arrow left facebook twitter linkedin medium menu play circle
May 14, 2019 - Claire Zhou

Post-Modeling: Your One-Stop Solution for Solving Fraud Problems in Real Time

DataVisor Quick Take

Putting advanced analytics, robust rule management and optimization, and efficient case management to work to defeat known and unknown fraud.

In this post, the third installment in our three-part series on fraud modeling, we’re going to look at what happens after models are deployed. This is in many ways the most important and far-reaching of our three stages of modeling, as this is where the real work for detection and preventing fraud takes place. It’s important to understand that merely having models running is no guarantee of success.

Once your model is deployed, you begin receiving results, but will you understand them, and know how to act on them? How much of a drain on resources and time will manual review be, and what kind of impact will this have on operational costs? Successful fraud prevention means managing so many components—modeling tools, rules engines, machine learning engines, decision engines, case management tools, and more. Each has a purpose, each contributes to the whole, and each is uniquely challenging to integrate and leverage.

What is needed, is a one-stop solution for comprehensive fraud management. This is why we built dCube, to give organizations the power, tools, and technologies they need to solve the entire spectrum of fraud management challenges within a single platform.

Advanced Analytics

dCube’s advanced analytics capabilities enable organizations to uncover new and emerging fraud patterns, fingerprint existing and new techniques, and monitor risk levels—all in real time.

Surfacing fraud patterns and trends
Using dCube, you can discover new attack patterns, and identify the techniques fraudsters are using to stage attacks against your business and your customers. You’ll have the ability to reveal the origins of different attacks while gaining insight into common malicious behaviors. Through being able to see changes in the number of fraudulent clusters or accounts over time, you’re able to monitor risk levels in real time.

Post-Modeling 1

Utilizing “geo distribution review,” you can create location-based strategies and draw on information of this type for localization measures.

Post-Modeling 2

Uncovering connections within fraud rings
Uncover entire groups of fraudulent accounts by surfacing the hidden links among accounts with correlated attributes. These accounts are clustered together based on strong feature similarities such as profile information, digital fingerprints, account behaviors, and more.

Post-Modeling 3

Leveraging explainable and actionable reason codes

Explore reason codes that highlight fraud rings and organized attacks. Review activity across accounts and receive detailed summaries of why a given account was detected. These human-understandable reasons will indicate whether a detected account is closely related to a group of suspicious accounts or if there are any anomalies in the account behavior.

Robust Rules Management and Optimization

The DataVisor Rules Engine provides advanced rule management and rule performance optimization, and the Feature Platform makes it possible to engineer complex features for advanced rule creation. With these capabilities, you can manage complex rules at scale with maximum flexibility and a streamlined workflow, and make ensembled decisions together with machine learning engines.

Post-Modeling 4

You’ll no longer need to engineer features specifically for all kinds of rules. Instead, you can directly leverage hundreds of pre-built features or engineer custom ones with the Feature Platform, and seamlessly use them to create advanced rules in a timely manner.

Post-Modeling 5

You’ll be able to deploy or pause rules in production, and organize rules using tags or rule sets. You can take a holistic view of the different rules in the system and see summary stats and history, including rule creator, deployment history, and more.

Additionally, you can track rule performance and detected accounts over time, test rules against historical data or with live data, validate detection quality, and analyze detection overlap with detection models.


Efficient Case Management

DataVisor Case Management empowers organizations to boost review efficiency by up to 100x through bulk action and seamless team collaboration, and make more informed decisions for cases that require additional review with access to complete histories.

The system provides a centralized hub for reviewers to efficiently assign and receive tasks, prioritize what to review, and analyze reason codes to accurately identify malicious activities.

Post-Modeling 7

You can review detailed detection reasons, user profiles, and suspicious patterns, and benefit from investigating correlated accounts all together at the group level. Teams are empowered to make accurate bulk decisions that apply for the entire fraud ring, and significantly improve review efficiency with more cases reviewed in less time.

Know the Unknown

The hallmarks of a successful fraud management strategy are speed, agility, and efficiency. The kind of real-time proactivity necessary to stay ahead of rapidly-evolving known and unknown fraud can only be realized through a sophisticated combination of hyper-modern tools, technologies, and solutions. Yet even with all this in place, success can be eroded by cumbersome processes, and the need to continuously juggle disparate system components.

It’s time to bid farewell to legacy solutions that lose ground every day to increasingly fleet fraudsters capable of launching massive attacks at scale and getting away clean before alarms are even sounded.

dCube introduces an era of seamless integration within a single platform, combines the best of automated and manual processes to power new levels of accuracy and performance, and enables the building and deployment of models so sophisticated that not even the cleverest of fraudsters stands a chance.


We’ve now covered pre-modelingmodeling, and—with this post—post-modeling. Please make sure to read all three posts, and then contact us to schedule a demo of dCube!

about Claire Zhou
Claire is a Senior Product Marketing Manager at DataVisor with over 5 years of marketing experience in security and fin-tech. She is passionate about empowering enterprise customers with AI-based solutions. Her expertise spans data analytics, cybersecurity, and fraud prevention. Claire has an MBA from UCLA.
about Claire Zhou
Claire is a Senior Product Marketing Manager at DataVisor with over 5 years of marketing experience in security and fin-tech. She is passionate about empowering enterprise customers with AI-based solutions. Her expertise spans data analytics, cybersecurity, and fraud prevention. Claire has an MBA from UCLA.