Building a Better Rules Engine

There were two key requirements we had for a rules engine for which I had never found an appropriate solution prior to DataVisor. The first requirement was the ability to integrate new data feeds as data elements in the rules engine. The second was complex event processing. I’ll describe both in a little more detail and give some examples to clarify exactly what I mean.

The integration challenge: A lot of teams and companies view rules engines as something they’ll build and deliver, but when it comes to supporting the rules engine over time, there’s a problem. A rules engine is a constantly growing and evolving product. It requires new data feeds, new actions, new clients, new functionality, workflows, etc. As a result, it was often difficult and expensive to get resources to support the rules engine over the long term. 

The ability to integrate new data feeds as data elements in the rules engine was difficult from a resourcing  and implementation standpoint. When the core product captured new data as a result of a newly launched feature, ideally, we wanted that data in the rules engine to write rules against. We’d often have to wait weeks or months to get those data feeds because supporting the rules engine wasn’t on any team’s roadmap. 

At DataVisor, on the other hand, there’s a team dedicated to supporting and helping with this. DataVisor also provides a supporting product called Feature Platform. Feature Platform enables the creation of new features that can be used in rules. So with the combination of Feature Platform and a dedicated team of DataVisor engineers and product managers, the ability to create new features and gain access to new data is literally just hours to days away.

The challenge of complex event processing: The second requirement, complex event processing, is an essential but often expensive feature of a rules engine. As the name suggests, it’s a complex process of calculating certain variables, and it can be expensive from a processing standpoint and difficult to implement from a real-time standpoint. 

For instance, suppose you want to know the sum of activity of a user or event, rather than simply the value of that variable at a specific time. For example, if I want to know the value of a listing at the time it’s posted, that’s easy. If I want to know the value of all that user’s listings currently or historically, however, that’s much more complicated. It takes complex event processing to find the answer to that query quickly. Luckily, DataVisor’s rules engine is architected in such a way that this processing is fairly simple and straightforward, working seamlessly within the rules.

DataVisor: Checking All the Boxes for Better Fraud Detection

So, two of my “must have” requirements were met with DataVisor, which made selecting DataVisor easy from a rules engine perspective. Additionally, DataVisor had state-of-the-art unsupervised machine learning technology for fraud detection and large-scale fraud attacks. DataVisor proved to be the best technology and value for our money and a decision I won’t regret. I believe in DataVisor’s mission and technology so much, I came to work here.

If you want to see exactly what DataVisor can do, now you can. Watch DataVisor fight fraud. Join our bi-weekly demo.

Photo of Steve Knopf
about Steve Knopf
Steve has over two decades of experience in Product Management and Trust & Safety business leadership and strategy. He previously held leadership roles at eBay, Letgo and OfferUp, helping to protect millions of users from fraud and other negative experiences. Prior to that, he was a Program Manager at Microsoft.
Photo of Steve Knopf
about Steve Knopf
Steve has over two decades of experience in Product Management and Trust & Safety business leadership and strategy. He previously held leadership roles at eBay, Letgo and OfferUp, helping to protect millions of users from fraud and other negative experiences. Prior to that, he was a Program Manager at Microsoft.